201
|
Abstract
We show that material plasticity can be precisely and efficiently predicted by deep-learning methods. This approach is fundamentally different from the century-old theory of continuum plasticity because it is not iteratively tracing the yield surface, neither does it require the notion of effective strain or stress at the macroscopic level. Instead, we use representative computer simulations of materials, including microstructure and constituents, load them along different deformation paths, and then learn the reversible, irreversible, and history-dependent phenomena directly from data. We demonstrate that complex phenomena such as distortional hardening can be predicted within 0.5% error. The generality of the methodology and widespread importance of plasticity in designing structures and materials make it useful to a myriad of fields. Plasticity theory aims at describing the yield loci and work hardening of a material under general deformation states. Most of its complexity arises from the nontrivial dependence of the yield loci on the complete strain history of a material and its microstructure. This motivated 3 ingenious simplifications that underpinned a century of developments in this field: 1) yield criteria describing yield loci location; 2) associative or nonassociative flow rules defining the direction of plastic flow; and 3) effective stress–strain laws consistent with the plastic work equivalence principle. However, 2 key complications arise from these simplifications. First, finding equations that describe these 3 assumptions for materials with complex microstructures is not trivial. Second, yield surface evolution needs to be traced iteratively, i.e., through a return mapping algorithm. Here, we show that these assumptions are not needed in the context of sequence learning when using recurrent neural networks, diverting the above-mentioned complications. This work offers an alternative to currently established plasticity formulations by providing the foundations for finding history- and microstructure-dependent constitutive models through deep learning.
Collapse
|
202
|
Kao JC. Considerations in using recurrent neural networks to probe neural dynamics. J Neurophysiol 2019; 122:2504-2521. [PMID: 31619125 DOI: 10.1152/jn.00467.2018] [Citation(s) in RCA: 13] [Impact Index Per Article: 2.6] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/22/2022] Open
Abstract
Recurrent neural networks (RNNs) are increasingly being used to model complex cognitive and motor tasks performed by behaving animals. RNNs are trained to reproduce animal behavior while also capturing key statistics of empirically recorded neural activity. In this manner, the RNN can be viewed as an in silico circuit whose computational elements share similar motifs with the cortical area it is modeling. Furthermore, because the RNN's governing equations and parameters are fully known, they can be analyzed to propose hypotheses for how neural populations compute. In this context, we present important considerations when using RNNs to model motor behavior in a delayed reach task. First, by varying the network's nonlinear activation and rate regularization, we show that RNNs reproducing single-neuron firing rate motifs may not adequately capture important population motifs. Second, we find that even when RNNs reproduce key neurophysiological features on both the single neuron and population levels, they can do so through distinctly different dynamical mechanisms. To distinguish between these mechanisms, we show that an RNN consistent with a previously proposed dynamical mechanism is more robust to input noise. Finally, we show that these dynamics are sufficient for the RNN to generalize to tasks it was not trained on. Together, these results emphasize important considerations when using RNN models to probe neural dynamics.NEW & NOTEWORTHY Artificial neurons in a recurrent neural network (RNN) may resemble empirical single-unit activity but not adequately capture important features on the neural population level. Dynamics of RNNs can be visualized in low-dimensional projections to provide insight into the RNN's dynamical mechanism. RNNs trained in different ways may reproduce neurophysiological motifs but do so with distinctly different mechanisms. RNNs trained to only perform a delayed reach task can generalize to perform tasks where the target is switched or the target location is changed.
Collapse
|
203
|
Heeger DJ, Mackey WE. Oscillatory recurrent gated neural integrator circuits (ORGaNICs), a unifying theoretical framework for neural dynamics. Proc Natl Acad Sci U S A 2019; 116:22783-22794. [PMID: 31636212 PMCID: PMC6842604 DOI: 10.1073/pnas.1911633116] [Citation(s) in RCA: 15] [Impact Index Per Article: 3.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/18/2022] Open
Abstract
Working memory is an example of a cognitive and neural process that is not static but evolves dynamically with changing sensory inputs; another example is motor preparation and execution. We introduce a theoretical framework for neural dynamics, based on oscillatory recurrent gated neural integrator circuits (ORGaNICs), and apply it to simulate key phenomena of working memory and motor control. The model circuits simulate neural activity with complex dynamics, including sequential activity and traveling waves of activity, that manipulate (as well as maintain) information during working memory. The same circuits convert spatial patterns of premotor activity to temporal profiles of motor control activity and manipulate (e.g., time warp) the dynamics. Derivative-like recurrent connectivity, in particular, serves to manipulate and update internal models, an essential feature of working memory and motor execution. In addition, these circuits incorporate recurrent normalization, to ensure stability over time and robustness with respect to perturbations of synaptic weights.
Collapse
|
204
|
Wu J, Wang W, Zhang J, Zhou B, Zhao W, Su Z, Gu X, Wu J, Zhou Z, Chen S. DeepHLApan: A Deep Learning Approach for Neoantigen Prediction Considering Both HLA-Peptide Binding and Immunogenicity. Front Immunol 2019; 10:2559. [PMID: 31736974 PMCID: PMC6838785 DOI: 10.3389/fimmu.2019.02559] [Citation(s) in RCA: 58] [Impact Index Per Article: 11.6] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/31/2019] [Accepted: 10/15/2019] [Indexed: 12/30/2022] Open
Abstract
Neoantigens play important roles in cancer immunotherapy. Current methods used for neoantigen prediction focus on the binding between human leukocyte antigens (HLAs) and peptides, which is insufficient for high-confidence neoantigen prediction. In this study, we apply deep learning techniques to predict neoantigens considering both the possibility of HLA-peptide binding (binding model) and the potential immunogenicity (immunogenicity model) of the peptide-HLA complex (pHLA). The binding model achieves comparable performance with other well-acknowledged tools on the latest Immune Epitope Database (IEDB) benchmark datasets and an independent mass spectrometry (MS) dataset. The immunogenicity model could significantly improve the prediction precision of neoantigens. The further application of our method to the mutations with pre-existing T-cell responses indicating its feasibility in clinical application. DeepHLApan is freely available at https://github.com/jiujiezz/deephlapan and http://biopharm.zju.edu.cn/deephlapan.
Collapse
|
205
|
Segura-Bedmar I, Raez P. Cohort selection for clinical trials using deep learning models. J Am Med Inform Assoc 2019; 26:1181-1188. [PMID: 31532478 PMCID: PMC6798560 DOI: 10.1093/jamia/ocz139] [Citation(s) in RCA: 12] [Impact Index Per Article: 2.4] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/15/2019] [Revised: 07/10/2019] [Accepted: 07/22/2019] [Indexed: 01/09/2023] Open
Abstract
OBJECTIVE The goal of the 2018 n2c2 shared task on cohort selection for clinical trials (track 1) is to identify which patients meet the selection criteria for clinical trials. Cohort selection is a particularly demanding task to which natural language processing and deep learning can make a valuable contribution. Our goal is to evaluate several deep learning architectures to deal with this task. MATERIALS AND METHODS Cohort selection can be formulated as a multilabeling problem whose goal is to determine which criteria are met for each patient record. We explore several deep learning architectures such as a simple convolutional neural network (CNN), a deep CNN, a recurrent neural network (RNN), and CNN-RNN hybrid architecture. Although our architectures are similar to those proposed in existing deep learning systems for text classification, our research also studies the impact of using a fully connected feedforward layer on the performance of these architectures. RESULTS The RNN and hybrid models provide the best results, though without statistical significance. The use of the fully connected feedforward layer improves the results for all the architectures, except for the hybrid architecture. CONCLUSIONS Despite the limited size of the dataset, deep learning methods show promising results in learning useful features for the task of cohort selection. Therefore, they can be used as a previous filter for cohort selection for any clinical trial with a minimum of human intervention, thus reducing the cost and time of clinical trials significantly.
Collapse
|
206
|
Ontology-Based Healthcare Named Entity Recognition from Twitter Messages Using a Recurrent Neural Network Approach. INTERNATIONAL JOURNAL OF ENVIRONMENTAL RESEARCH AND PUBLIC HEALTH 2019; 16:ijerph16193628. [PMID: 31569654 PMCID: PMC6801946 DOI: 10.3390/ijerph16193628] [Citation(s) in RCA: 17] [Impact Index Per Article: 3.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 08/25/2019] [Revised: 09/12/2019] [Accepted: 09/24/2019] [Indexed: 01/04/2023]
Abstract
Named Entity Recognition (NER) in the healthcare domain involves identifying and categorizing disease, drugs, and symptoms for biosurveillance, extracting their related properties and activities, and identifying adverse drug events appearing in texts. These tasks are important challenges in healthcare. Analyzing user messages in social media networks such as Twitter can provide opportunities to detect and manage public health events. Twitter provides a broad range of short messages that contain interesting information for information extraction. In this paper, we present a Health-Related Named Entity Recognition (HNER) task using healthcare-domain ontology that can recognize health-related entities from large numbers of user messages from Twitter. For this task, we employ a deep learning architecture which is based on a recurrent neural network (RNN) with little feature engineering. To achieve our goal, we collected a large number of Twitter messages containing health-related information, and detected biomedical entities from the Unified Medical Language System (UMLS). A bidirectional long short-term memory (BiLSTM) model learned rich context information, and a convolutional neural network (CNN) was used to produce character-level features. The conditional random field (CRF) model predicted a sequence of labels that corresponded to a sequence of inputs, and the Viterbi algorithm was used to detect health-related entities from Twitter messages. We provide comprehensive results giving valuable insights for identifying medical entities in Twitter for various applications. The BiLSTM-CRF model achieved a precision of 93.99%, recall of 73.31%, and F1-score of 81.77% for disease or syndrome HNER; a precision of 90.83%, recall of 81.98%, and F1-score of 87.52% for sign or symptom HNER; and a precision of 94.85%, recall of 73.47%, and F1-score of 84.51% for pharmacologic substance named entities. The ontology-based manual annotation results show that it is possible to perform high-quality annotation despite the complexity of medical terminology and the lack of context in tweets.
Collapse
|
207
|
Han BK, Ryu JK, Kim SC. Context-Aware Winter Sports Based on Multivariate Sequence Learning. SENSORS 2019; 19:s19153296. [PMID: 31357531 PMCID: PMC6696288 DOI: 10.3390/s19153296] [Citation(s) in RCA: 8] [Impact Index Per Article: 1.6] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 06/15/2019] [Revised: 07/22/2019] [Accepted: 07/23/2019] [Indexed: 11/16/2022]
Abstract
In this paper, we present an intelligent system that is capable of estimating the status of a player engaging in winter activities based on the sequence analysis of multivariate time-series sensor signals. Among the winter activities, this paper mainly focuses on downhill winter sports such as alpine skiing and snowboarding. Assuming that the mechanical vibrations generated by physical interaction between the ground surface and ski/snowboard in motion can describe the ground conditions and playing contexts, we utilize inertial and vibration signals to categorize the motion context. For example, the proposed system estimates whether the player is sitting on a ski lift or standing on the escalator, or skiing on wet or snowy ground, etc. To measure the movement of a player during a game or on the move, we develop a custom embedded system comprising a motion sensor and piezo transducer. The captured multivariate sequence signals are then trained in a supervised fashion. We adopt artificial neural network approaches (e.g., 1D convolutional neural network, and gated recurrent neural networks, such as long short-term memory and gated recurrent units). The experimental results validate the feasibility of the proposed approach.
Collapse
|
208
|
A Low-Delay Lightweight Recurrent Neural Network (LLRNN) for Rotating Machinery Fault Diagnosis. SENSORS 2019; 19:s19143109. [PMID: 31337108 PMCID: PMC6679287 DOI: 10.3390/s19143109] [Citation(s) in RCA: 19] [Impact Index Per Article: 3.8] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 05/15/2019] [Revised: 06/18/2019] [Accepted: 07/05/2019] [Indexed: 11/17/2022]
Abstract
Fault diagnosis is critical to ensuring the safety and reliable operation of rotating machinery systems. Long short-term memory networks (LSTM) have received a great deal of attention in this field. Most of the LSTM-based fault diagnosis methods have too many parameters and calculation, resulting in large memory occupancy and high calculation delay. Thus, this paper proposes a low-delay lightweight recurrent neural network (LLRNN) model for mechanical fault diagnosis, based on a special LSTM cell structure with a forget gate. The input vibration signal is segmented into several shorter sub-signals in order to shorten the length of the time sequence. Then, these sub-signals are sent into the network directly and converted into the final diagnostic results without any manual participation. Compared with some existing methods, our experiments illustrate that the proposed method has less memory space occupancy and lower computational delay while maintaining the same level of accuracy.
Collapse
|
209
|
Zhou X, Xu Z, Li S. Collision-Free Compliance Control for Redundant Manipulators: An Optimization Case. Front Neurorobot 2019; 13:50. [PMID: 31396070 PMCID: PMC6662470 DOI: 10.3389/fnbot.2019.00050] [Citation(s) in RCA: 5] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/30/2019] [Accepted: 06/24/2019] [Indexed: 11/24/2022] Open
Abstract
Force control of manipulators could enhance compliance and execution capabilities, and has become a key issue in the field of robotic control. However, it is challenging for redundant manipulators, especially when there exist risks of collisions. In this paper, we propose a collision-free compliance control strategy based on recurrent neural networks. Inspired by impedance control, the position-force control task is rebuilt as a reference command of task-space velocities, by combing kinematic properties, the compliance controller is then described as an equality constraint in joint velocity level. As to collision avoidance strategy, both robot and obstacles are approximately described as two sets of key points, and the distances between those points are used to scale the feasible workspace. In order to save unnecessary energy consumption while reducing impact of possible collisions, the secondary task is chosen to minimize joint velocities. Then a RNN with provable convergence is established to solve the constraint-optimization problem in realtime. Numerical results validate the effectiveness of the proposed controller.
Collapse
|
210
|
Xu Z, Zhou X, Li S. Deep Recurrent Neural Networks Based Obstacle Avoidance Control for Redundant Manipulators. Front Neurorobot 2019; 13:47. [PMID: 31333442 PMCID: PMC6622359 DOI: 10.3389/fnbot.2019.00047] [Citation(s) in RCA: 16] [Impact Index Per Article: 3.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 04/15/2019] [Accepted: 06/17/2019] [Indexed: 11/27/2022] Open
Abstract
Obstacle avoidance is an important subject in the control of robot manipulators, but is remains challenging for robots with redundant degrees of freedom, especially when there exist complex physical constraints. In this paper, we propose a novel controller based on deep recurrent neural networks. By abstracting robots and obstacles into critical point sets respectively, the distance between the robot and obstacles can be described in a simpler way, then the obstacle avoidance strategy is established in form of inequality constraints by general class-K functions. Using minimal-velocity-norm (MVN) scheme, the control problem is formulated as a quadratic-programming case under multiple constraints. Then a deep recurrent neural network considering system models is established to solve the QP problem online. Theoretical conduction and numerical simulations show that the controller is capable of avoiding static or dynamic obstacles, while tracking the predefined trajectories under physical constraints.
Collapse
|
211
|
Masse NY, Yang GR, Song HF, Wang XJ, Freedman DJ. Circuit mechanisms for the maintenance and manipulation of information in working memory. Nat Neurosci 2019; 22:1159-1167. [PMID: 31182866 PMCID: PMC7321806 DOI: 10.1038/s41593-019-0414-3] [Citation(s) in RCA: 91] [Impact Index Per Article: 18.2] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/09/2018] [Accepted: 04/22/2019] [Indexed: 11/09/2022]
Abstract
Recently it has been proposed that information in working memory (WM) may not always be stored in persistent neuronal activity but can be maintained in 'activity-silent' hidden states, such as synaptic efficacies endowed with short-term synaptic plasticity. To test this idea computationally, we investigated recurrent neural network models trained to perform several WM-dependent tasks, in which WM representation emerges from learning and is not a priori assumed to depend on self-sustained persistent activity. We found that short-term synaptic plasticity can support the short-term maintenance of information, provided that the memory delay period is sufficiently short. However, in tasks that require actively manipulating information, persistent activity naturally emerges from learning, and the amount of persistent activity scales with the degree of manipulation required. These results shed insight into the current debate on WM encoding and suggest that persistent activity can vary markedly between short-term memory tasks with different cognitive demands.
Collapse
|
212
|
Wang L, Li K, Chen X, Hu XP. Application of Convolutional Recurrent Neural Network for Individual Recognition Based on Resting State fMRI Data. Front Neurosci 2019; 13:434. [PMID: 31118882 PMCID: PMC6504790 DOI: 10.3389/fnins.2019.00434] [Citation(s) in RCA: 12] [Impact Index Per Article: 2.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/09/2018] [Accepted: 04/15/2019] [Indexed: 11/28/2022] Open
Abstract
In most task and resting state fMRI studies, a group consensus is often sought, where individual variability is considered a nuisance. None the less, biological variability is an important factor that cannot be ignored and is gaining more attention in the field. One recent development is the individual identification based on static functional connectome. While the original work was based on the static connectome, subsequent efforts using recurrent neural networks (RNN) demonstrated that the inclusion of temporal features greatly improved identification accuracy. Given that convolutional RNN (ConvRNN) seamlessly integrates spatial and temporal features, the present work applied ConvRNN for individual identification with resting state fMRI data. Our result demonstrates ConvRNN achieving a higher identification accuracy than conventional RNN, likely due to better extraction of local features between neighboring ROIs. Furthermore, given that each convolutional output assembles in-place features, they provide a natural way for us to visualize the informative spatial pattern and temporal information, opening up a promising new avenue for analyzing fMRI data.
Collapse
|
213
|
Wang H, Zhao S, Dong Q, Cui Y, Chen Y, Han J, Xie L, Liu T. Recognizing Brain States Using Deep Sparse Recurrent Neural Network. IEEE TRANSACTIONS ON MEDICAL IMAGING 2019; 38:1058-1068. [PMID: 30369441 PMCID: PMC6508593 DOI: 10.1109/tmi.2018.2877576] [Citation(s) in RCA: 27] [Impact Index Per Article: 5.4] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 05/29/2023]
Abstract
Brain activity is a dynamic combination of different sensory responses and thus brain activity/state is continuously changing over time. However, the brain's dynamical functional states recognition at fast time-scales in task fMRI data have been rarely explored. In this paper, we propose a novel 5-layer deep sparse recurrent neural network (DSRNN) model to accurately recognize the brain states across the whole scan session. Specifically, the DSRNN model includes an input layer, one fully-connected layer, two recurrent layers, and a softmax output layer. The proposed framework has been tested on seven task fMRI data sets of Human Connectome Project. Extensive experiment results demonstrate that the proposed DSRNN model can accurately identify the brain's state in different task fMRI data sets and significantly outperforms other auto-correlation methods or non-temporal approaches in the dynamic brain state recognition accuracy. In general, the proposed DSRNN offers a new methodology for basic neuroscience and clinical research.
Collapse
|
214
|
Song Y, Kim I. Spatio-Temporal Action Detection in Untrimmed Videos by Using Multimodal Features and Region Proposals. SENSORS 2019; 19:s19051085. [PMID: 30832433 PMCID: PMC6427216 DOI: 10.3390/s19051085] [Citation(s) in RCA: 3] [Impact Index Per Article: 0.6] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 12/10/2018] [Revised: 02/23/2019] [Accepted: 02/25/2019] [Indexed: 12/02/2022]
Abstract
This paper proposes a novel deep neural network model for solving the spatio-temporal-action-detection problem, by localizing all multiple-action regions and classifying the corresponding actions in an untrimmed video. The proposed model uses a spatio-temporal region proposal method to effectively detect multiple-action regions. First, in the temporal region proposal, anchor boxes were generated by targeting regions expected to potentially contain actions. Unlike the conventional temporal region proposal methods, the proposed method uses a complementary two-stage method to effectively detect the temporal regions of the respective actions occurring asynchronously. In addition, to detect a principal agent performing an action among the people appearing in a video, the spatial region proposal process was used. Further, coarse-level features contain comprehensive information of the whole video and have been frequently used in conventional action-detection studies. However, they cannot provide detailed information of each person performing an action in a video. In order to overcome the limitation of coarse-level features, the proposed model additionally learns fine-level features from the proposed action tubes in the video. Various experiments conducted using the LIRIS-HARL and UCF-10 datasets confirm the high performance and effectiveness of the proposed deep neural network model.
Collapse
|
215
|
Chen G, Chen J, Lienen M, Conradt J, Röhrbein F, Knoll AC. FLGR: Fixed Length Gists Representation Learning for RNN-HMM Hybrid-Based Neuromorphic Continuous Gesture Recognition. Front Neurosci 2019; 13:73. [PMID: 30809114 PMCID: PMC6380225 DOI: 10.3389/fnins.2019.00073] [Citation(s) in RCA: 7] [Impact Index Per Article: 1.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/30/2018] [Accepted: 01/23/2019] [Indexed: 11/18/2022] Open
Abstract
A neuromorphic vision sensors is a novel passive sensing modality and frameless sensors with several advantages over conventional cameras. Frame-based cameras have an average frame-rate of 30 fps, causing motion blur when capturing fast motion, e.g., hand gesture. Rather than wastefully sending entire images at a fixed frame rate, neuromorphic vision sensors only transmit the local pixel-level changes induced by the movement in a scene when they occur. This leads to advantageous characteristics, including low energy consumption, high dynamic range, a sparse event stream and low response latency. In this study, a novel representation learning method was proposed: Fixed Length Gists Representation (FLGR) learning for event-based gesture recognition. Previous methods accumulate events into video frames in a time duration (e.g., 30 ms) to make the accumulated image-level representation. However, the accumulated-frame-based representation waives the friendly event-driven paradigm of neuromorphic vision sensor. New representation are urgently needed to fill the gap in non-accumulated-frame-based representation and exploit the further capabilities of neuromorphic vision. The proposed FLGR is a sequence learned from mixture density autoencoder and preserves the nature of event-based data better. FLGR has a data format of fixed length, and it is easy to feed to sequence classifier. Moreover, an RNN-HMM hybrid was proposed to address the continuous gesture recognition problem. Recurrent neural network (RNN) was applied for FLGR sequence classification while hidden Markov model (HMM) is employed for localizing the candidate gesture and improving the result in a continuous sequence. A neuromorphic continuous hand gestures dataset (Neuro ConGD Dataset) was developed with 17 hand gestures classes for the community of the neuromorphic research. Hopefully, FLGR can inspire the study on the event-based highly efficient, high-speed, and high-dynamic-range sequence classification tasks.
Collapse
|
216
|
Bayesian Recurrent Neural Network Models for Forecasting and Quantifying Uncertainty in Spatial-Temporal Data. ENTROPY 2019; 21:e21020184. [PMID: 33266899 PMCID: PMC7514666 DOI: 10.3390/e21020184] [Citation(s) in RCA: 41] [Impact Index Per Article: 8.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 12/29/2018] [Revised: 02/03/2019] [Accepted: 02/12/2019] [Indexed: 11/20/2022]
Abstract
Recurrent neural networks (RNNs) are nonlinear dynamical models commonly used in the machine learning and dynamical systems literature to represent complex dynamical or sequential relationships between variables. Recently, as deep learning models have become more common, RNNs have been used to forecast increasingly complicated systems. Dynamical spatio-temporal processes represent a class of complex systems that can potentially benefit from these types of models. Although the RNN literature is expansive and highly developed, uncertainty quantification is often ignored. Even when considered, the uncertainty is generally quantified without the use of a rigorous framework, such as a fully Bayesian setting. Here we attempt to quantify uncertainty in a more formal framework while maintaining the forecast accuracy that makes these models appealing, by presenting a Bayesian RNN model for nonlinear spatio-temporal forecasting. Additionally, we make simple modifications to the basic RNN to help accommodate the unique nature of nonlinear spatio-temporal data. The proposed model is applied to a Lorenz simulation and two real-world nonlinear spatio-temporal forecasting applications.
Collapse
|
217
|
Xie J, Liu X, Dajun Zeng D. Mining e-cigarette adverse events in social media using Bi-LSTM recurrent neural network with word embedding representation. J Am Med Inform Assoc 2019; 25:72-80. [PMID: 28505280 DOI: 10.1093/jamia/ocx045] [Citation(s) in RCA: 28] [Impact Index Per Article: 5.6] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/17/2016] [Accepted: 04/11/2017] [Indexed: 02/04/2023] Open
Abstract
Objective Recent years have seen increased worldwide popularity of e-cigarette use. However, the risks of e-cigarettes are underexamined. Most e-cigarette adverse event studies have achieved low detection rates due to limited subject sample sizes in the experiments and surveys. Social media provides a large data repository of consumers' e-cigarette feedback and experiences, which are useful for e-cigarette safety surveillance. However, it is difficult to automatically interpret the informal and nontechnical consumer vocabulary about e-cigarettes in social media. This issue hinders the use of social media content for e-cigarette safety surveillance. Recent developments in deep neural network methods have shown promise for named entity extraction from noisy text. Motivated by these observations, we aimed to design a deep neural network approach to extract e-cigarette safety information in social media. Methods Our deep neural language model utilizes word embedding as the representation of text input and recognizes named entity types with the state-of-the-art Bidirectional Long Short-Term Memory (Bi-LSTM) Recurrent Neural Network. Results Our Bi-LSTM model achieved the best performance compared to 3 baseline models, with a precision of 94.10%, a recall of 91.80%, and an F-measure of 92.94%. We identified 1591 unique adverse events and 9930 unique e-cigarette components (ie, chemicals, flavors, and devices) from our research testbed. Conclusion Although the conditional random field baseline model had slightly better precision than our approach, our Bi-LSTM model achieved much higher recall, resulting in the best F-measure. Our method can be generalized to extract medical concepts from social media for other medical applications.
Collapse
|
218
|
Karako K, Chen Y, Tang W. On medical application of neural networks trained with various types of data. Biosci Trends 2018; 12:553-559. [PMID: 30555113 DOI: 10.5582/bst.2018.01264] [Citation(s) in RCA: 9] [Impact Index Per Article: 1.5] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/05/2022]
Abstract
Neural networks have garnered attention over the past few years. A neural network is a typical model of machine learning that is used to identify visual patterns. Neural networks are used to solve a wide variety of problems, including image recognition problems and time series prediction problems. In addition, neural networks have been applied to medicine over the past few years. This paper classifies the ways in which neural networks have been applied to medicine based on the type of data used to train those networks. Applications of neural networks to medicine can be categorized two types: automated diagnosis and physician aids. Considering the number of patients per physician, neural networks could be used to diagnose diseases related to the vascular system, heart, brain, spinal column, head, neck, and tumors/cancer in three fields: vascular and interventional radiology, interventional cardiology, and neuroradiology. Lastly, this paper also considers areas of medicine where neural networks can be effectively applied in the future.
Collapse
|
219
|
Road Surface Classification Using a Deep Ensemble Network with Sensor Feature Selection. SENSORS 2018; 18:s18124342. [PMID: 30544855 PMCID: PMC6308656 DOI: 10.3390/s18124342] [Citation(s) in RCA: 15] [Impact Index Per Article: 2.5] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 11/29/2018] [Revised: 12/06/2018] [Accepted: 12/07/2018] [Indexed: 11/21/2022]
Abstract
Deep learning is a fast-growing field of research, in particular, for autonomous application. In this study, a deep learning network based on various sensor data is proposed for identifying the roads where the vehicle is driving. Long-Short Term Memory (LSTM) unit and ensemble learning are utilized for network design and a feature selection technique is applied such that unnecessary sensor data could be excluded without a loss of performance. Real vehicle experiments were carried out for the learning and verification of the proposed deep learning structure. The classification performance was verified through four different test roads. The proposed network shows the classification accuracy of 94.6% in the test data.
Collapse
|
220
|
Azizi S, Bayat S, Yan P, Tahmasebi A, Kwak JT, Xu S, Turkbey B, Choyke P, Pinto P, Wood B, Mousavi P, Abolmaesumi P. Deep Recurrent Neural Networks for Prostate Cancer Detection: Analysis of Temporal Enhanced Ultrasound. IEEE TRANSACTIONS ON MEDICAL IMAGING 2018; 37:2695-2703. [PMID: 29994471 PMCID: PMC7983161 DOI: 10.1109/tmi.2018.2849959] [Citation(s) in RCA: 21] [Impact Index Per Article: 3.5] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Subscribe] [Scholar Register] [Indexed: 05/29/2023]
Abstract
Temporal enhanced ultrasound (TeUS), comprising the analysis of variations in backscattered signals from a tissue over a sequence of ultrasound frames, has been previously proposed as a new paradigm for tissue characterization. In this paper, we propose to use deep recurrent neural networks (RNN) to explicitly model the temporal information in TeUS. By investigating several RNN models, we demonstrate that long short-term memory (LSTM) networks achieve the highest accuracy in separating cancer from benign tissue in the prostate. We also present algorithms for in-depth analysis of LSTM networks. Our in vivo study includes data from 255 prostate biopsy cores of 157 patients. We achieve area under the curve, sensitivity, specificity, and accuracy of 0.96, 0.76, 0.98, and 0.93, respectively. Our result suggests that temporal modeling of TeUS using RNN can significantly improve cancer detection accuracy over previously presented works.
Collapse
|
221
|
Idei H, Murata S, Chen Y, Yamashita Y, Tani J, Ogata T. A Neurorobotics Simulation of Autistic Behavior Induced by Unusual Sensory Precision. COMPUTATIONAL PSYCHIATRY (CAMBRIDGE, MASS.) 2018; 2:164-182. [PMID: 30627669 PMCID: PMC6317752 DOI: 10.1162/cpsy_a_00019] [Citation(s) in RCA: 17] [Impact Index Per Article: 2.8] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 12/27/2017] [Accepted: 07/17/2018] [Indexed: 01/27/2023]
Abstract
Recently, applying computational models developed in cognitive science to psychiatric disorders has been recognized as an essential approach for understanding cognitive mechanisms underlying psychiatric symptoms. Autism spectrum disorder is a neurodevelopmental disorder that is hypothesized to affect information processes in the brain involving the estimation of sensory precision (uncertainty), but the mechanism by which observed symptoms are generated from such abnormalities has not been thoroughly investigated. Using a humanoid robot controlled by a neural network using a precision-weighted prediction error minimization mechanism, it is suggested that both increased and decreased sensory precision could induce the behavioral rigidity characterized by resistance to change that is characteristic of autistic behavior. Specifically, decreased sensory precision caused any error signals to be disregarded, leading to invariability of the robot's intention, while increased sensory precision caused an excessive response to error signals, leading to fluctuations and subsequent fixation of intention. The results may provide a system-level explanation of mechanisms underlying different types of behavioral rigidity in autism spectrum and other psychiatric disorders. In addition, our findings suggest that symptoms caused by decreased and increased sensory precision could be distinguishable by examining the internal experience of patients and neural activity coding prediction error signals in the biological brain.
Collapse
|
222
|
Wang H, Xie K, Lian Z, Cui Y, Chen Y, Zhang J, Xie L, Tsien J, Liu T. Large-Scale Circuitry Interactions Upon Earthquake Experiences Revealed by Recurrent Neural Networks. IEEE Trans Neural Syst Rehabil Eng 2018; 26:2115-2125. [PMID: 30296236 PMCID: PMC6298947 DOI: 10.1109/tnsre.2018.2872919] [Citation(s) in RCA: 5] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/09/2022]
Abstract
Brain dynamics has recently received increasing interest due to its significant importance in basic and clinical neurosciences. However, due to inherent difficulties in both data acquisition and data analysis methods, studies on large-scale brain dynamics of mouse with local field potential (LFP) recording are very rare. In this paper, we did a series of works on modeling large-scale mouse brain dynamic activities responding to fearful earthquake. Based on LFP recording data from 13 brain regions that are closely related to fear learning and memory and the effective Bayesian connectivity change point model, we divided the response time series into four stages: "Before," "Earthquake," "Recovery," and "After." We first reported the changes in power and theta-gamma coupling during stage transitions. Then, a recurrent neural network model was designed to model the functional dynamics in these thirteen brain regions and six frequency bands in response to the fear stimulus. Interestingly, our results showed that the functional brain connectivities in theta and gamma bands exhibited distinct response processes: in theta band, there is a separated-united-separated alternation in whole-brain connectivity and a low-high-low change in connectivity strength; however, gamma bands have a united-separated-united transition and a high-low-high alternation in connectivity pattern and strength. In general, our results offer a novel perspective in studying functional brain dynamics under fearful stimulus and reveal its relationship to the brain's structural connectivity substrates.
Collapse
|
223
|
SmartFall: A Smartwatch-Based Fall Detection System Using Deep Learning. SENSORS 2018; 18:s18103363. [PMID: 30304768 PMCID: PMC6210545 DOI: 10.3390/s18103363] [Citation(s) in RCA: 60] [Impact Index Per Article: 10.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 08/31/2018] [Revised: 10/04/2018] [Accepted: 10/04/2018] [Indexed: 11/17/2022]
Abstract
This paper presents SmartFall, an Android app that uses accelerometer data collected from a commodity-based smartwatch Internet of Things (IoT) device to detect falls. The smartwatch is paired with a smartphone that runs the SmartFall application, which performs the computation necessary for the prediction of falls in real time without incurring latency in communicating with a cloud server, while also preserving data privacy. We experimented with both traditional (Support Vector Machine and Naive Bayes) and non-traditional (Deep Learning) machine learning algorithms for the creation of fall detection models using three different fall datasets (Smartwatch, Notch, Farseeing). Our results show that a Deep Learning model for fall detection generally outperforms more traditional models across the three datasets. This is attributed to the Deep Learning model’s ability to automatically learn subtle features from the raw accelerometer data that are not available to Naive Bayes and Support Vector Machine, which are restricted to learning from a small set of extracted features manually specified. Furthermore, the Deep Learning model exhibits a better ability to generalize to new users when predicting falls, an important quality of any model that is to be successful in the real world. We also present a three-layer open IoT system architecture used in SmartFall, which can be easily adapted for the collection and analysis of other sensor data modalities (e.g., heart rate, skin temperature, walking patterns) that enables remote monitoring of a subject’s wellbeing.
Collapse
|
224
|
Kuroki S, Isomura T. Task-Related Synaptic Changes Localized to Small Neuronal Population in Recurrent Neural Network Cortical Models. Front Comput Neurosci 2018; 12:83. [PMID: 30344485 PMCID: PMC6182086 DOI: 10.3389/fncom.2018.00083] [Citation(s) in RCA: 4] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/13/2018] [Accepted: 09/12/2018] [Indexed: 12/17/2022] Open
Abstract
Humans have flexible control over cognitive functions depending on the context. Several studies suggest that the prefrontal cortex (PFC) controls this cognitive flexibility, but the detailed underlying mechanisms remain unclear. Recent developments in machine learning techniques allow simple PFC models written as a recurrent neural network to perform various behavioral tasks like humans and animals. Computational modeling allows the estimation of neuronal parameters that are crucial for performing the tasks, which cannot be observed by biologic experiments. To identify salient neural-network features for flexible cognition tasks, we compared four PFC models using a context-dependent integration task. After training the neural networks with the task, we observed highly plastic synapses localized to a small neuronal population in all models. In three of the models, the neuronal units containing these highly plastic synapses contributed most to the performance. No common tendencies were observed in the distribution of synaptic strengths among the four models. These results suggest that task-dependent plastic synaptic changes are more important for accomplishing flexible cognitive tasks than the structures of the constructed synaptic networks.
Collapse
|
225
|
Spoerer CJ, McClure P, Kriegeskorte N. Corrigendum: Recurrent Convolutional Neural Networks: A Better Model of Biological Object Recognition. Front Psychol 2018; 9:1695. [PMID: 30250446 PMCID: PMC6142017 DOI: 10.3389/fpsyg.2018.01695] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.3] [Reference Citation Analysis] [Key Words] [Track Full Text] [Download PDF] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/09/2018] [Accepted: 08/22/2018] [Indexed: 12/02/2022] Open
|
226
|
Liu M, Cheng D, Yan W. Classification of Alzheimer's Disease by Combination of Convolutional and Recurrent Neural Networks Using FDG-PET Images. Front Neuroinform 2018; 12:35. [PMID: 29970996 PMCID: PMC6018166 DOI: 10.3389/fninf.2018.00035] [Citation(s) in RCA: 94] [Impact Index Per Article: 15.7] [Reference Citation Analysis] [Abstract] [Key Words] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/20/2017] [Accepted: 05/28/2018] [Indexed: 01/17/2023] Open
Abstract
Alzheimer's disease (AD) is an irreversible brain degenerative disorder affecting people aged older than 65 years. Currently, there is no effective cure for AD, but its progression can be delayed with some treatments. Accurate and early diagnosis of AD is vital for the patient care and development of future treatment. Fluorodeoxyglucose positrons emission tomography (FDG-PET) is a functional molecular imaging modality, which proves to be powerful to help understand the anatomical and neural changes of brain related to AD. Most existing methods extract the handcrafted features from images, and then design a classifier to distinguish AD from other groups. These methods highly depends on the preprocessing of brain images, including image rigid registration and segmentation. Motivated by the success of deep learning in image classification, this paper proposes a new classification framework based on combination of 2D convolutional neural networks (CNN) and recurrent neural networks (RNNs), which learns the intra-slice and inter-slice features for classification after decomposition of the 3D PET image into a sequence of 2D slices. The 2D CNNs are built to capture the features of image slices while the gated recurrent unit (GRU) of RNN is cascaded to learn and integrate the inter-slice features for image classification. No rigid registration and segmentation are required for PET images. Our method is evaluated on the baseline FDG-PET images acquired from 339 subjects including 93 AD patients, 146 mild cognitive impairments (MCI) and 100 normal controls (NC) from Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Experimental results show that the proposed method achieves an area under receiver operating characteristic curve (AUC) of 95.3% for AD vs. NC classification and 83.9% for MCI vs. NC classification, demonstrating the promising classification performance.
Collapse
|
227
|
Human midcingulate cortex encodes distributed representations of task progress. Proc Natl Acad Sci U S A 2018; 115:6398-6403. [PMID: 29866834 DOI: 10.1073/pnas.1803650115] [Citation(s) in RCA: 19] [Impact Index Per Article: 3.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 01/11/2023] Open
Abstract
The function of midcingulate cortex (MCC) remains elusive despite decades of investigation and debate. Complicating matters, individual MCC neurons respond to highly diverse task-related events, and MCC activation is reported in most human neuroimaging studies employing a wide variety of task manipulations. Here we investigate this issue by applying a model-based cognitive neuroscience approach involving neural network simulations, functional magnetic resonance imaging, and representational similarity analysis. We demonstrate that human MCC encodes distributed, dynamically evolving representations of extended, goal-directed action sequences. These representations are uniquely sensitive to the stage and identity of each sequence, indicating that MCC sustains contextual information necessary for discriminating between task states. These results suggest that standard univariate approaches for analyzing MCC function overlook the major portion of task-related information encoded by this brain area and point to promising new avenues for investigation.
Collapse
|
228
|
Shi J, Wen H, Zhang Y, Han K, Liu Z. Deep recurrent neural network reveals a hierarchy of process memory during dynamic natural vision. Hum Brain Mapp 2018; 39:2269-2282. [PMID: 29436055 PMCID: PMC5895512 DOI: 10.1002/hbm.24006] [Citation(s) in RCA: 25] [Impact Index Per Article: 4.2] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/16/2017] [Revised: 12/15/2017] [Accepted: 02/06/2018] [Indexed: 02/05/2023] Open
Abstract
The human visual cortex extracts both spatial and temporal visual features to support perception and guide behavior. Deep convolutional neural networks (CNNs) provide a computational framework to model cortical representation and organization for spatial visual processing, but unable to explain how the brain processes temporal information. To overcome this limitation, we extended a CNN by adding recurrent connections to different layers of the CNN to allow spatial representations to be remembered and accumulated over time. The extended model, or the recurrent neural network (RNN), embodied a hierarchical and distributed model of process memory as an integral part of visual processing. Unlike the CNN, the RNN learned spatiotemporal features from videos to enable action recognition. The RNN better predicted cortical responses to natural movie stimuli than the CNN, at all visual areas, especially those along the dorsal stream. As a fully observable model of visual processing, the RNN also revealed a cortical hierarchy of temporal receptive window, dynamics of process memory, and spatiotemporal representations. These results support the hypothesis of process memory, and demonstrate the potential of using the RNN for in-depth computational understanding of dynamic natural vision.
Collapse
|
229
|
Chen S, Hu X. Individual Identification Using the Functional Brain Fingerprint Detected by the Recurrent Neural Network. Brain Connect 2018; 8:197-204. [PMID: 29634323 DOI: 10.1089/brain.2017.0561] [Citation(s) in RCA: 17] [Impact Index Per Article: 2.8] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/12/2022] Open
Abstract
Individual identification based on brain function has gained traction in literature. Investigating individual differences in brain function can provide additional insights into the brain. In this work, we introduce a recurrent neural network-based model for identifying individuals based on only a short segment of resting-state functional magnetic resonance imaging data. In addition, we demonstrate how the global signal and differences in atlases affect individual identifiability. Furthermore, we investigate neural network features that exhibit the uniqueness of each individual. The results indicate that our model is able to identify individuals based on neural features and provides additional information regarding brain dynamics.
Collapse
|
230
|
The Synaptic Properties of Cells Define the Hallmarks of Interval Timing in a Recurrent Neural Network. J Neurosci 2018; 38:4186-4199. [PMID: 29615484 DOI: 10.1523/jneurosci.2651-17.2018] [Citation(s) in RCA: 22] [Impact Index Per Article: 3.7] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/13/2017] [Revised: 03/06/2018] [Accepted: 03/11/2018] [Indexed: 11/21/2022] Open
Abstract
Extensive research has described two key features of interval timing. The bias property is associated with accuracy and implies that time is overestimated for short intervals and underestimated for long intervals. The scalar property is linked to precision and states that the variability of interval estimates increases as a function of interval duration. The neural mechanisms behind these properties are not well understood. Here we implemented a recurrent neural network that mimics a cortical ensemble and includes cells that show paired-pulse facilitation and slow inhibitory synaptic currents. The network produces interval selective responses and reproduces both bias and scalar properties when a Bayesian decoder reads its activity. Notably, the interval-selectivity, timing accuracy, and precision of the network showed complex changes as a function of the decay time constants of the modeled synaptic properties and the level of background activity of the cells. These findings suggest that physiological values of the time constants for paired-pulse facilitation and GABAb, as well as the internal state of the network, determine the bias and scalar properties of interval timing.SIGNIFICANCE STATEMENT Timing is a fundamental element of complex behavior, including music and language. Temporal processing in a wide variety of contexts shows two primary features: time estimates exhibit a shift toward the mean (the bias property) and are more variable for longer intervals (the scalar property). We implemented a recurrent neural network that includes long-lasting synaptic currents, which cannot only produce interval-selective responses but also follow the bias and scalar properties. Interestingly, only physiological values of the time constants for paired-pulse facilitation and GABAb, as well as intermediate background activity within the network can reproduce the two key features of interval timing.
Collapse
|
231
|
Anumula J, Neil D, Delbruck T, Liu SC. Feature Representations for Neuromorphic Audio Spike Streams. Front Neurosci 2018; 12:23. [PMID: 29479300 PMCID: PMC5811520 DOI: 10.3389/fnins.2018.00023] [Citation(s) in RCA: 37] [Impact Index Per Article: 6.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/07/2017] [Accepted: 01/11/2018] [Indexed: 11/24/2022] Open
Abstract
Event-driven neuromorphic spiking sensors such as the silicon retina and the silicon cochlea encode the external sensory stimuli as asynchronous streams of spikes across different channels or pixels. Combining state-of-art deep neural networks with the asynchronous outputs of these sensors has produced encouraging results on some datasets but remains challenging. While the lack of effective spiking networks to process the spike streams is one reason, the other reason is that the pre-processing methods required to convert the spike streams to frame-based features needed for the deep networks still require further investigation. This work investigates the effectiveness of synchronous and asynchronous frame-based features generated using spike count and constant event binning in combination with the use of a recurrent neural network for solving a classification task using N-TIDIGITS18 dataset. This spike-based dataset consists of recordings from the Dynamic Audio Sensor, a spiking silicon cochlea sensor, in response to the TIDIGITS audio dataset. We also propose a new pre-processing method which applies an exponential kernel on the output cochlea spikes so that the interspike timing information is better preserved. The results from the N-TIDIGITS18 dataset show that the exponential features perform better than the spike count features, with over 91% accuracy on the digit classification task. This accuracy corresponds to an improvement of at least 2.5% over the use of spike count features, establishing a new state of the art for this dataset.
Collapse
|
232
|
Yang X, Zhang J, Yoshizoe K, Terayama K, Tsuda K. ChemTS: an efficient python library for de novo molecular generation. SCIENCE AND TECHNOLOGY OF ADVANCED MATERIALS 2017; 18:972-976. [PMID: 29435094 PMCID: PMC5801530 DOI: 10.1080/14686996.2017.1401424] [Citation(s) in RCA: 102] [Impact Index Per Article: 14.6] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Subscribe] [Scholar Register] [Received: 09/29/2017] [Revised: 11/01/2017] [Accepted: 11/02/2017] [Indexed: 05/23/2023]
Abstract
Automatic design of organic materials requires black-box optimization in a vast chemical space. In conventional molecular design algorithms, a molecule is built as a combination of predetermined fragments. Recently, deep neural network models such as variational autoencoders and recurrent neural networks (RNNs) are shown to be effective in de novo design of molecules without any predetermined fragments. This paper presents a novel Python library ChemTS that explores the chemical space by combining Monte Carlo tree search and an RNN. In a benchmarking problem of optimizing the octanol-water partition coefficient and synthesizability, our algorithm showed superior efficiency in finding high-scoring molecules. ChemTS is available at https://github.com/tsudalab/ChemTS.
Collapse
|
233
|
Quiroga MDM, Morris AP, Krekelberg B. Adaptation without Plasticity. Cell Rep 2017; 17:58-68. [PMID: 27681421 DOI: 10.1016/j.celrep.2016.08.089] [Citation(s) in RCA: 17] [Impact Index Per Article: 2.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/25/2016] [Revised: 06/25/2016] [Accepted: 08/25/2016] [Indexed: 11/30/2022] Open
Abstract
Sensory adaptation is a phenomenon in which neurons are affected not only by their immediate input but also by the sequence of preceding inputs. In visual cortex, for example, neurons shift their preferred orientation after exposure to an oriented stimulus. This adaptation is traditionally attributed to plasticity. We show that a recurrent network generates tuning curve shifts observed in cat and macaque visual cortex, even when all synaptic weights and intrinsic properties in the model are fixed. This demonstrates that, in a recurrent network, adaptation on timescales of hundreds of milliseconds does not require plasticity. Given the ubiquity of recurrent connections, this phenomenon likely contributes to responses observed across cortex and shows that plasticity cannot be inferred solely from changes in tuning on these timescales. More broadly, our findings show that recurrent connections can endow a network with a powerful mechanism to store and integrate recent contextual information.
Collapse
|
234
|
Bhalla US. Dendrites, deep learning, and sequences in the hippocampus. Hippocampus 2017; 29:239-251. [PMID: 29024221 DOI: 10.1002/hipo.22806] [Citation(s) in RCA: 10] [Impact Index Per Article: 1.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/09/2017] [Revised: 10/06/2017] [Accepted: 10/10/2017] [Indexed: 11/06/2022]
Abstract
The hippocampus places us both in time and space. It does so over remarkably large spans: milliseconds to years, and centimeters to kilometers. This works for sensory representations, for memory, and for behavioral context. How does it fit in such wide ranges of time and space scales, and keep order among the many dimensions of stimulus context? A key organizing principle for a wide sweep of scales and stimulus dimensions is that of order in time, or sequences. Sequences of neuronal activity are ubiquitous in sensory processing, in motor control, in planning actions, and in memory. Against this strong evidence for the phenomenon, there are currently more models than definite experiments about how the brain generates ordered activity. The flip side of sequence generation is discrimination. Discrimination of sequences has been extensively studied at the behavioral, systems, and modeling level, but again physiological mechanisms are fewer. It is against this backdrop that I discuss two recent developments in neural sequence computation, that at face value share little beyond the label "neural." These are dendritic sequence discrimination, and deep learning. One derives from channel physiology and molecular signaling, the other from applied neural network theory - apparently extreme ends of the spectrum of neural circuit detail. I suggest that each of these topics has deep lessons about the possible mechanisms, scales, and capabilities of hippocampal sequence computation.
Collapse
|
235
|
Chaisangmongkon W, Swaminathan SK, Freedman DJ, Wang XJ. Computing by Robust Transience: How the Fronto-Parietal Network Performs Sequential, Category-Based Decisions. Neuron 2017; 93:1504-1517.e4. [PMID: 28334612 DOI: 10.1016/j.neuron.2017.03.002] [Citation(s) in RCA: 87] [Impact Index Per Article: 12.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/24/2016] [Revised: 09/30/2016] [Accepted: 02/27/2017] [Indexed: 10/19/2022]
Abstract
Decision making involves dynamic interplay between internal judgements and external perception, which has been investigated in delayed match-to-category (DMC) experiments. Our analysis of neural recordings shows that, during DMC tasks, LIP and PFC neurons demonstrate mixed, time-varying, and heterogeneous selectivity, but previous theoretical work has not established the link between these neural characteristics and population-level computations. We trained a recurrent network model to perform DMC tasks and found that the model can remarkably reproduce key features of neuronal selectivity at the single-neuron and population levels. Analysis of the trained networks elucidates that robust transient trajectories of the neural population are the key driver of sequential categorical decisions. The directions of trajectories are governed by network self-organized connectivity, defining a "neural landscape" consisting of a task-tailored arrangement of slow states and dynamical tunnels. With this model, we can identify functionally relevant circuit motifs and generalize the framework to solve other categorization tasks.
Collapse
|
236
|
Xiao L, Zhang Y, Liao B, Zhang Z, Ding L, Jin L. A Velocity-Level Bi-Criteria Optimization Scheme for Coordinated Path Tracking of Dual Robot Manipulators Using Recurrent Neural Network. Front Neurorobot 2017; 11:47. [PMID: 28928651 PMCID: PMC5591439 DOI: 10.3389/fnbot.2017.00047] [Citation(s) in RCA: 10] [Impact Index Per Article: 1.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/24/2017] [Accepted: 08/17/2017] [Indexed: 11/13/2022] Open
Abstract
A dual-robot system is a robotic device composed of two robot arms. To eliminate the joint-angle drift and prevent the occurrence of high joint velocity, a velocity-level bi-criteria optimization scheme, which includes two criteria (i.e., the minimum velocity norm and the repetitive motion), is proposed and investigated for coordinated path tracking of dual robot manipulators. Specifically, to realize the coordinated path tracking of dual robot manipulators, two subschemes are first presented for the left and right robot manipulators. After that, such two subschemes are reformulated as two general quadratic programs (QPs), which can be formulated as one unified QP. A recurrent neural network (RNN) is thus presented to solve effectively the unified QP problem. At last, computer simulation results based on a dual three-link planar manipulator further validate the feasibility and the efficacy of the velocity-level optimization scheme for coordinated path tracking using the recurrent neural network.
Collapse
|
237
|
Ding L, Xiao L, Liao B, Lu R, Peng H. An Improved Recurrent Neural Network for Complex-Valued Systems of Linear Equation and Its Application to Robotic Motion Tracking. Front Neurorobot 2017; 11:45. [PMID: 28919855 PMCID: PMC5585159 DOI: 10.3389/fnbot.2017.00045] [Citation(s) in RCA: 7] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/30/2017] [Accepted: 08/11/2017] [Indexed: 11/13/2022] Open
Abstract
To obtain the online solution of complex-valued systems of linear equation in complex domain with higher precision and higher convergence rate, a new neural network based on Zhang neural network (ZNN) is investigated in this paper. First, this new neural network for complex-valued systems of linear equation in complex domain is proposed and theoretically proved to be convergent within finite time. Then, the illustrative results show that the new neural network model has the higher precision and the higher convergence rate, as compared with the gradient neural network (GNN) model and the ZNN model. Finally, the application for controlling the robot using the proposed method for the complex-valued systems of linear equation is realized, and the simulation results verify the effectiveness and superiorness of the new neural network for the complex-valued systems of linear equation.
Collapse
|
238
|
Spoerer CJ, McClure P, Kriegeskorte N. Recurrent Convolutional Neural Networks: A Better Model of Biological Object Recognition. Front Psychol 2017; 8:1551. [PMID: 28955272 PMCID: PMC5600938 DOI: 10.3389/fpsyg.2017.01551] [Citation(s) in RCA: 73] [Impact Index Per Article: 10.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/03/2017] [Accepted: 08/25/2017] [Indexed: 11/13/2022] Open
Abstract
Feedforward neural networks provide the dominant model of how the brain performs visual object recognition. However, these networks lack the lateral and feedback connections, and the resulting recurrent neuronal dynamics, of the ventral visual pathway in the human and non-human primate brain. Here we investigate recurrent convolutional neural networks with bottom-up (B), lateral (L), and top-down (T) connections. Combining these types of connections yields four architectures (B, BT, BL, and BLT), which we systematically test and compare. We hypothesized that recurrent dynamics might improve recognition performance in the challenging scenario of partial occlusion. We introduce two novel occluded object recognition tasks to test the efficacy of the models, digit clutter (where multiple target digits occlude one another) and digit debris (where target digits are occluded by digit fragments). We find that recurrent neural networks outperform feedforward control models (approximately matched in parametric complexity) at recognizing objects, both in the absence of occlusion and in all occlusion conditions. Recurrent networks were also found to be more robust to the inclusion of additive Gaussian noise. Recurrent neural networks are better in two respects: (1) they are more neurobiologically realistic than their feedforward counterparts; (2) they are better in terms of their ability to recognize objects, especially under challenging conditions. This work shows that computer vision can benefit from using recurrent convolutional architectures and suggests that the ubiquitous recurrent connections in biological brains are essential for task performance.
Collapse
|
239
|
Yu Z, Moirangthem DS, Lee M. Continuous Timescale Long-Short Term Memory Neural Network for Human Intent Understanding. Front Neurorobot 2017; 11:42. [PMID: 28878646 PMCID: PMC5572368 DOI: 10.3389/fnbot.2017.00042] [Citation(s) in RCA: 20] [Impact Index Per Article: 2.9] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/11/2017] [Accepted: 08/07/2017] [Indexed: 11/13/2022] Open
Abstract
Understanding of human intention by observing a series of human actions has been a challenging task. In order to do so, we need to analyze longer sequences of human actions related with intentions and extract the context from the dynamic features. The multiple timescales recurrent neural network (MTRNN) model, which is believed to be a kind of solution, is a useful tool for recording and regenerating a continuous signal for dynamic tasks. However, the conventional MTRNN suffers from the vanishing gradient problem which renders it impossible to be used for longer sequence understanding. To address this problem, we propose a new model named Continuous Timescale Long-Short Term Memory (CTLSTM) in which we inherit the multiple timescales concept into the Long-Short Term Memory (LSTM) recurrent neural network (RNN) that addresses the vanishing gradient problem. We design an additional recurrent connection in the LSTM cell outputs to produce a time-delay in order to capture the slow context. Our experiments show that the proposed model exhibits better context modeling ability and captures the dynamic features on multiple large dataset classification tasks. The results illustrate that the multiple timescales concept enhances the ability of our model to handle longer sequences related with human intentions and hence proving to be more suitable for complex tasks, such as intention recognition.
Collapse
|
240
|
Mark S, Romani S, Jezek K, Tsodyks M. Theta-paced flickering between place-cell maps in the hippocampus: A model based on short-term synaptic plasticity. Hippocampus 2017; 27:959-970. [PMID: 28558154 PMCID: PMC5575492 DOI: 10.1002/hipo.22743] [Citation(s) in RCA: 12] [Impact Index Per Article: 1.7] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/09/2016] [Revised: 05/16/2017] [Accepted: 05/18/2017] [Indexed: 01/29/2023]
Abstract
Hippocampal place cells represent different environments with distinct neural activity patterns. Following an abrupt switch between two familiar configurations of visual cues defining two environments, the hippocampal neural activity pattern switches almost immediately to the corresponding representation. Surprisingly, during a transient period following the switch to the new environment, occasional fast transitions between the two activity patterns (flickering) were observed (Jezek, Henriksen, Treves, Moser, & Moser, 2011). Here we show that an attractor neural network model of place cells with connections endowed with short‐term synaptic plasticity can account for this phenomenon. A memory trace of the recent history of network activity is maintained in the state of the synapses, allowing the network to temporarily reactivate the representation of the previous environment in the absence of the corresponding sensory cues. The model predicts that the number of flickering events depends on the amplitude of the ongoing theta rhythm and the distance between the current position of the animal and its position at the time of cue switching. We test these predictions with new analysis of experimental data. These results suggest a potential role of short‐term synaptic plasticity in recruiting the activity of different cell assemblies and in shaping hippocampal activity of behaving animals.
Collapse
|
241
|
Choi E, Schuetz A, Stewart WF, Sun J. Using recurrent neural network models for early detection of heart failure onset. J Am Med Inform Assoc 2017; 24:361-370. [PMID: 27521897 PMCID: PMC5391725 DOI: 10.1093/jamia/ocw112] [Citation(s) in RCA: 327] [Impact Index Per Article: 46.7] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/27/2016] [Revised: 06/30/2016] [Accepted: 07/05/2016] [Indexed: 11/15/2022] Open
Abstract
Objective We explored whether use of deep learning to model temporal relations among events in electronic health records (EHRs) would improve model performance in predicting initial diagnosis of heart failure (HF) compared to conventional methods that ignore temporality. Materials and Methods Data were from a health system's EHR on 3884 incident HF cases and 28 903 controls, identified as primary care patients, between May 16, 2000, and May 23, 2013. Recurrent neural network (RNN) models using gated recurrent units (GRUs) were adapted to detect relations among time-stamped events (eg, disease diagnosis, medication orders, procedure orders, etc.) with a 12- to 18-month observation window of cases and controls. Model performance metrics were compared to regularized logistic regression, neural network, support vector machine, and K-nearest neighbor classifier approaches. Results Using a 12-month observation window, the area under the curve (AUC) for the RNN model was 0.777, compared to AUCs for logistic regression (0.747), multilayer perceptron (MLP) with 1 hidden layer (0.765), support vector machine (SVM) (0.743), and K-nearest neighbor (KNN) (0.730). When using an 18-month observation window, the AUC for the RNN model increased to 0.883 and was significantly higher than the 0.834 AUC for the best of the baseline methods (MLP). Conclusion Deep learning models adapted to leverage temporal relations appear to improve performance of models for detection of incident heart failure with a short observation window of 12-18 months.
Collapse
|
242
|
Zhao R, Yan R, Wang J, Mao K. Learning to Monitor Machine Health with Convolutional Bi-Directional LSTM Networks. SENSORS 2017; 17:s17020273. [PMID: 28146106 PMCID: PMC5336098 DOI: 10.3390/s17020273] [Citation(s) in RCA: 106] [Impact Index Per Article: 15.1] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 11/24/2016] [Accepted: 01/24/2017] [Indexed: 11/29/2022]
Abstract
In modern manufacturing systems and industries, more and more research efforts have been made in developing effective machine health monitoring systems. Among various machine health monitoring approaches, data-driven methods are gaining in popularity due to the development of advanced sensing and data analytic techniques. However, considering the noise, varying length and irregular sampling behind sensory data, this kind of sequential data cannot be fed into classification and regression models directly. Therefore, previous work focuses on feature extraction/fusion methods requiring expensive human labor and high quality expert knowledge. With the development of deep learning methods in the last few years, which redefine representation learning from raw data, a deep neural network structure named Convolutional Bi-directional Long Short-Term Memory networks (CBLSTM) has been designed here to address raw sensory data. CBLSTM firstly uses CNN to extract local features that are robust and informative from the sequential input. Then, bi-directional LSTM is introduced to encode temporal information. Long Short-Term Memory networks (LSTMs) are able to capture long-term dependencies and model sequential data, and the bi-directional structure enables the capture of past and future contexts. Stacked, fully-connected layers and the linear regression layer are built on top of bi-directional LSTMs to predict the target value. Here, a real-life tool wear test is introduced, and our proposed CBLSTM is able to predict the actual tool wear based on raw sensory data. The experimental results have shown that our model is able to outperform several state-of-the-art baseline methods.
Collapse
|
243
|
Song HF, Yang GR, Wang XJ. Reward-based training of recurrent neural networks for cognitive and value-based tasks. eLife 2017; 6:e21492. [PMID: 28084991 PMCID: PMC5293493 DOI: 10.7554/elife.21492] [Citation(s) in RCA: 98] [Impact Index Per Article: 14.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 09/13/2016] [Accepted: 01/12/2017] [Indexed: 01/27/2023] Open
Abstract
Trained neural network models, which exhibit features of neural activity recorded from behaving animals, may provide insights into the circuit mechanisms of cognitive functions through systematic analysis of network activity and connectivity. However, in contrast to the graded error signals commonly used to train networks through supervised learning, animals learn from reward feedback on definite actions through reinforcement learning. Reward maximization is particularly relevant when optimal behavior depends on an animal's internal judgment of confidence or subjective preferences. Here, we implement reward-based training of recurrent neural networks in which a value network guides learning by using the activity of the decision network to predict future reward. We show that such models capture behavioral and electrophysiological findings from well-known experimental paradigms. Our work provides a unified framework for investigating diverse cognitive and value-based computations, and predicts a role for value representation that is essential for learning, but not executing, a task.
Collapse
|
244
|
Guo X, Merrikh-Bayat F, Gao L, Hoskins BD, Alibart F, Linares-Barranco B, Theogarajan L, Teuscher C, Strukov DB. Modeling and Experimental Demonstration of a Hopfield Network Analog-to-Digital Converter with Hybrid CMOS/Memristor Circuits. Front Neurosci 2015; 9:488. [PMID: 26732664 PMCID: PMC4689862 DOI: 10.3389/fnins.2015.00488] [Citation(s) in RCA: 40] [Impact Index Per Article: 4.4] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/20/2015] [Accepted: 12/07/2015] [Indexed: 11/17/2022] Open
Abstract
The purpose of this work was to demonstrate the feasibility of building recurrent artificial neural networks with hybrid complementary metal oxide semiconductor (CMOS)/memristor circuits. To do so, we modeled a Hopfield network implementing an analog-to-digital converter (ADC) with up to 8 bits of precision. Major shortcomings affecting the ADC's precision, such as the non-ideal behavior of CMOS circuitry and the specific limitations of memristors, were investigated and an effective solution was proposed, capitalizing on the in-field programmability of memristors. The theoretical work was validated experimentally by demonstrating the successful operation of a 4-bit ADC circuit implemented with discrete Pt/TiO2- x /Pt memristors and CMOS integrated circuit components.
Collapse
|
245
|
Heikkinen H, Sharifian F, Vigario R, Vanni S. Feedback to distal dendrites links fMRI signals to neural receptive fields in a spiking network model of the visual cortex. J Neurophysiol 2015; 114:57-69. [PMID: 25925319 DOI: 10.1152/jn.00169.2015] [Citation(s) in RCA: 5] [Impact Index Per Article: 0.6] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/19/2015] [Accepted: 04/27/2015] [Indexed: 11/22/2022] Open
Abstract
The blood oxygenation level-dependent (BOLD) response has been strongly associated with neuronal activity in the brain. However, some neuronal tuning properties are consistently different from the BOLD response. We studied the spatial extent of neural and hemodynamic responses in the primary visual cortex, where the BOLD responses spread and interact over much longer distances than the small receptive fields of individual neurons would predict. Our model shows that a feedforward-feedback loop between V1 and a higher visual area can account for the observed spread of the BOLD response. In particular, anisotropic landing of inputs to compartmental neurons were necessary to account for the BOLD signal spread, while retaining realistic spiking responses. Our work shows that simple dendrites can separate tuning at the synapses and at the action potential output, thus bridging the BOLD signal to the neural receptive fields with high fidelity.
Collapse
|
246
|
Seyed-Allaei H. Phase diagram of spiking neural networks. Front Comput Neurosci 2015; 9:19. [PMID: 25788885 PMCID: PMC4349167 DOI: 10.3389/fncom.2015.00019] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.1] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/26/2014] [Accepted: 02/02/2015] [Indexed: 11/20/2022] Open
Abstract
In computer simulations of spiking neural networks, often it is assumed that every two neurons of the network are connected by a probability of 2%, 20% of neurons are inhibitory and 80% are excitatory. These common values are based on experiments, observations, and trials and errors, but here, I take a different perspective, inspired by evolution, I systematically simulate many networks, each with a different set of parameters, and then I try to figure out what makes the common values desirable. I stimulate networks with pulses and then measure their: dynamic range, dominant frequency of population activities, total duration of activities, maximum rate of population and the occurrence time of maximum rate. The results are organized in phase diagram. This phase diagram gives an insight into the space of parameters – excitatory to inhibitory ratio, sparseness of connections and synaptic weights. This phase diagram can be used to decide the parameters of a model. The phase diagrams show that networks which are configured according to the common values, have a good dynamic range in response to an impulse and their dynamic range is robust in respect to synaptic weights, and for some synaptic weights they oscillates in α or β frequencies, independent of external stimuli.
Collapse
|
247
|
Hayakawa T, Kaneko T, Aoyagi T. A biologically plausible learning rule for the Infomax on recurrent neural networks. Front Comput Neurosci 2014; 8:143. [PMID: 25505404 PMCID: PMC4243565 DOI: 10.3389/fncom.2014.00143] [Citation(s) in RCA: 5] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/17/2014] [Accepted: 10/21/2014] [Indexed: 11/22/2022] Open
Abstract
A fundamental issue in neuroscience is to understand how neuronal circuits in the cerebral cortex play their functional roles through their characteristic firing activity. Several characteristics of spontaneous and sensory-evoked cortical activity have been reproduced by Infomax learning of neural networks in computational studies. There are, however, still few models of the underlying learning mechanisms that allow cortical circuits to maximize information and produce the characteristics of spontaneous and sensory-evoked cortical activity. In the present article, we derive a biologically plausible learning rule for the maximization of information retained through time in dynamics of simple recurrent neural networks. Applying the derived learning rule in a numerical simulation, we reproduce the characteristics of spontaneous and sensory-evoked cortical activity: cell-assembly-like repeats of precise firing sequences, neuronal avalanches, spontaneous replays of learned firing sequences and orientation selectivity observed in the primary visual cortex. We further discuss the similarity between the derived learning rule and the spike timing-dependent plasticity of cortical neurons.
Collapse
|
248
|
Zheng P, Triesch J. Robust development of synfire chains from multiple plasticity mechanisms. Front Comput Neurosci 2014; 8:66. [PMID: 25071537 PMCID: PMC4074894 DOI: 10.3389/fncom.2014.00066] [Citation(s) in RCA: 32] [Impact Index Per Article: 3.2] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/15/2014] [Accepted: 06/02/2014] [Indexed: 11/13/2022] Open
Abstract
Biological neural networks are shaped by a large number of plasticity mechanisms operating at different time scales. How these mechanisms work together to sculpt such networks into effective information processing circuits is still poorly understood. Here we study the spontaneous development of synfire chains in a self-organizing recurrent neural network (SORN) model that combines a number of different plasticity mechanisms including spike-timing-dependent plasticity, structural plasticity, as well as homeostatic forms of plasticity. We find that the network develops an abundance of feed-forward motifs giving rise to synfire chains. The chains develop into ring-like structures, which we refer to as "synfire rings." These rings emerge spontaneously in the SORN network and allow for stable propagation of activity on a fast time scale. A single network can contain multiple non-overlapping rings suppressing each other. On a slower time scale activity switches from one synfire ring to another maintaining firing rate homeostasis. Overall, our results show how the interaction of multiple plasticity mechanisms might give rise to the robust formation of synfire chains in biological neural networks.
Collapse
|
249
|
Cavallari S, Panzeri S, Mazzoni A. Comparison of the dynamics of neural interactions between current-based and conductance-based integrate-and-fire recurrent networks. Front Neural Circuits 2014; 8:12. [PMID: 24634645 PMCID: PMC3943173 DOI: 10.3389/fncir.2014.00012] [Citation(s) in RCA: 29] [Impact Index Per Article: 2.9] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/10/2013] [Accepted: 02/07/2014] [Indexed: 11/13/2022] Open
Abstract
Models of networks of Leaky Integrate-and-Fire (LIF) neurons are a widely used tool for theoretical investigations of brain function. These models have been used both with current- and conductance-based synapses. However, the differences in the dynamics expressed by these two approaches have been so far mainly studied at the single neuron level. To investigate how these synaptic models affect network activity, we compared the single neuron and neural population dynamics of conductance-based networks (COBNs) and current-based networks (CUBNs) of LIF neurons. These networks were endowed with sparse excitatory and inhibitory recurrent connections, and were tested in conditions including both low- and high-conductance states. We developed a novel procedure to obtain comparable networks by properly tuning the synaptic parameters not shared by the models. The so defined comparable networks displayed an excellent and robust match of first order statistics (average single neuron firing rates and average frequency spectrum of network activity). However, these comparable networks showed profound differences in the second order statistics of neural population interactions and in the modulation of these properties by external inputs. The correlation between inhibitory and excitatory synaptic currents and the cross-neuron correlation between synaptic inputs, membrane potentials and spike trains were stronger and more stimulus-modulated in the COBN. Because of these properties, the spike train correlation carried more information about the strength of the input in the COBN, although the firing rates were equally informative in both network models. Moreover, the network activity of COBN showed stronger synchronization in the gamma band, and spectral information about the input higher and spread over a broader range of frequencies. These results suggest that the second order statistics of network dynamics depend strongly on the choice of synaptic model.
Collapse
|
250
|
Neftci E, Das S, Pedroni B, Kreutz-Delgado K, Cauwenberghs G. Event-driven contrastive divergence for spiking neuromorphic systems. Front Neurosci 2014; 7:272. [PMID: 24574952 PMCID: PMC3922083 DOI: 10.3389/fnins.2013.00272] [Citation(s) in RCA: 113] [Impact Index Per Article: 11.3] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/07/2013] [Accepted: 12/22/2013] [Indexed: 11/13/2022] Open
Abstract
Restricted Boltzmann Machines (RBMs) and Deep Belief Networks have been demonstrated to perform efficiently in a variety of applications, such as dimensionality reduction, feature learning, and classification. Their implementation on neuromorphic hardware platforms emulating large-scale networks of spiking neurons can have significant advantages from the perspectives of scalability, power dissipation and real-time interfacing with the environment. However, the traditional RBM architecture and the commonly used training algorithm known as Contrastive Divergence (CD) are based on discrete updates and exact arithmetics which do not directly map onto a dynamical neural substrate. Here, we present an event-driven variation of CD to train a RBM constructed with Integrate & Fire (I&F) neurons, that is constrained by the limitations of existing and near future neuromorphic hardware platforms. Our strategy is based on neural sampling, which allows us to synthesize a spiking neural network that samples from a target Boltzmann distribution. The recurrent activity of the network replaces the discrete steps of the CD algorithm, while Spike Time Dependent Plasticity (STDP) carries out the weight updates in an online, asynchronous fashion. We demonstrate our approach by training an RBM composed of leaky I&F neurons with STDP synapses to learn a generative model of the MNIST hand-written digit dataset, and by testing it in recognition, generation and cue integration tasks. Our results contribute to a machine learning-driven approach for synthesizing networks of spiking neurons capable of carrying out practical, high-level functionality.
Collapse
|