Enhancing ASD detection accuracy: a combined approach of machine learning and deep learning models with natural language processing

Purpose

The main aim of our study was to explore the utility of artificial intelligence (AI) in diagnosing autism spectrum disorder (ASD). The study primarily focused on using machine learning (ML) and deep learning (DL) models to detect ASD potential cases by analyzing text inputs, especially from social media platforms like Twitter. This is to overcome the ongoing challenges in ASD diagnosis, such as the requirement for specialized professionals and extensive resources. Timely identification, particularly in children, is essential to provide immediate intervention and support, thereby improving the quality of life for affected individuals.

Methods

We employed natural language processing (NLP) techniques along with ML models like decision trees, extreme gradient boosting (XGB), k-nearest neighbors algorithm (KNN), and DL models such as recurrent neural networks (RNN), long short-term memory (LSTM), bidirectional long short-term memory (Bi-LSTM), bidirectional encoder representations from transformers (BERT and BERTweet). We extracted a dataset of 404,627 tweets from Twitter users using the platform's API and classified them based on whether they were written by individuals claiming to have ASD (ASD users) or by those without ASD (non-ASD users). From this dataset, we used a subset of 90,000 tweets (45,000 from each classification group) for the training and testing of these models.

Results

The application of our AI models yielded promising results, with the predictive model reaching an accuracy of almost 88% when classifying texts that potentially originated from individuals with ASD.

Conclusion

Our research demonstrated the potential of using AI, particularly DL models, in enhancing the accuracy of ASD detection and diagnosis. This innovative approach signifies the critical role AI can play in advancing early diagnostic techniques, enabling better patient outcomes and underlining the importance of early identification of ASD, especially in children.

EM-COGLOAD: An investigation into age and cognitive load detection using eye tracking and deep learning

Alzheimer's Disease is the most prevalent neurodegenerative disease, and is a leading cause of disability among the elderly. Eye movement behaviour demonstrates potential as a non-invasive biomarker for Alzheimer's Disease, with changes detectable at an early stage after initial onset. This paper introduces a new publicly available dataset: EM-COGLOAD (available at https://osf.io/zjtdq/, DOI: 10.17605/OSF.IO/ZJTDQ). A dual-task paradigm was used to create effects of declined cognitive performance in 75 healthy adults as they carried out visual tracking tasks. Their eye movement was recorded, and time series classification of the extracted eye movement traces was explored using a range of deep learning techniques. The results of this showed that convolutional neural networks were able to achieve an accuracy of 87.5% when distinguishing between eye movement under low and high cognitive load, and 76% when distinguishing between the oldest and youngest age groups.

Benchmarking feature selection and feature extraction methods to improve the performances of machine-learning algorithms for patient classification using metabolomics biomedical data

Objective

Classification tasks are an open challenge in the field of biomedicine. While several machine-learning techniques exist to accomplish this objective, several peculiarities associated with biomedical data, especially when it comes to omics measurements, prevent their use or good performance achievements. Omics approaches aim to understand a complex biological system through systematic analysis of its content at the molecular level. On the other hand, omics data are heterogeneous, sparse and affected by the classical "curse of dimensionality" problem, i.e. having much fewer observation, samples (n) than omics features (p). Furthermore, a major problem with multi-omics data is the imbalance either at the class or feature level. The objective of this work is to study whether feature extraction and/or feature selection techniques can improve the performances of classification machine-learning algorithms on omics measurements.

Methods

Among all omics, metabolomics has emerged as a powerful tool in cancer research, facilitating a deeper understanding of the complex metabolic landscape associated with tumorigenesis and tumor progression. Thus, we selected three publicly available metabolomics datasets, and we applied several feature extraction techniques both linear and non-linear, coupled or not with feature selection methods, and evaluated the performances regarding patient classification in the different configurations for the three datasets.

Results

We provide general workflow and guidelines on when to use those techniques depending on the characteristics of the data available. To further test the extension of our approach to other omics data, we have included a transcriptomics and a proteomics data. Overall, for all datasets, we showed that applying supervised feature selection improves the performances of feature extraction methods for classification purposes. Scripts used to perform all analyses are available at: https://github.com/Plant-Net/Metabolomic_project/.

Identifying factors controlling cellular uptake of gold nanoparticles by machine learning

There is strong interest to improve the therapeutic potential of gold nanoparticles (GNPs) while ensuring their safe development. The utility of GNPs in medicine requires a molecular-level understanding of how GNPs interact with biological systems. Despite considerable research efforts devoted to monitoring the internalisation of GNPs, there is still insufficient understanding of the factors responsible for the variability in GNP uptake in different cell types. Data-driven models are useful for identifying the sources of this variability. Here, we trained multiple machine learning models on 2077 data points for 193 individual nanoparticles from 59 independent studies to predict cellular uptake level of GNPs and compared different algorithms for their efficacies of prediction. The five ensemble learners (Xgboost, random forest, bootstrap aggregation, gradient boosting, light gradient boosting machine) made the best predictions of GNP uptake, accounting for 80-90% of the variance in the test data. The models identified particle size, zeta potential, GNP concentration and exposure duration as the most important drivers of cellular uptake. We expect this proof-of-concept study will foster the more effective use of accumulated cellular uptake data for GNPs and minimise any methodological bias in individual studies that may lead to under- or over-estimation of cellular internalisation rates.

SERT-StructNet: Protein secondary structure prediction method based on multi-factor hybrid deep model

Protein secondary structure prediction (PSSP) is a pivotal research endeavour that plays a crucial role in the comprehensive elucidation of protein functions and properties. Current prediction methodologies are focused on deep-learning techniques, particularly focusing on multi-factor features. Diverging from existing approaches, in this study, we placed special emphasis on the effects of amino acid properties and protein secondary structure propensity scores (SSPs) on secondary structure during the meticulous selection of multi-factor features. This differential feature-selection strategy results in a distinctive and effective amalgamation of the sequence and property features. To harness these multi-factor features optimally, we introduced a hybrid deep feature extraction model. The model initially employs mechanisms such as dilated convolution (D-Conv) and a channel attention network (SENet) for local feature extraction and targeted channel enhancement. Subsequently, a combination of recurrent neural network variants (BiGRU and BiLSTM), along with a transformer module, was employed to achieve global bidirectional information consideration and feature enhancement. This approach to multi-factor feature input and multi-level feature processing enabled a comprehensive exploration of intricate associations among amino acid residues in protein sequences, yielding a Q 3 accuracy of 84.9% and an Sov score of 85.1%. The overall performance surpasses that of the comparable methods. This study introduces a novel and efficient method for determining the PSSP domain, which is poised to deepen our understanding of the practical applications of protein molecular structures.

DeepCORE: An interpretable multi-view deep neural network model to detect co-operative regulatory elements

Gene transcription is an essential process involved in all aspects of cellular functions with significant impact on biological traits and diseases. This process is tightly regulated by multiple elements that co-operate to jointly modulate the transcription levels of target genes. To decipher the complicated regulatory network, we present a novel multi-view attention-based deep neural network that models the relationship between genetic, epigenetic, and transcriptional patterns and identifies co-operative regulatory elements (COREs). We applied this new method, named DeepCORE, to predict transcriptomes in various tissues and cell lines, which outperformed the state-of-the-art algorithms. Furthermore, DeepCORE contains an interpreter that extracts the attention values embedded in the deep neural network, maps the attended regions to putative regulatory elements, and infers COREs based on correlated attentions. The identified COREs are significantly enriched with known promoters and enhancers. Novel regulatory elements discovered by DeepCORE showed epigenetic signatures consistent with the status of histone modification marks.

Predicting Knee Joint Contact Forces During Normal Walking Using Kinematic Inputs With a Long-Short Term Neural Network

Knee joint contact forces are commonly estimated via surrogate measures (i.e., external knee adduction moments or musculoskeletal modeling). Despite its capabilities, modeling is not optimal for clinicians or persons with limited experience. The purpose of this study was to design a novel prediction method for knee joint contact forces that is simplistic in terms of required inputs. This study included marker trajectories and instrumented knee forces during normal walking from the "Grand Challenge" (n = 6) and "CAMS" (n = 2) datasets. Inverse kinematics were used to derive stance phase hip (sagittal, frontal, transverse), knee (sagittal, frontal), ankle (sagittal), and trunk (frontal) kinematics. A long-short term memory network (LSTM) was created using matlab to predict medial and lateral knee force waveforms using combinations of the kinematics. The Grand Challenge and CAMS datasets trained and tested the network, respectively. Musculoskeletal modeling forces were derived using static optimization and joint reaction tools in OpenSim. Waveform accuracy was determined as the proportion of variance and root-mean-square error between network predictions and in vivo data. The LSTM network was highly accurate for medial forces (R2 = 0.77, RMSE = 0.27 BW) and required only frontal hip and knee and sagittal hip and ankle kinematics. Modeled medial force predictions were excellent (R2 = 0.77, RMSE = 0.33 BW). Lateral force predictions were poor for both methods (LSTM R2 = 0.18, RMSE = 0.08 BW; modeling R2 = 0.21, RMSE = 0.54 BW). The designed LSTM network outperformed most reports of musculoskeletal modeling, including those reached in this study, revealing knee joint forces can accurately be predicted by using only kinematic input variables.

MMDB: Multimodal dual-branch model for multi-functional bioactive peptide prediction

Bioactive peptides can hinder oxidative processes and microbial spoilage in foodstuffs and play important roles in treating diverse diseases and disorders. While most of the methods focus on single-functional bioactive peptides and have obtained promising prediction performance, it is still a significant challenge to accurately detect complex and diverse functions simultaneously with the quick increase of multi-functional bioactive peptides. In contrast to previous research on multi-functional bioactive peptide prediction based solely on sequence, we propose a novel multimodal dual-branch (MMDB) lightweight deep learning model that designs two different branches to effectively capture the complementary information of peptide sequence and structural properties. Specifically, a multi-scale dilated convolution with Bi-LSTM branch is presented to effectively model the different scales sequence properties of peptides while a multi-layer convolution branch is proposed to capture structural information. To the best of our knowledge, this is the first effective extraction of peptide sequence features using multi-scale dilated convolution without parameter increase. Multimodal features from both branches are integrated via a fully connected layer for multi-label classification. Compared to state-of-the-art methods, our MMDB model exhibits competitive results across metrics, with a 9.1% Coverage increase and 5.3% and 3.5% improvements in Precision and Accuracy, respectively.

Water quality prediction based on sparse dataset using enhanced machine learning

Water quality in surface bodies remains a pressing issue worldwide. While some regions have rich water quality data, less attention is given to areas that lack sufficient data. Therefore, it is crucial to explore novel ways of managing source-oriented surface water pollution in scenarios with infrequent data collection such as weekly or monthly. Here we showed sparse-dataset-based prediction of water pollution using machine learning. We investigated the efficacy of a traditional Recurrent Neural Network alongside three Long Short-Term Memory (LSTM) models, integrated with the Load Estimator (LOADEST). The research was conducted at a river-lake confluence, an area with intricate hydrological patterns. We found that the Self-Attentive LSTM (SA-LSTM) model outperformed the other three machine learning models in predicting water quality, achieving Nash-Sutcliffe Efficiency (NSE) scores of 0.71 for CODMn and 0.57 for NH3N when utilizing LOADEST-augmented water quality data (referred to as the SA-LSTM-LOADEST model). The SA-LSTM-LOADEST model improved upon the standalone SA-LSTM model by reducing the Root Mean Square Error (RMSE) by 24.6% for CODMn and 21.3% for NH3N. Furthermore, the model maintained its predictive accuracy when data collection intervals were extended from weekly to monthly. Additionally, the SA-LSTM-LOADEST model demonstrated the capability to forecast pollution loads up to ten days in advance. This study shows promise for improving water quality modeling in regions with limited monitoring capabilities.

Working memory capacity for continuous events: The root of temporal compression in episodic memory?

Remembering the unfolding of past episodes usually takes less time than their actual duration. In this study, we evaluated whether such temporal compression emerges when continuous events are too long to be fully held in working memory. To do so, we asked 90 young adults to watch and mentally replay video clips showing people performing a continuous action (e.g., turning a car jack) that lasted 3, 6, 9, 12, or 15 s. For each clip, participants had to carefully watch the event and then to mentally replay it as accurately and precisely as possible. Results showed that mental replay durations increased with event duration but in a non-linear manner: they were close to the actual event duration for short videos (3-9 s), but significantly smaller for longer videos (12 and 15 s). These results suggest that working memory is temporally limited in its capacity to represent continuous events, which could in part explain why the unfolding of events is temporally compressed in episodic memory.

Facilitating interaction between partial differential equation-based dynamics and unknown dynamics for regional wind speed prediction

Regional wind speed prediction is an important spatiotemporal prediction problem which is crucial for optimizing wind power utilization. Nevertheless, the complex dynamics of wind speed pose a formidable challenge to prediction tasks. The evolving dynamics of wind could be governed by underlying physical principles that can be described by partial differential equations (PDE). This study proposes a novel approach called PDE-assisted network (PaNet) for regional wind speed prediction. In PaNet, a new architecture is devised, incorporating both PDE-based dynamics (PDE dynamics) and unknown dynamics. Specifically, this architecture establishes interactions between the two dynamics, regulated by an inter-dynamics communication unit that controls interactions through attention gates. Additionally, recognizing the significance of the initial state for PDE dynamics, an adaptive frequency-gated unit is introduced to generate a suitable initial state for the PDE dynamics by selecting essential frequency components. To evaluate the predictive performance of PaNet, this study conducts comprehensive experiments on two real-world wind speed datasets. The experimental results indicated that the proposed method is superior to other baseline methods.

A self-supervised network for image denoising and watermark removal

In image watermark removal, popular methods depend on given reference non-watermark images in a supervised way to remove watermarks. However, reference non-watermark images are difficult to be obtained in the real world. At the same time, they often suffer from the influence of noise when captured by digital devices. To resolve these issues, in this paper, we present a self-supervised network for image denoising and watermark removal (SSNet). SSNet uses a parallel network in a self-supervised learning way to remove noise and watermarks. Specifically, each sub-network contains two sub-blocks. The upper sub-network uses the first sub-block to remove noise, according to noise-to-noise. Then, the second sub-block in the upper sub-network is used to remove watermarks, according to the distributions of watermarks. To prevent the loss of important information, the lower sub-network is used to simultaneously learn noise and watermarks in a self-supervised learning way. Moreover, two sub-networks interact via attention to extract more complementary salient information. The proposed method does not depend on paired images to learn a blind denoising and watermark removal model, which is very meaningful for real applications. Also, it is more effective than the popular image watermark removal methods in public datasets. Codes can be found at https://github.com/hellloxiaotian/SSNet.

A novel predefined-time neurodynamic approach for mixed variational inequality problems and applications

In this paper, we propose a novel neurodynamic approach with predefined-time stability that offers a solution to address mixed variational inequality problems. Our approach introduces an adjustable time parameter, thereby enhancing flexibility and applicability compared to conventional fixed-time stability methods. By satisfying certain conditions, the proposed approach is capable of converging to a unique solution within a predefined-time, which sets it apart from fixed-time stability and finite-time stability approaches. Furthermore, our approach can be extended to address a wide range of mathematical optimization problems, including variational inequalities, nonlinear complementarity problems, sparse signal recovery problems, and nash equilibria seeking problems in noncooperative games. We provide numerical simulations to validate the theoretical derivation and showcase the effectiveness and feasibility of our proposed method.

Enhanced deep unrolling networks for snapshot compressive hyperspectral imaging

Snapshot compressive hyperspectral imaging necessitates the reconstruction of a complete hyperspectral image from its compressive snapshot measurement, presenting a challenging inverse problem. This paper proposes an enhanced deep unrolling neural network, called EDUNet, to tackle this problem. The EDUNet is constructed via the deep unrolling of a proximal gradient descent algorithm and introduces two innovative modules for gradient-driven update and proximal mapping reflectivity. The gradient-driven update module leverages a memory-assistant descent approach inspired by momentum-based acceleration techniques, for enhancing the unrolled reconstruction process and improving convergence. The proximal mapping is modeled by a sub-network with a cross-stage spectral self-attention, which effectively exploits the inherent self-similarities present in hyperspectral images along the spectral axis. It also enhances feature flow throughout the network, contributing to reconstruction performance gain. Furthermore, we introduce a spectral geometry consistency loss, encouraging EDUNet to prioritize the geometric layouts of spectral curves, leading to a more precise capture of spectral information in hyperspectral images. Experiments are conducted using three benchmark datasets including KAIST, ICVL, and Harvard, along with some real data, comprising a total of 73 samples. The experimental results demonstrate that EDUNet outperforms 15 competing models across four metrics including PSNR, SSIM, SAM, and ERGAS.

Predicting intubation for intensive care units patients: A deep learning approach to improve patient management

OBJECTIVE

For patients in the Intensive Care Unit (ICU), the timing of intubation has a significant association with patients' outcomes. However, accurate prediction of the timing of intubation remains an unsolved challenge due to the noisy, sparse, heterogeneous, and unbalanced nature of ICU data. In this study, our objective is to develop a workflow for pre-processing ICU data and to develop a customized deep learning model to predict the need for intubation.

METHODS

To improve the prediction accuracy, we transform the intubation prediction task into a time series classification task. We carefully design a sequence of data pre-processing steps to handle the multimodal noisy data. Firstly, we discretize the sequential data and address missing data using interpolation. Next, we employ a sampling strategy to address data imbalance and standardize the data to facilitate faster model convergence. Furthermore, we employ the feature selection technique and propose an ensemble model to combine features learned by different deep learning models.

RESULTS

The performance is evaluated on Medical Information Mart for Intensive Care (MIMIC)-III, an ICU dataset. Our proposed Deep Feature Fusion method achieves an area under the curve (AUC) of the receiver operating curve (ROC) of 0.8953, surpassing the performance of other deep learning and traditional machine learning models.

CONCLUSION

Our proposed Deep Feature Fusion method proves to be a viable approach for predicting intubation and outperforms other deep learning and classical machine learning models. The study confirms that high-frequency time-varying indicators, particularly Mean Blood Pressure (MeanBP) and peripheral oxygen saturation (SpO2), are significant risk factors for predicting intubation.

Dynamic changes in task preparation in a multi-task environment: The task transformation paradigm

A key element of human flexible behavior concerns the ability to continuously predict and prepare for sudden changes in tasks or actions. Here, we tested whether people can dynamically modulate task preparation processes and decision-making strategies when the identity of a to-be-performed task becomes uncertain. To this end, we developed a new paradigm where participants need to prepare for one of nine tasks on each trial. Crucially, in some blocks, the task being prepared could suddenly shift to a different task after a longer cue-target interval, by changing either the stimulus category or categorization rule that defined the initial task. We found that participants were able to dynamically modulate task preparation in the face of this task uncertainty. A second experiment shows that these changes in behavior were not simply a function of decreasing task expectancy, but rather of increasing switch expectancy. Finally, in the third and fourth experiment, we demonstrate that these dynamic modulations can be applied in a compositional manner, depending on whether either only the stimulus category or categorization rule would be expected to change.

Dose-dependent changes in orientation amplitude maps in the cat visual cortex after propofol bolus injections

A general intravenous anesthetic propofol (2,6-diisopropylphenol) is widely used in clinical, veterinary practice and animal experiments. It activates gamma- aminobutyric acid (GABAa) receptors. Though the cerebral cortex is one of the major targets of propofol action, no study of dose dependency of propofol action on cat visual cortex was performed yet. Also, no such investigation was done until now using intrinsic signal optical imaging. Here, we report for the first time on the dependency of optical signal in the visual cortex (area 17/area 18) on the propofol dose. Optical imaging of intrinsic responses to visual stimuli was performed in cats before and after propofol bolus injections at different doses on the background of continuous propofol infusion. Orientation amplitude maps were recorded. We found that amplitude of optical signal significantly decreased after a bolus dose of propofol. The effect was dose- and time-dependent producing stronger suppression of optical signal under the highest bolus propofol doses and short time interval after injection. In each hemisphere, amplitude at cardinal and oblique orientations decreased almost equally. However, surprisingly, amplitude at cardinal orientations in the ipsilateral hemisphere was depressed stronger than in contralateral cortex at most time intervals. As the magnitude of optical signal represents the strength of orientation tuned component, these our data give new insights on the mechanisms of generation of orientation selectivity. Our results also provide new data toward understanding brain dynamics under anesthesia and suggest a recommendation for conducting intrinsic signal optical imaging experiments on cortical functioning under propofol anesthesia.

PatchCL-AE: Anomaly detection for medical images using patch-wise contrastive learning-based auto-encoder

Anomaly detection is an important yet challenging task in medical image analysis. Most anomaly detection methods are based on reconstruction, but the performance of reconstruction-based methods is limited due to over-reliance on pixel-level losses. To address the limitation, we propose a patch-wise contrastive learning-based auto-encoder for medical anomaly detection. The key contribution is the patch-wise contrastive learning loss that provides supervision on local semantics to enforce semantic consistency between corresponding input-output patches. Contrastive learning pulls corresponding patch pairs closer while pushing non-corresponding ones apart between input and output, enabling the model to learn local normal features better and improve discriminability on anomalous regions. Additionally, we design an anomaly score based on local semantic discrepancies to pinpoint abnormalities by comparing feature difference rather than pixel variations. Extensive experiments on three public datasets (i.e., brain MRI, retinal OCT, and chest X-ray) achieve state-of-the-art performance, with our method achieving over 99% AUC on retinal and brain images. Both the contrastive patch-wise supervision and patch-discrepancy score provide targeted advancements to overcome the weaknesses in existing approaches.

Goal-oriented inference of environment from redundant observations

The agent learns to organize decision behavior to achieve a behavioral goal, such as reward maximization, and reinforcement learning is often used for this optimization. Learning an optimal behavioral strategy is difficult under the uncertainty that events necessary for learning are only partially observable, called as Partially Observable Markov Decision Process (POMDP). However, the real-world environment also gives many events irrelevant to reward delivery and an optimal behavioral strategy. The conventional methods in POMDP, which attempt to infer transition rules among the entire observations, including irrelevant states, are ineffective in such an environment. Supposing Redundantly Observable Markov Decision Process (ROMDP), here we propose a method for goal-oriented reinforcement learning to efficiently learn state transition rules among reward-related "core states" from redundant observations. Starting with a small number of initial core states, our model gradually adds new core states to the transition diagram until it achieves an optimal behavioral strategy consistent with the Bellman equation. We demonstrate that the resultant inference model outperforms the conventional method for POMDP. We emphasize that our model only containing the core states has high explainability. Furthermore, the proposed method suits online learning as it suppresses memory consumption and improves learning speed.

4 mC site recognition algorithm based on pruned pre-trained DNABert-Pruning model and fused artificial feature encoding

DNA 4 mC plays a crucial role in the genetic expression process of organisms. However, existing deep learning algorithms have shortcomings in the ability to represent DNA sequence features. In this paper, we propose a 4 mC site identification algorithm, DNABert-4mC, based on a fusion of the pruned pre-training DNABert-Pruning model and artificial feature encoding to identify 4 mC sites. The algorithm prunes and compresses the DNABert model, resulting in the pruned pre-training model DNABert-Pruning. This model reduces the number of parameters and removes redundancy from output features, yielding more precise feature representations while upholding accuracy.Simultaneously, the algorithm constructs an artificial feature encoding module to assist the DNABert-Pruning model in feature representation, effectively supplementing the information that is missing from the pre-trained features. The algorithm also introduces the AFF-4mC fusion strategy, which combines artificial feature encoding with the DNABert-Pruning model, to improve the feature representation capability of DNA sequences in multi-semantic spaces and better extract 4 mC sites and the distribution of nucleotide importance within the sequence. In experiments on six independent test sets, the DNABert-4mC algorithm achieved an average AUC value of 93.81%, outperforming seven other advanced algorithms with improvements of 2.05%, 5.02%, 11.32%, 5.90%, 12.02%, 2.42% and 2.34%, respectively.

Spinach genomes reveal migration history and candidate genes for important crop traits

Spinach (Spinacia oleracea) is an important leafy crop possessing notable economic value and health benefits. Current genomic resources include reference genomes and genome-wide association studies. However, the worldwide genetic relationships and the migration history of the crop remained uncertain, and genome-wide association studies have produced extensive gene lists related to agronomic traits. Here, we re-analysed the sequenced genomes of 305 cultivated and wild spinach accessions to unveil the phylogeny and history of cultivated spinach and to explore genetic variation in relation to phenotypes. In contrast to previous studies, we employed machine learning methods (based on Extreme Gradient Boosting, XGBoost) to detect variants that are collectively associated with agronomic traits. Variant-based cluster analyses revealed three primary spinach groups in the Middle East, Asia and Europe/US. Combining admixture analysis and allele-sharing statistics, migration routes of spinach from the Middle East to Europe and Asia are presented. Using XGBoost machine learning models we predict genomic variants influencing bolting time, flowering time, petiole color, and leaf surface texture and propose candidate genes for each trait. This study enhances our understanding of the history and phylogeny of domesticated spinach and provides valuable information on candidate genes for future genetic improvement of the crop.

Advances in machine learning with chemical language models in molecular property and reaction outcome predictions

Molecular properties and reactions form the foundation of chemical space. Over the years, innumerable molecules have been synthesized, a smaller fraction of them found immediate applications, while a larger proportion served as a testimony to creative and empirical nature of the domain of chemical science. With increasing emphasis on sustainable practices, it is desirable that a target set of molecules are synthesized preferably through a fewer empirical attempts instead of a larger library, to realize an active candidate. In this front, predictive endeavors using machine learning (ML) models built on available data acquire high timely significance. Prediction of molecular property and reaction outcome remain one of the burgeoning applications of ML in chemical science. Among several methods of encoding molecular samples for ML models, the ones that employ language like representations are gaining steady popularity. Such representations would additionally help adopt well-developed natural language processing (NLP) models for chemical applications. Given this advantageous background, herein we describe several successful chemical applications of NLP focusing on molecular property and reaction outcome predictions. From relatively simpler recurrent neural networks (RNNs) to complex models like transformers, different network architecture have been leveraged for tasks such as de novo drug design, catalyst generation, forward and retro-synthesis predictions. The chemical language model (CLM) provides promising avenues toward a broad range of applications in a time and cost-effective manner. While we showcase an optimistic outlook of CLMs, attention is also placed on the persisting challenges in reaction domain, which would optimistically be addressed by advanced algorithms tailored to chemical language and with increased availability of high-quality datasets.

Novel high-dense microelectrode array based multimodal bioelectronic monitoring system for cardiac arrhythmia re-entry analysis

In recent decades, significant progress has been made in the treatment of heart diseases, particularly in the field of personalized medicine. Despite the development of genetic tests, phenotyping and risk stratification are performed based on clinical findings and invasive in vivo techniques, such as stimulation conduction mapping techniques and programmed ventricular pacing. Consequently, label-free non-invasive in vitro functional analysis systems are urgently needed for more accurate and effective in vitro risk stratification, model-based therapy planning, and clinical safety profile evaluation of drugs. To overcome these limitations, a novel multilayer high-density microelectrode array (HD-MEA), with an optimized configuration of 512 sensing and 4 pacing electrodes on a sensor area of 100 mm2, was developed for the bioelectronic detection of re-entry arrhythmia patterns. Together with a co-developed front-end, we monitored label-free and in parallel cardiac electrophysiology based on field potential monitoring and mechanical contraction using impedance spectroscopy at the same microelectrode. In proof of principle experiments, human induced pluripotent stem cell (hiPS)-derived cardiomyocytes were cultured on HD-MEAs and used to demonstrate the sensitive quantification of contraction strength modulation by cardioactive drugs such as blebbistatin (IC50 = 4.2 μM), omecamtiv and levosimendan. Strikingly, arrhythmia-typical rotor patterns (re-entry) can be induced by optimized electrical stimulation sequences and detected with high spatial resolution. Therefore, we provide a novel cardiac re-entry analysis system as a promising reference point for diagnostic approaches based on in vitro assays using patient-specific hiPS-derived cardiomyocytes.

Deep learning-assisted mass spectrometry imaging for preliminary screening and pre-classification of psychoactive substances

Currently, it is of great urgency to develop a rapid pre-classification and screening method for suspected drugs as the constantly springing up of new psychoactive substances. In most researches, psychoactive substances classification approaches depended on the similar chemical structures and pharmacological action with known drugs. Such approaches could not face the complicated circumstance of emerging new psychoactive substances. Herein, mass spectrometry imaging and convolutional neural networks (CNN) were used for preliminary screening and pre-classification of suspected psychoactive substances. Mass spectrometry imaging was performed simultaneously on two brain slices as one was from blank group and another one was from psychoactive substance-induced group. Then, fused neurotransmitter variation mass spectrometry images (Nv-MSIs) reflecting the difference of neurotransmitters between two slices were achieved through two homemade programs. A CNN model was developed to classify the Nv-MSIs. Compared with traditional classification methods, CNN achieved better estimation accuracy and required minimal data preprocessing. Also, the specific region on Nv-MSIs and weight of each neurotransmitter that affected the classification most could be unraveled by CNN. Finally, the method was successfully applied to assist the identification of a new psychoactive substance seized recently. This sample was identified as cannabinoids, which greatly promoted the screening process.

A multifactorial study of mass movement in the hilly and gully Loess Plateau based on intensive field surveys and remote sensing techniques

Mass movements, driven by various non-linearly correlated factors, exhibit high randomness, posing vast difficulties for field observations and subsequent investigations into the underlying mechanisms. In this study 157 mass movement incidents (including collapses, slump and spalling) and their primary influencing factors were surveyed in a small catchment of the hilly and gully Loess Plateau, China, through intensive field investigations and remote sensing techniques. The spatial pattern of mass movement and its relation with the influencing factors were assessed, while the relative impact of different factors was studied using the canonical correlation analysis. Results showed that 1) Mass movements predominantly occurred on gully slopes steeper than 70°. Collapses were the main type of mass movement, accounting for 87.9 % of the number of samples. 2) With regard to the impact of individual factors, rainstorms (rainfall intensity >50 mm day-1) significantly enhanced the occurrence frequency, erosion area and erosion volume of mass movement. The occurrence frequency and erosion area / volume were highest at a soil dry bulk density of 1.34 g cm-3 and 1.54 g cm-3, respectively. Mass movement occurred most frequently on unvegetated or unrooted gully slopes, where the resisting effect of vegetation on mass movement was absent. Gully slopes with smooth rather than rugged profiles were also found to be typical areas of mass movement. The occurrence frequency of mass movement decreased with the elevated topographic wetness index (TWI) and distance to slope top and increased with the distance to channels. 3) For the relative impact of different factors, rainfall and shear strength were key factors facilitating and resisting the onset of mass movement, respectively, while topography exerted the greatest influence on the erosion area and volume. This study revealed the relative influence of different factors on occurrence and scale of mass movement, providing a useful reference for modelling and control of the problem.

Diagnosis and classification of intestinal diseases with urine by surface-enhanced Raman spectroscopy

Intestinal Disease (ID) is often characterized by clinical symptoms such as malabsorption, intestinal dysfunction, and injury. If treatment is not timely, it will increase the risk of cancer. Early diagnosis of ID is the key to cure it. There are certain limitations of the conventional diagnostic methods, such as low sensitivity and specificity. Therefore, development of a highly sensitive, non-invasive diagnostic method for ID is extremely important. Urine samples are easier to collect and more sensitive to changes in biomolecules than other pathological diagnostic samples such as tissue and blood. In this paper, a diagnostic method of ID with urine by surface-enhanced Raman spectroscopy (SERS) is proposed. A classification model between ID patients and healthy controls (HC) and a classification model between different pathological types of ID (i.e., benign intestinal disease (BID) and colorectal cancer (CRC)) are established. Here, 830 urine samples, including 100 HC, 443 BID, and 287 CRC, were investigated by SERS. The ID/HC classification model was developed by analyzing the SERS spectra of 150 ID and 100 HC, while BID/CRC classification model was built with 300 BID and 150 CRC patients by principal component analysis (PCA)-support vector machines (SVM). The two established models were internally verified by leave-one-out-cross-validation (LOOCV). Finally, the BID/CRC classification model was further evaluated by 143 BID and 137 CRC patients as an external test set. It shows that the accuracy of the classification model validated by the LOOCV for ID/HC and BID/CRC is 86.4% and 85.56%, respectively. And the accuracy of the BID/CRC classification model with external test set is 82.14%. It shows that high accuracy can be achieved with these two established classification models. It indicates that ID patients in the general population can be identified and BID and CRC patients can be further classified with measuring urine by SERS. It shows that the proposed diagnostic method and established classification models provide valuable information for clinicians to early diagnose ID patients and analyze different stages of ID.

Encoding-related Brain Activity Predicts Subsequent Trial-level Control of Proactive Interference in Working Memory

Proactive interference (PI) appears when familiar information interferes with newly acquired information and is a major cause of forgetting in working memory. It has been proposed that encoding of item-context associations might help mitigate familiarity-based PI. Here, we investigate whether encoding-related brain activation could predict subsequent level of PI at retrieval using trial-specific parametric modulation. Participants were scanned with event-related fMRI while performing a 2-back working memory task with embedded 3-back lures designed to induce PI. We found that the ability to control interference in working memory was modulated by level of activation in the left inferior frontal gyrus, left hippocampus, and bilateral caudate nucleus during encoding. These results provide insight to the processes underlying control of PI in working memory and suggest that encoding of temporal context details support subsequent interference control.

Data-driven learning of chaotic dynamical systems using Discrete-Temporal Sobolev Networks

We introduce the Discrete-Temporal Sobolev Network (DTSN), a neural network loss function that assists dynamical system forecasting by minimizing variational differences between the network output and the training data via a temporal Sobolev norm. This approach is entirely data-driven, architecture agnostic, and does not require derivative information from the estimated system. The DTSN is particularly well suited to chaotic dynamical systems as it minimizes noise in the network output which is crucial for such sensitive systems. For our test cases we consider discrete approximations of the Lorenz-63 system and the Chua circuit. For the network architectures we use the Long Short-Term Memory (LSTM) and the Transformer. The performance of the DTSN is compared with the standard MSE loss for both architectures, as well as with the Physics Informed Neural Network (PINN) loss for the LSTM. The DTSN loss is shown to substantially improve accuracy for both architectures, while requiring less information than the PINN and without noticeably increasing computational time, thereby demonstrating its potential to improve neural network forecasting of dynamical systems.

Growing Like a Tree: Finding Trunks From Graph Skeleton Trees

The message-passing paradigm has served as the foundation of graph neural networks (GNNs) for years, making them achieve great success in a wide range of applications. Despite its elegance, this paradigm presents several unexpected challenges for graph-level tasks, such as the long-range problem, information bottleneck, over-squashing phenomenon, and limited expressivity. In this study, we aim to overcome these major challenges and break the conventional "node- and edge-centric" mindset in graph-level tasks. To this end, we provide an in-depth theoretical analysis of the causes of the information bottleneck from the perspective of information influence. Building on the theoretical results, we offer unique insights to break this bottleneck and suggest extracting a skeleton tree from the original graph, followed by propagating information in a distinctive manner on this tree. Drawing inspiration from natural trees, we further propose to find trunks from graph skeleton trees to create powerful graph representations and develop the corresponding framework for graph-level tasks. Extensive experiments on multiple real-world datasets demonstrate the superiority of our model. Comprehensive experimental analyses further highlight its capability of capturing long-range dependencies and alleviating the over-squashing problem, thereby providing novel insights into graph-level tasks.

Dimension Reduction With Prior Information for Knowledge Discovery

This paper addresses the problem of mapping high-dimensional data to a low-dimensional space, in the presence of other known features. This problem is ubiquitous in science and engineering as there are often controllable/measurable features in most applications. To solve this problem, this paper proposes a broad class of methods, which is referred to as conditional multidimensional scaling (MDS). An algorithm for optimizing the objective function of conditional MDS is also developed. The convergence of this algorithm is proven under mild assumptions. Conditional MDS is illustrated with kinship terms, facial expressions, textile fabrics, car-brand perception, and cylinder machining examples. These examples demonstrate the advantages of conditional MDS over conventional dimension reduction in improving the estimation quality of the reduced-dimension space and simplifying visualization and knowledge discovery tasks. Computer codes for this work are available in the open-source cml R package.

Deep hybrid modeling of a HEK293 process: Combining long short-term memory networks with first principles equations

The combination of physical equations with deep learning is becoming a promising methodology for bioprocess digitalization. In this paper, we investigate for the first time the combination of long short-term memory (LSTM) networks with first principles equations in a hybrid workflow to describe human embryonic kidney 293 (HEK293) culture dynamics. Experimental data of 27 extracellular state variables in 20 fed-batch HEK293 cultures were collected in a parallel high throughput 250 mL cultivation system in an industrial process development setting. The adaptive moment estimation method with stochastic regularization and cross-validation were employed for deep learning. A total of 784 hybrid models with varying deep neural network architectures, depths, layers sizes and node activation functions were compared. In most scenarios, hybrid LSTM models outperformed classical hybrid Feedforward Neural Network (FFNN) models in terms of training and testing error. Hybrid LSTM models revealed to be less sensitive to data resampling than FFNN hybrid models. As disadvantages, Hybrid LSTM models are in general more complex (higher number of parameters) and have a higher computation cost than FFNN hybrid models. The hybrid model with the highest prediction accuracy consisted in a LSTM network with seven internal states connected in series with dynamic material balance equations. This hybrid model correctly predicted the dynamics of the 27 state variables (R2 = 0.93 in the test data set), including biomass, key substrates, amino acids and metabolic by-products for around 10 cultivation days.

Respiratory modulation of the heart rate: A potential biomarker of cardiorespiratory function in human

In recent decade, wearable digital devices have shown potentials for the discovery of novel biomarkers of humans' physiology and behavior. Heart rate (HR) and respiration rate (RR) are most crucial bio-signals in humans' digital phenotyping research. HR is a continuous and non-invasive proxy to autonomic nervous system and ample evidence pinpoints the critical role of respiratory modulation of cardiac function. In the present study, we recorded longitudinal (7 days, 4.63 ± 1.52) HR and RR of 89 freely behaving human subjects (Female: 39, age 57.28 ± 5.67, Male: 50, age 58.48 ± 6.32) and analyzed their dynamics using linear models and information theoretic measures. While HR's linear and nonlinear characteristics were expressed within the plane of the HR-RR directed flow of information (HR→RR - RR→HR), their dynamics were determined by its RR→HR axis. More importantly, RR→HR quantified the effect of alcohol consumption on individuals' cardiorespiratory function independent of their consumed amount of alcohol, thereby signifying the presence of this habit in their daily life activities. The present findings provided evidence for the critical role of the respiratory modulation of HR, which was previously only studied in non-human animals. These results can contribute to humans' phenotyping research by presenting RR→HR as a digital diagnosis/prognosis marker of humans' cardiorespiratory pathology.

A new perspective on positive symptoms: expression of damage or self-defence mechanism of the brain?

Usually, positive neurological symptoms are considered as the consequence of a mere, afinalistic and abnormal increase in function of specific brain areas. However, according to the Theory of Active Inference, which argues that action and perception constitute a loop that updates expectations according to a Bayesian model, the brain is rather an explorer that formulates hypotheses and tests them to assess the correspondence between internal models and reality. Moreover, the cerebral cortex is characterised by a continuous "conflict" between different brain areas, which constantly attempt to expand in order to acquire more of the limited available computational resources, by means of their dopamine-induced neuroplasticity. Thus, it has recently been suggested that dreams, during rapid eye movement sleep (REMS), protect visual brain areas (deprived of their stimuli during rest) from being conquered by other normally stimulated ones. It is therefore conceivable that positive symptoms also have a functional importance for the brain. We evaluate supporting literature data of a 'defensive' role of positive symptoms and the relevance of dopamine-induced neuroplasticity in the context of neurodegenerative and psychiatric diseases. Furthermore, the possible functional significance of idiopathic REMS-related behavioural disorder as well as phantom limb syndrome is examined. We suggest that positive neurological symptoms are not merely a passive expression of a damage, but active efforts, related to dopamine-induced plasticity, to maintain a correct relationship between the external world and its brain representation, thus preventing healthy cortical areas from ousting injured ones.

Frequency Domain Channel-Wise Attack to CNN Classifiers in Motor Imagery Brain-Computer Interfaces

OBJECTIVE

Convolutional neural network (CNN), a classical structure in deep learning, has been commonly deployed in the motor imagery brain-computer interface (MIBCI). Many methods have been proposed to evaluate the vulnerability of such CNN models, primarily by attacking them using direct temporal perturbations. In this work, we propose a novel attacking approach based on perturbations in the frequency domain instead.

METHODS

For a given natural MI trial in the frequency domain, the proposed approach, called frequency domain channel-wise attack (FDCA), generates perturbations at each channel one after another to fool the CNN classifiers. The advances of this strategy are two-fold. First, instead of focusing on the temporal domain, perturbations are generated in the frequency domain where discriminative patterns can be extracted for motor imagery (MI) classification tasks. Second, the perturbing optimization is performed based on differential evolution algorithm in a black-box scenario where detailed model knowledge is not required.

RESULTS

Experimental results demonstrate the effectiveness of the proposed FDCA which achieves a significantly higher success rate than the baselines and existing methods in attacking three major CNN classifiers on four public MI benchmarks.

CONCLUSION

Perturbations generated in the frequency domain yield highly competitive results in attacking MIBCI deployed by CNN models even in a black-box setting, where the model information is well-protected.

SIGNIFICANCE

To our best knowledge, existing MIBCI attack approaches are all gradient-based methods and require details about the victim model, e.g., the parameters and objective function. We provide a more flexible strategy that does not require model details but still produces an effective attack outcome.

One-step Bayesian example-dependent cost classification: The OsC-MLP method

Example-dependent cost classification problems are those where the decision costs depend not only on the true and the attributed classes but also on the sample features. Discriminative algorithms that carry out such classification tasks must take this dependence into account. In some applications, the decision costs are known for the training set but not in production, which complicates the problem. In this paper, we introduce a new one-step Bayesian formulation to train Neural Networks and solve the above limitation for binary cases with one-step Learning Machines, avoiding the drawbacks that unknown analytical forms of the example-dependent costs create. The formulation is based on defining an artificial likelihood ratio by using the available training classification costs in its definition, and proposes a test that does not require the values of the costs for unseen samples. Furthermore, it also includes Bayesian rebalancing mechanisms to combat the negative effects of class imbalance. Experimental results support the consistency and effectiveness of the corresponding algorithms.

Expectation Modulates Repetition Suppression at Late But Not Early Stages during Visual Word Recognition: Evidence from Event-related Potentials

Visual word recognition is commonly rapid and efficient, incorporating top-down predictive processing mechanisms. Neuroimaging studies with face stimuli suggest that repetition suppression (RS) reflects predictive processing at the neural level, as this effect is larger when repetitions are more frequent, that is, more expected. It remains unclear, however, at the temporal level whether and how RS and its modulation by expectation occur in visual word recognition. To address this gap, the present study aimed to investigate the presence and time course of these effects during visual word recognition using EEG. Thirty-six native Cantonese speakers were presented with pairs of Chinese written words and performed a nonlinguistic oddball task. The second word of a pair was either a repetition of the first or a different word (alternation). In repetition blocks, 75% of trials were repetitions and 25% were alternations, whereas the reverse was true in alternation blocks. Topographic analysis of variance of EEG at each time point showed robust RS effects in three time windows (141-227 msec, 242-445 msec, and 467-513 msec) reflecting facilitation of visual word recognition. Importantly, the modulation of RS by expectation was observed at the late rather than early intervals (334-387 msec, 465-550 msec, and 559-632 msec) and more than 100 msec after the first RS effects. In the predictive coding view of RS, only late repetition effects are modulated by expectation, whereas early RS effects may be mediated by lower-level predictions. Taken together, our findings provide the first EEG evidence revealing distinct temporal dynamics of RS effects and repetition probability on RS effects in visual processing of Chinese words.

Biological motion perception in the theoretical framework of perceptual decision-making: An event-related potential study

Biological motion perception plays a critical role in various decisions in daily life. Failure to decide accordingly in such a perceptual task could have life-threatening consequences. Neurophysiology and computational modeling studies suggest two processes mediating perceptual decision-making. One of these signals is associated with the accumulation of sensory evidence and the other with response selection. Recent EEG studies with humans have introduced an event-related potential called Centroparietal Positive Potential (CPP) as a neural marker aligned with the sensory evidence accumulation while effectively distinguishing it from motor-related lateralized readiness potential (LRP). The present study aims to investigate the neural mechanisms of biological motion perception in the framework of perceptual decision-making, which has been overlooked before. More specifically, we examine whether CPP would track the coherence of the biological motion stimuli and could be distinguished from the LRP signal. We recorded EEG from human participants while they performed a direction discrimination task of a point-light walker stimulus embedded in various levels of noise. Our behavioral findings revealed shorter reaction times and reduced miss rates as the coherence of the stimuli increased. In addition, CPP tracked the coherence of the biological motion stimuli with a tendency to reach a common level during the response, albeit with a later onset than the previously reported results in random-dot motion paradigms. Furthermore, CPP was distinguished from the LRP signal based on its temporal profile. Overall, our results suggest that the mechanisms underlying perceptual decision-making generalize to more complex and socially significant stimuli like biological motion.

Source Localization and Spectrum Analyzing of EEG in Stuttering State upon Dysfluent Utterances

Purpose: The present study which addressed adults who stutter (AWS) attempted to investigate power spectral dynamics in the stuttering state by answering the questions using quantitative electroencephalography (qEEG). Method: A 64-channel electroencephalography (EEG) setup was used for data acquisition at 20 AWS. Since the speech, especially stuttering, causes significant noise in the EEG, 2 conditions of speech preparation (SP) and imagined speech (IS) were considered. EEG signals were decomposed into 6 bands. The corresponding sources were localized using the standard low-resolution electromagnetic tomography (sLORETA) tool in both fluent and dysfluent states. Results: Significant differences were noted after analyzing the time-locked EEG signals in fluent and dysfluent utterances. Consistent with previous studies, poor alpha and beta suppression in SP and IS conditions were localized in the left frontotemporal areas in a dysfluent state. This was partly true for the right frontal regions. In the theta range, disfluency was concurrence with increased activation in the left and right motor areas. Increased delta power in the left and right motor areas as well as increased beta2 power over left parietal regions was notable EEG features upon fluent speech. Conclusion: Based on the present findings and those of earlier studies, explaining the neural circuitries involved in stuttering probably requires an examination of the entire frequency spectrum involved in speech.

Prediction of medial knee contact force using multisource fusion recurrent neural network and transfer learning

Estimation of knee contact force (KCF) during gait provides essential information to evaluate knee joint function. Machine learning has been employed to estimate KCF because of the advantages of low computational cost and real-time. However, the existing machine learning models do not adequately consider gait-related data's temporal-dependent, multidimensional, and highly heterogeneous nature. This study is aimed at developing a multisource fusion recurrent neural network to predict the medial condyle KCF. First, a multisource fusion long short-term memory (MF-LSTM) model was established. Then, we developed a transfer learning strategy based on the MF-LSTM model for subject-specific medial KCF prediction. Four subjects with instrumented tibial prostheses were obtained from the literature. The results showed that the MF-LSTM model could predict medial KCF to a certain high level of accuracy (the mean of ρ = 0.970). The transfer learning model improved the prediction accuracy (the mean of ρ = 0.987). This study shows that the MF-LSTM model is a powerful and accurate computational tool for medial KCF prediction. Introducing transfer learning techniques could further improve the prediction performance for the target subject. This coupling strategy can help clinicians accurately estimate and track joint contact forces in real time.

Self-Training Boosted Multi-Factor Matching Network for Composed Image Retrieval

The composed image retrieval (CIR) task aims to retrieve the desired target image for a given multimodal query, i.e., a reference image with its corresponding modification text. The key limitations encountered by existing efforts are two aspects: 1) ignoring the multiple query-target matching factors; 2) ignoring the potential unlabeled reference-target image pairs in existing benchmark datasets. To address these two limitations is non-trivial due to the following challenges: 1) how to effectively model the multiple matching factors in a latent way without direct supervision signals; 2) how to fully utilize the potential unlabeled reference-target image pairs to improve the generalization ability of the CIR model. To address these challenges, in this work, we first propose a CLIP-Transformer based muLtI-factor Matching Network (LIMN), which consists of three key modules: disentanglement-based latent factor tokens mining, dual aggregation-based matching token learning, and dual query-target matching modeling. Thereafter, we design an iterative dual self-training paradigm to further enhance the performance of LIMN by fully utilizing the potential unlabeled reference-target image pairs in a weakly-supervised manner. Specifically, we denote the iterative dual self-training paradigm enhanced LIMN as LIMN+. Extensive experiments on four datasets, including FashionIQ, Shoes, CIRR, and Fashion200 K, show that our proposed LIMN and LIMN+ significantly surpass the state-of-the-art baselines.

Should I Stay or Should I Go? Cognitive Modeling of Left-Turn Gap Acceptance Decisions in Human Drivers

OBJECTIVE

We aim to bridge the gap between naturalistic studies of driver behavior and modern cognitive and neuroscientific accounts of decision making by modeling the cognitive processes underlying left-turn gap acceptance by human drivers.

BACKGROUND

Understanding decisions of human drivers is essential for the development of safe and efficient transportation systems. Current models of decision making in drivers provide little insight into the underlying cognitive processes. On the other hand, laboratory studies of abstract, highly controlled tasks point towards noisy evidence accumulation as a key mechanism governing decision making. However, it is unclear whether the cognitive processes implicated in these tasks are as paramount to decisions that are ingrained in more complex behaviors, such as driving.

RESULTS

The drivers' probability of accepting the available gap increased with the size of the gap; importantly, response time increased with time gap but not distance gap. The generalized drift-diffusion model explained the observed decision outcomes and response time distributions, as well as substantial individual differences in those. Through cross-validation, we demonstrate that the model not only explains the data, but also generalizes to out-of-sample conditions.

CONCLUSION

Our results suggest that dynamic evidence accumulation is an essential mechanism underlying left-turn gap acceptance decisions in human drivers, and exemplify how simple cognitive process models can help to understand human behavior in complex real-world tasks.

APPLICATION

Potential applications of our results include real-time prediction of human behavior by automated vehicles and simulating realistic human-like behaviors in virtual environments for automated vehicles.

Hebbian dreaming for small datasets

The dreaming Hopfield model constitutes a generalization of the Hebbian paradigm for neural networks, that is able to perform on-line learning when "awake" and also to account for off-line "sleeping" mechanisms. The latter have been shown to enhance storing in such a way that, in the long sleep-time limit, this model can reach the maximal storage capacity achievable by networks equipped with symmetric pairwise interactions. In this paper, we inspect the minimal amount of information that must be supplied to such a network to guarantee a successful generalization, and we test it both on random synthetic and on standard structured datasets (i.e., MNIST, Fashion-MNIST and Olivetti). By comparing these minimal thresholds of information with those required by the standard (i.e., always "awake") Hopfield model, we prove that the present network can save up to ∼90% of the dataset size, yet preserving the same performance of the standard counterpart. This suggests that sleep may play a pivotal role in explaining the gap between the large volumes of data required to train artificial neural networks and the relatively small volumes needed by their biological counterparts. Further, we prove that the model Cost function (typically used in statistical mechanics) admits a representation in terms of a standard Loss function (typically used in machine learning) and this allows us to analyze its emergent computational skills both theoretically and computationally: a quantitative picture of its capabilities as a function of its control parameters is achieved and consistency between the two approaches is highlighted. The resulting network is an associative memory for pattern recognition tasks that learns from examples on-line, generalizes correctly (in suitable regions of its control parameters) and optimizes its storage capacity by off-line sleeping: such a reduction of the training cost can be inspiring toward sustainable AI and in situations where data are relatively sparse.

Models for predicting vehicle emissions: A comprehensive review

Air pollution is a primary concern, causing around 7 million premature deaths annually, with traffic-related sources contributing 23 %-45 % of emissions. While several studies have surveyed vehicle emission models, they are either outdated or focus on specific data-driven models. This paper systematically reviews vehicle emission prediction models, comparing traditional approaches with data-driven emission models. The traditional emission models can be divided into average-speed, modal, and other models, noting their reliance on empirical assumptions and parameters that may not be universally applicable. In contrast, we delve into data-driven models utilizing dynamometer and on-road test data for time-series and spatial-temporal predictions. The application of these models is discussed across various scenarios, highlighting the progress and gap. We observed that traditional models, primarily estimating total traffic emissions in study regions, lack micro-level detail crucial for tailored decisions. The direct link between road emission model accuracy and input data quality poses challenges in disaggregating on-road vehicle emission inventories. Due to unique transportation instruments, traffic fleet components, and patterns, exploring the effects of emission-reduction policies in specific cities or regions is urgent. Vehicle characteristics, environmental conditions, traffic scenarios, and prediction scales are common effect factors, while instantaneous driving profiles prove effective in model calibration. In data-driven models, ANN outperforms in estimating emissions and performance of low-power diesel engines with errors not exceeding 5 %. However, no single data-driven method performed excellently in predicting all pollutants. Besides, integrated methods utilizing LSTM, GRU, and RNN outperform individual models. To enhance prediction accuracy considering the inherent connectivity of road networks and spatiotemporal variation patterns of vehicle emissions, GCN is an emerging approach for capturing spatial-temporal relationships based on remote sensing data. Moreover, limited data-driven studies have been performed to forecast particle matter emissions, the main contributors to urban pollution, calling for more attention for future research.

Advancing brain-inspired computing with hybrid neural networks

Brain-inspired computing, drawing inspiration from the fundamental structure and information-processing mechanisms of the human brain, has gained significant momentum in recent years. It has emerged as a research paradigm centered on brain-computer dual-driven and multi-network integration. One noteworthy instance of this paradigm is the hybrid neural network (HNN), which integrates computer-science-oriented artificial neural networks (ANNs) with neuroscience-oriented spiking neural networks (SNNs). HNNs exhibit distinct advantages in various intelligent tasks, including perception, cognition and learning. This paper presents a comprehensive review of HNNs with an emphasis on their origin, concepts, biological perspective, construction framework and supporting systems. Furthermore, insights and suggestions for potential research directions are provided aiming to propel the advancement of the HNN paradigm.

An ADMM-LSTM framework for short-term load forecasting

Accurate short-term load forecasting (STLF) is crucial for maintaining reliable and efficient operations within power systems. With the continuous increase in volume and variety of energy data provided by renewables, electric vehicles and other sources, long short-term memory (LSTM) has emerged as an attractive approach for STLF due to its superiorities in extracting the dynamic temporal information. However, traditional LSTM training methods rely on stochastic gradient methods that have several limitations. This paper presents an innovative LSTM optimization framework via the alternating direction method of multipliers (ADMM) for STLF, dubbed ADMM-LSTM. Explicitly, we train the LSTM network distributedly by the ADMM algorithm. More specifically, we introduce a novel approach to update the parameters in the ADMM-LSTM framework, using a backward-forward order, significantly reducing computational time. Additionally, within the proposed framework, the solution to each subproblem is achieved by utilizing either the proximal point algorithm or local linear approximation, preventing the need for supplementary numerical solvers. This approach confers several advantages, including avoiding issues associated with exploding or vanishing gradients, thanks to the inherent gradient-free characteristics of ADMM-LSTM. Furthermore, we offer a comprehensive theoretical analysis that elucidates the convergence properties inherent to the ADMM-LSTM framework. This analysis provides a deeper understanding of the algorithm's convergence behavior. Lastly, the efficacy of our method is substantiated through a series of experiments conducted on two publicly available datasets. The experimental results demonstrate the superior performance of our approach when compared to existing methods.

Efficient spiking neural network design via neural architecture search

Spiking neural networks (SNNs) are brain-inspired models that utilize discrete and sparse spikes to transmit information, thus having the property of energy efficiency. Recent advances in learning algorithms have greatly improved SNN performance due to the automation of feature engineering. While the choice of neural architecture plays a significant role in deep learning, the current SNN architectures are mainly designed manually, which is a time-consuming and error-prone process. In this paper, we propose a spiking neural architecture search (NAS) method that can automatically find efficient SNNs. To tackle the challenge of long search time faced by SNNs when utilizing NAS, the proposed NAS encodes candidate architectures in a branchless spiking supernet which significantly reduces the computation requirements in the search process. Considering that real-world tasks prefer efficient networks with optimal accuracy under a limited computational budget, we propose a Synaptic Operation (SynOps)-aware optimization to automatically find the computationally efficient subspace of the supernet. Experimental results show that, in less search time, our proposed NAS can find SNNs with higher accuracy and lower computational cost than state-of-the-art SNNs. We also conduct experiments to validate the search process and the trade-off between accuracy and computational cost.

A predictive coding model of the N400

The N400 event-related component has been widely used to investigate the neural mechanisms underlying real-time language comprehension. However, despite decades of research, there is still no unifying theory that can explain both its temporal dynamics and functional properties. In this work, we show that predictive coding - a biologically plausible algorithm for approximating Bayesian inference - offers a promising framework for characterizing the N400. Using an implemented predictive coding computational model, we demonstrate how the N400 can be formalized as the lexico-semantic prediction error produced as the brain infers meaning from the linguistic form of incoming words. We show that the magnitude of lexico-semantic prediction error mirrors the functional sensitivity of the N400 to various lexical variables, priming, contextual effects, as well as their higher-order interactions. We further show that the dynamics of the predictive coding algorithm provides a natural explanation for the temporal dynamics of the N400, and a biologically plausible link to neural activity. Together, these findings directly situate the N400 within the broader context of predictive coding research. More generally, they raise the possibility that the brain may use the same computational mechanism for inference across linguistic and non-linguistic domains.

Advances in machine intelligence-driven virtual screening approaches for big-data

Virtual screening (VS) is an integral and ever-evolving domain of drug discovery framework. The VS is traditionally classified into ligand-based (LB) and structure-based (SB) approaches. Machine intelligence or artificial intelligence has wide applications in the drug discovery domain to reduce time and resource consumption. In combination with machine intelligence algorithms, VS has emerged into revolutionarily progressive technology that learns within robust decision orders for data curation and hit molecule screening from large VS libraries in minutes or hours. The exponential growth of chemical and biological data has evolved as "big-data" in the public domain demands modern and advanced machine intelligence-driven VS approaches to screen hit molecules from ultra-large VS libraries. VS has evolved from an individual approach (LB and SB) to integrated LB and SB techniques to explore various ligand and target protein aspects for the enhanced rate of appropriate hit molecule prediction. Current trends demand advanced and intelligent solutions to handle enormous data in drug discovery domain for screening and optimizing hits or lead with fewer or no false positive hits. Following the big-data drift and tremendous growth in computational architecture, we presented this review. Here, the article categorized and emphasized individual VS techniques, detailed literature presented for machine learning implementation, modern machine intelligence approaches, and limitations and deliberated the future prospects.

Topographical Features of Pediatric Electroencephalography during High Initial Concentration Sevoflurane for Inhalational Induction of Anesthesia

BACKGROUND

High-density electroencephalographic (EEG) monitoring remains underutilized in clinical anesthesia, despite its obvious utility in unraveling the profound physiologic impact of these agents on central nervous system functioning. In school-aged children, the routine practice of rapid induction with high concentrations of inspiratory sevoflurane is commonplace, given its favorable efficacy and tolerance profile. However, few studies investigate topographic EEG during the critical timepoint coinciding with loss of responsiveness-a key moment for anesthesiologists in their everyday practice. The authors hypothesized that high initial sevoflurane inhalation would better precipitate changes in brain regions due to inhomogeneities in maturation across three different age groups compared with gradual stepwise paradigms utilized by other investigators. Knowledge of these changes may inform strategies for agent titration in everyday clinical settings.

METHODS

A total of 37 healthy children aged 5 to 10 yr underwent induction with 4% or greater sevoflurane in high-flow oxygen. Perturbations in anesthetic state were investigated in 23 of these children using 64-channel EEG with the Hjorth Laplacian referencing scheme. Topographical maps illustrated absolute, relative, and total band power across three age groups: 5 to 6 yr (n = 7), 7 to 8 yr (n = 8), and 9 to 10 yr (n = 8).

RESULTS

Spectral analysis revealed a large shift in total power driven by increased delta oscillations. Well-described topographic patterns of anesthesia, e.g., frontal predominance, paradoxical beta excitation, and increased slow activity, were evident in the topographic maps. However, there were no statistically significant age-related changes in spectral power observed in a midline electrode subset between the groups when responsiveness was lost compared to the resting state.

CONCLUSIONS

High initial concentration sevoflurane induction causes large-scale topographic effects on the pediatric EEG. Within the minute after unresponsiveness, this dosage may perturb EEG activity in children to an extent where age-related differences are not discernible.

EDITOR’S PERSPECTIVE

On the pitfalls of Batch Normalization for end-to-end video learning: A study on surgical workflow analysis

Batch Normalization's (BN) unique property of depending on other samples in a batch is known to cause problems in several tasks, including sequence modeling. Yet, BN-related issues are hardly studied for long video understanding, despite the ubiquitous use of BN in CNNs (Convolutional Neural Networks) for feature extraction. Especially in surgical workflow analysis, where the lack of pretrained feature extractors has led to complex, multi-stage training pipelines, limited awareness of BN issues may have hidden the benefits of training CNNs and temporal models end to end. In this paper, we analyze pitfalls of BN in video learning, including issues specific to online tasks such as a 'cheating' effect in anticipation. We observe that BN's properties create major obstacles for end-to-end learning. However, using BN-free backbones, even simple CNN-LSTMs beat the state of the art on three surgical workflow benchmarks by utilizing adequate end-to-end training strategies which maximize temporal context. We conclude that awareness of BN's pitfalls is crucial for effective end-to-end learning in surgical tasks. By reproducing results on natural-video datasets, we hope our insights will benefit other areas of video learning as well. Code is available at: https://gitlab.com/nct_tso_public/pitfalls_bn.