Analyzing biological and artificial neural networks: challenges with opportunities for synergy?

Homeostatic synaptic normalization optimizes learning in network models of neural population codes

Studying and understanding the code of large neural populations hinge on accurate statistical models of population activity. A novel class of models, based on learning to weigh sparse nonlinear Random Projections (RP) of the population, has demonstrated high accuracy, efficiency, and scalability. Importantly, these RP models have a clear and biologically plausible implementation as shallow neural networks. We present a new class of RP models that are learned by optimizing the randomly selected sparse projections themselves. This 'reshaping' of projections is akin to changing synaptic connections in just one layer of the corresponding neural circuit model. We show that Reshaped RP models are more accurate and efficient than the standard RP models in recapitulating the code of tens of cortical neurons from behaving monkeys. Incorporating more biological features and utilizing synaptic normalization in the learning process, results in accurate models that are more efficient. Remarkably, these models exhibit homeostasis in firing rates and total synaptic weights of projection neurons. We further show that these sparse homeostatic reshaped RP models outperform fully connected neural network models. Thus, our new scalable, efficient, and highly accurate population code models are not only biologically plausible but are actually optimized due to their biological features. These findings suggest a dual functional role of synaptic normalization in neural circuits: maintaining spiking and synaptic homeostasis while concurrently optimizing network performance and efficiency in encoding information and learning.

Proposal for a Mechanistic Disease Conceptualization in Clinical Neurosciences: The Neural Network Components (NNC) Model

ABSTRACT

Clinical neurosciences, and psychiatry specifically, have been challenged by the lack of a comprehensive and practical framework that explains the core mechanistic processes of variable psychiatric presentations. Current conceptualization and classification of psychiatric presentations are primarily centered on a non-biologically based clinical descriptive approach. Despite various attempts, advances in neuroscience research have not led to an improved conceptualization or mechanistic classification of psychiatric disorders. This perspective article proposes a new-work-in-progress-framework for conceptualizing psychiatric presentations based on neural network components (NNC). This framework could guide the development of mechanistic disease classification, improve understanding of underpinning pathology, and provide specific intervention targets. This model also has the potential to dissolve artificial barriers between the fields of psychiatry and neurology.

Lexical-Semantic Content, Not Syntactic Structure, Is the Main Contributor to ANN-Brain Similarity of fMRI Responses in the Language Network

Representations from artificial neural network (ANN) language models have been shown to predict human brain activity in the language network. To understand what aspects of linguistic stimuli contribute to ANN-to-brain similarity, we used an fMRI data set of responses to n = 627 naturalistic English sentences (Pereira et al., 2018) and systematically manipulated the stimuli for which ANN representations were extracted. In particular, we (i) perturbed sentences' word order, (ii) removed different subsets of words, or (iii) replaced sentences with other sentences of varying semantic similarity. We found that the lexical-semantic content of the sentence (largely carried by content words) rather than the sentence's syntactic form (conveyed via word order or function words) is primarily responsible for the ANN-to-brain similarity. In follow-up analyses, we found that perturbation manipulations that adversely affect brain predictivity also lead to more divergent representations in the ANN's embedding space and decrease the ANN's ability to predict upcoming tokens in those stimuli. Further, results are robust as to whether the mapping model is trained on intact or perturbed stimuli and whether the ANN sentence representations are conditioned on the same linguistic context that humans saw. The critical result-that lexical-semantic content is the main contributor to the similarity between ANN representations and neural ones-aligns with the idea that the goal of the human language system is to extract meaning from linguistic strings. Finally, this work highlights the strength of systematic experimental manipulations for evaluating how close we are to accurate and generalizable models of the human language network.

Influence of training and expertise on deep neural network attention and human attention during a medical image classification task

In many different domains, experts can make complex decisions after glancing very briefly at an image. However, the perceptual mechanisms underlying expert performance are still largely unknown. Recently, several machine learning algorithms have been shown to outperform human experts in specific tasks. But these algorithms often behave as black boxes and their information processing pipeline remains unknown. This lack of transparency and interpretability is highly problematic in applications involving human lives, such as health care. One way to "open the black box" is to compute an artificial attention map from the model, which highlights the pixels of the input image that contributed the most to the model decision. In this work, we directly compare human visual attention to machine visual attention when performing the same visual task. We have designed a medical diagnosis task involving the detection of lesions in small bowel endoscopic images. We collected eye movements from novices and gastroenterologist experts while they classified medical images according to their relevance for Crohn's disease diagnosis. We trained three state-of-the-art deep learning models on our carefully labeled dataset. Both humans and machine performed the same task. We extracted artificial attention with six different post hoc methods. We show that the model attention maps are significantly closer to human expert attention maps than to novices', especially for pathological images. As the model gets trained and its performance gets closer to the human experts, the similarity between model and human attention increases. Through the understanding of the similarities between the visual decision-making process of human experts and deep neural networks, we hope to inform both the training of new doctors and the architecture of new algorithms.

Neurobiological Causal Models of Language Processing

The language faculty is physically realized in the neurobiological infrastructure of the human brain. Despite significant efforts, an integrated understanding of this system remains a formidable challenge. What is missing from most theoretical accounts is a specification of the neural mechanisms that implement language function. Computational models that have been put forward generally lack an explicit neurobiological foundation. We propose a neurobiologically informed causal modeling approach which offers a framework for how to bridge this gap. A neurobiological causal model is a mechanistic description of language processing that is grounded in, and constrained by, the characteristics of the neurobiological substrate. It intends to model the generators of language behavior at the level of implementational causality. We describe key features and neurobiological component parts from which causal models can be built and provide guidelines on how to implement them in model simulations. Then we outline how this approach can shed new light on the core computational machinery for language, the long-term storage of words in the mental lexicon and combinatorial processing in sentence comprehension. In contrast to cognitive theories of behavior, causal models are formulated in the "machine language" of neurobiology which is universal to human cognition. We argue that neurobiological causal modeling should be pursued in addition to existing approaches. Eventually, this approach will allow us to develop an explicit computational neurobiology of language.

Hierarchical VAEs provide a normative account of motion processing in the primate brain

The relationship between perception and inference, as postulated by Helmholtz in the 19th century, is paralleled in modern machine learning by generative models like Variational Autoencoders (VAEs) and their hierarchical variants. Here, we evaluate the role of hierarchical inference and its alignment with brain function in the domain of motion perception. We first introduce a novel synthetic data framework, Retinal Optic Flow Learning (ROFL), which enables control over motion statistics and their causes. We then present a new hierarchical VAE and test it against alternative models on two downstream tasks: (i) predicting ground truth causes of retinal optic flow (e.g., self-motion); and (ii) predicting the responses of neurons in the motion processing pathway of primates. We manipulate the model architectures (hierarchical versus non-hierarchical), loss functions, and the causal structure of the motion stimuli. We find that hierarchical latent structure in the model leads to several improvements. First, it improves the linear decodability of ground truth factors and does so in a sparse and disentangled manner. Second, our hierarchical VAE outperforms previous state-of-the-art models in predicting neuronal responses and exhibits sparse latent-to-neuron relationships. These results depend on the causal structure of the world, indicating that alignment between brains and artificial neural networks depends not only on architecture but also on matching ecologically relevant stimulus statistics. Taken together, our results suggest that hierarchical Bayesian inference underlines the brain's understanding of the world, and hierarchical VAEs can effectively model this understanding.

Convolutional neural networks for vision neuroscience: significance, developments, and outstanding issues

Convolutional Neural Networks (CNN) are a class of machine learning models predominately used in computer vision tasks and can achieve human-like performance through learning from experience. Their striking similarities to the structural and functional principles of the primate visual system allow for comparisons between these artificial networks and their biological counterparts, enabling exploration of how visual functions and neural representations may emerge in the real brain from a limited set of computational principles. After considering the basic features of CNNs, we discuss the opportunities and challenges of endorsing CNNs as in silico models of the primate visual system. Specifically, we highlight several emerging notions about the anatomical and physiological properties of the visual system that still need to be systematically integrated into current CNN models. These tenets include the implementation of parallel processing pathways from the early stages of retinal input and the reconsideration of several assumptions concerning the serial progression of information flow. We suggest design choices and architectural constraints that could facilitate a closer alignment with biology provide causal evidence of the predictive link between the artificial and biological visual systems. Adopting this principled perspective could potentially lead to new research questions and applications of CNNs beyond modeling object recognition.

Investigating the concept of representation in the neural and psychological sciences

The concept of representation is commonly treated as indispensable to research on brains, behavior, and cognition. Nevertheless, systematic evidence about the ways the concept is applied remains scarce. We present the results of an experiment aimed at elucidating what researchers mean by "representation." Participants were an international group of psychologists, neuroscientists, and philosophers (N = 736). Applying elicitation methodology, participants responded to a survey with experimental scenarios aimed at invoking applications of "representation" and five other ways of describing how the brain responds to stimuli. While we find little disciplinary variation in the application of "representation" and other expressions (e.g., "about" and "carry information"), the results suggest that researchers exhibit uncertainty about what sorts of brain activity involve representations or not; they also prefer non-representational, causal characterizations of the brain's response to stimuli. Potential consequences of these findings are explored, such as reforming or eliminating the concept of representation from use.

Characterizing neural coding performance for populations of sensory neurons: comparing a weighted spike distance metrics to other analytical methods

The identity of sensory stimuli is encoded in the spatio-temporal patterns of responses of the encoding neural population. For stimuli to be discriminated reliably, differences in population responses must be accurately decoded by downstream networks. Several methods to compare patterns of responses have been used by neurophysiologists to characterize the accuracy of the sensory responses studied. Among the most widely used analyses, we note methods based on Euclidean distances or on spike metric distances. Methods based on artificial neural networks and machine learning that recognize and/or classify specific input patterns have also gained popularity. Here, we first compare these three strategies using datasets from three different model systems: the moth olfactory system, the electrosensory system of gymnotids, and leaky-integrate-and-fire (LIF) model responses. We show that the input-weighting procedure inherent to artificial neural networks allows the efficient extraction of information relevant to stimulus discrimination. To combine the convenience of methods such as spike metric distances but leverage the advantages of weighting the inputs, we propose a measure based on geometric distances where each dimension is weighted proportionally to how informative it is. We show that the result of this Weighted Euclidian Distance (WED) analysis performs as well or better than the artificial neural network we tested and outperforms the more traditional spike distance metrics. We applied information theoretic analysis to LIF responses and compared their encoding accuracy with the discrimination accuracy quantified through this WED analysis. We show a high degree of correlation between discrimination accuracy and information content, and that our weighting procedure allowed the efficient use of information present to perform the discrimination task. We argue that our proposed measure provides the flexibility and ease of use sought by neurophysiologists while providing a more powerful way to extract relevant information than more traditional methods.

The neuroconnectionist research programme

Artificial neural networks (ANNs) inspired by biology are beginning to be widely used to model behavioural and neural data, an approach we call 'neuroconnectionism'. ANNs have been not only lauded as the current best models of information processing in the brain but also criticized for failing to account for basic cognitive functions. In this Perspective article, we propose that arguing about the successes and failures of a restricted set of current ANNs is the wrong approach to assess the promise of neuroconnectionism for brain science. Instead, we take inspiration from the philosophy of science, and in particular from Lakatos, who showed that the core of a scientific research programme is often not directly falsifiable but should be assessed by its capacity to generate novel insights. Following this view, we present neuroconnectionism as a general research programme centred around ANNs as a computational language for expressing falsifiable theories about brain computation. We describe the core of the programme, the underlying computational framework and its tools for testing specific neuroscientific hypotheses and deriving novel understanding. Taking a longitudinal view, we review past and present neuroconnectionist projects and their responses to challenges and argue that the research programme is highly progressive, generating new and otherwise unreachable insights into the workings of the brain.

Lexical semantic content, not syntactic structure, is the main contributor to ANN-brain similarity of fMRI responses in the language network

Representations from artificial neural network (ANN) language models have been shown to predict human brain activity in the language network. To understand what aspects of linguistic stimuli contribute to ANN-to-brain similarity, we used an fMRI dataset of responses to n=627 naturalistic English sentences (Pereira et al., 2018) and systematically manipulated the stimuli for which ANN representations were extracted. In particular, we i) perturbed sentences' word order, ii) removed different subsets of words, or iii) replaced sentences with other sentences of varying semantic similarity. We found that the lexical semantic content of the sentence (largely carried by content words) rather than the sentence's syntactic form (conveyed via word order or function words) is primarily responsible for the ANN-to-brain similarity. In follow-up analyses, we found that perturbation manipulations that adversely affect brain predictivity also lead to more divergent representations in the ANN's embedding space and decrease the ANN's ability to predict upcoming tokens in those stimuli. Further, results are robust to whether the mapping model is trained on intact or perturbed stimuli, and whether the ANN sentence representations are conditioned on the same linguistic context that humans saw. The critical result-that lexical-semantic content is the main contributor to the similarity between ANN representations and neural ones-aligns with the idea that the goal of the human language system is to extract meaning from linguistic strings. Finally, this work highlights the strength of systematic experimental manipulations for evaluating how close we are to accurate and generalizable models of the human language network.

Double Model Following Adaptive Control for a Complex Dynamical Network

This paper formulates and solves a new problem of the double model following adaptive control (MFAC) of nodes and links in a complex dynamical network (CDN). This is different from most existing studies on CDN and MFAC. Inspired by the concept of composite systems, the CDN with dynamic links is regarded as an interconnected system composed of an interconnected node group (NG) and link group (LG). Guided by the above-mentioned new idea of viewing a CDN from the perspective of composite systems, by means of Lyapunov theory and proposed related mathematical preliminaries, a new adaptive control scheme is proposed for NG. In addition, to remove the restriction that the states of links in a CDN are unavailable due to physical constraints, technical restraints, and expensive measurement costs, we synthesize the coupling term in LG with the proposed adaptive control scheme for NG, such that the problem of double MFAC of nodes and links in CDN is solved. Finally, a simulation example is presented to verify the theoretical results.

Recent Advances at the Interface of Neuroscience and Artificial Neural Networks

Biological neural networks adapt and learn in diverse behavioral contexts. Artificial neural networks (ANNs) have exploited biological properties to solve complex problems. However, despite their effectiveness for specific tasks, ANNs are yet to realize the flexibility and adaptability of biological cognition. This review highlights recent advances in computational and experimental research to advance our understanding of biological and artificial intelligence. In particular, we discuss critical mechanisms from the cellular, systems, and cognitive neuroscience fields that have contributed to refining the architecture and training algorithms of ANNs. Additionally, we discuss how recent work used ANNs to understand complex neuronal correlates of cognition and to process high throughput behavioral data.

A deep neural network model of the primate superior colliculus for emotion recognition

Although sensory processing is pivotal to nearly every theory of emotion, the evaluation of the visual input as 'emotional' (e.g. a smile as signalling happiness) has been traditionally assumed to take place in supramodal 'limbic' brain regions. Accordingly, subcortical structures of ancient evolutionary origin that receive direct input from the retina, such as the superior colliculus (SC), are traditionally conceptualized as passive relay centres. However, mounting evidence suggests that the SC is endowed with the necessary infrastructure and computational capabilities for the innate recognition and initial categorization of emotionally salient features from retinal information. Here, we built a neurobiologically inspired convolutional deep neural network (DNN) model that approximates physiological, anatomical and connectional properties of the retino-collicular circuit. This enabled us to characterize and isolate the initial computations and discriminations that the DNN model of the SC can perform on facial expressions, based uniquely on the information it directly receives from the virtual retina. Trained to discriminate facial expressions of basic emotions, our model matches human error patterns and above chance, yet suboptimal, classification accuracy analogous to that reported in patients with V1 damage, who rely on retino-collicular pathways for non-conscious vision of emotional attributes. When presented with gratings of different spatial frequencies and orientations never 'seen' before, the SC model exhibits spontaneous tuning to low spatial frequencies and reduced orientation discrimination, as can be expected from the prevalence of the magnocellular (M) over parvocellular (P) projections. Likewise, face manipulation that biases processing towards the M or P pathway affects expression recognition in the SC model accordingly, an effect that dovetails with variations of activity in the human SC purposely measured with ultra-high field functional magnetic resonance imaging. Lastly, the DNN generates saliency maps and extracts visual features, demonstrating that certain face parts, like the mouth or the eyes, provide higher discriminative information than other parts as a function of emotional expressions like happiness and sadness. The present findings support the contention that the SC possesses the necessary infrastructure to analyse the visual features that define facial emotional stimuli also without additional processing stages in the visual cortex or in 'limbic' areas. This article is part of the theme issue 'Cracking the laugh code: laughter through the lens of biology, psychology and neuroscience'.

The Construction of ITP Diagnostic Modeling Based on the Expressions of Hub Genes Associated with M1 Polarization of Macrophages

Purpose

Primary immune thrombocytopenia (ITP) is an immune disease with a diagnosis of exclusion, since no validated biomarkers have been identified. In this study, we explored biomarkers associated with the development of ITP from an immune perspective to inform the clinical diagnosis.

Patients and Methods

Differentially expressed genes (DEGs) between normal and ITP samples were analyzed using limma package. Random forest algorithm and LASSO regression were further used to screen for DEGs associated with ITP. The expression of these hub genes was validated by PCR. The relationship between DEGs and immunity was explored by enrichment analysis. Immune cell infiltration in ITP was analyzed by CIBERSORT and ssGSEA, and the relationship between DEGs and infiltrating immune cells was analyzed by Spearman’s rank correlation analysis. Finally, a diagnostic model related to DEGs was constructed by the neural network, and its efficiency was detected by the ROC curve.

Results

After screening the GEO database and validation by PCR analysis, The expression of CTH and TAF8 were higher and while OSBP2 expression was lower in ITP patients compared to normal subjects (P<0.05). GO enrichment analysis showed that these DEGs were associated with inflammatory immune-related diseases, and KEGG analysis showed that they mainly regulated signaling pathways such as JAK-STAT. CIBERSORT and ssGSEA analyses showed that these DEGs were mainly associated with macrophage M1 polarization. The expression of CTH and TAF8 were positively correlated with M1 expression, while OSBP2 was negatively correlated with M1 expression. The ROC curve showed high accuracy of the neural network model [AUC= 0.939, 95% CI (0.8–1)].

Conclusion

Our results suggest that CTH, TAF8, and OSBP2 can be used as effective diagnostic biomarkers of ITP.

Buice M. MouseNet: A biologically constrained convolutional neural network model for the mouse visual cortex

Convolutional neural networks trained on object recognition derive inspiration from the neural architecture of the visual system in mammals, and have been used as models of the feedforward computation performed in the primate ventral stream. In contrast to the deep hierarchical organization of primates, the visual system of the mouse has a shallower arrangement. Since mice and primates are both capable of visually guided behavior, this raises questions about the role of architecture in neural computation. In this work, we introduce a novel framework for building a biologically constrained convolutional neural network model of the mouse visual cortex. The architecture and structural parameters of the network are derived from experimental measurements, specifically the 100-micrometer resolution interareal connectome, the estimates of numbers of neurons in each area and cortical layer, and the statistics of connections between cortical layers. This network is constructed to support detailed task-optimized models of mouse visual cortex, with neural populations that can be compared to specific corresponding populations in the mouse brain. Using a well-studied image classification task as our working example, we demonstrate the computational capability of this mouse-sized network. Given its relatively small size, MouseNet achieves roughly 2/3rds the performance level on ImageNet as VGG16. In combination with the large scale Allen Brain Observatory Visual Coding dataset, we use representational similarity analysis to quantify the extent to which MouseNet recapitulates the neural representation in mouse visual cortex. Importantly, we provide evidence that optimizing for task performance does not improve similarity to the corresponding biological system beyond a certain point. We demonstrate that the distributions of some physiological quantities are closer to the observed distributions in the mouse brain after task training. We encourage the use of the MouseNet architecture by making the code freely available.

Forecasting COVID-19 Epidemic Trends by Combining a Neural Network with Rt Estimation

On 31 December 2019, a cluster of pneumonia cases of unknown etiology was reported in Wuhan (China). The cases were declared to be Coronavirus Disease 2019 (COVID-19) by the World Health Organization (WHO). COVID-19 has been defined as SARS Coronavirus 2 (SARS-CoV-2). Some countries, e.g., Italy, France, and the United Kingdom (UK), have been subjected to frequent restrictions for preventing the spread of infection, contrary to other ones, e.g., the United States of America (USA) and Sweden. The restrictions afflicted the evolution of trends with several perturbations that destabilized its normal evolution. Globally, Rt has been used to estimate time-varying reproduction numbers during epidemics. Methods: This paper presents a solution based on Deep Learning (DL) for the analysis and forecasting of epidemic trends in new positive cases of SARS-CoV-2 (COVID-19). It combined a neural network (NN) and an Rt estimation by adjusting the data produced by the output layer of the NN on the related Rt estimation. Results: Tests were performed on datasets related to the following countries: Italy, the USA, France, the UK, and Sweden. Positive case registration was retrieved between 24 February 2020 and 11 January 2022. Tests performed on the Italian dataset showed that our solution reduced the Mean Absolute Percentage Error (MAPE) by 28.44%, 39.36%, 22.96%, 17.93%, 28.10%, and 24.50% compared to other ones with the same configuration but that were based on the LSTM, GRU, RNN, ARIMA (1,0,3), and ARIMA (7,2,4) models, or an NN without applying the Rt as a corrective index. It also reduced MAPE by 17.93%, the Mean Absolute Error (MAE) by 34.37%, and the Root Mean Square Error (RMSE) by 43.76% compared to the same model without the adjustment performed by the Rt. Furthermore, it allowed an average MAPE reduction of 5.37%, 63.10%, 17.84%, and 14.91% on the datasets related to the USA, France, the UK, and Sweden, respectively.

Towards a neural network hypothesis for functional cognitive disorders: an extension of the Overfitted Brain Hypothesis

Introduction: Whilst the empirical understanding of functional cognitive disorders (FCD) has advanced in recent years, theoretical and conceptual models have evolved more slowly. Existing frameworks for FCD are based on models of other functional neurological disorders or of metacognitive processes and are recognised to lack mechanistic precision.Methods: In this article, a novel application to FCD of Hoel's Overfitted Brain Hypothesis of the evolved function of dreaming is attempted.Results: This posits that the empirically observed sleep disturbance in FCD entails impaired dreaming which causes the brain to be overfitted and hence unable to generalise appropriately, producing mismatch between memory expectations and memory performance.Conclusions: This formulation of FCD is based on considerations derived from the study of neural networks and shares commonalities with Bayesian models of functional neurological disorders. Additionally, it has implications for future hypothesis-driven research in FCD and suggests a pragmatic basis for management strategies.

Decoding violated sensory expectations from the auditory cortex of anaesthetised mice: Hierarchical recurrent neural network depicts separate 'danger' and 'safety' units

The ability to respond appropriately to sensory information received from the external environment is among the most fundamental capabilities of central nervous systems. In the auditory domain, processes underlying this behaviour are studied by measuring auditory‐evoked electrophysiology during sequences of sounds with predetermined regularities. Identifying neural correlates of ensuing auditory novelty responses is supported by research in experimental animals. In the present study, we reanalysed epidural field potential recordings from the auditory cortex of anaesthetised mice during frequency and intensity oddball stimulation. Multivariate pattern analysis (MVPA) and hierarchical recurrent neural network (RNN) modelling were adopted to explore these data with greater resolution than previously considered using conventional methods. Time‐wise and generalised temporal decoding MVPA approaches revealed previously underestimated asymmetry between responses to sound‐level transitions in the intensity oddball paradigm, in contrast with tone frequency changes. After training, the cross‐validated RNN model architecture with four hidden layers produced output waveforms in response to simulated auditory inputs that were strongly correlated with grand‐average auditory‐evoked potential waveforms (r² > .9). Units in hidden layers were classified based on their temporal response properties and characterised using principal component analysis and sample entropy. These demonstrated spontaneous alpha rhythms, sound onset and offset responses and putative ‘safety’ and ‘danger’ units activated by relatively inconspicuous and salient changes in auditory inputs, respectively. The hypothesised existence of corresponding biological neural sources is naturally derived from this model. If proven, this could have significant implications for prevailing theories of auditory processing.

The geometry of robustness in spiking neural networks

Neural systems are remarkably robust against various perturbations, a phenomenon that still requires a clear explanation. Here, we graphically illustrate how neural networks can become robust. We study spiking networks that generate low-dimensional representations, and we show that the neurons’ subthreshold voltages are confined to a convex region in a lower-dimensional voltage subspace, which we call a 'bounding box'. Any changes in network parameters (such as number of neurons, dimensionality of inputs, firing thresholds, synaptic weights, or transmission delays) can all be understood as deformations of this bounding box. Using these insights, we show that functionality is preserved as long as perturbations do not destroy the integrity of the bounding box. We suggest that the principles underlying robustness in these networks — low-dimensional representations, heterogeneity of tuning, and precise negative feedback — may be key to understanding the robustness of neural systems at the circuit level.

Dynamic Self-Rectifying Liquid Metal-Semiconductor Heterointerfaces: A Platform for Development of Bioinspired Afferent Systems

The assembly of geometrically complex and dynamically active liquid metal/semiconductor heterointerfaces has drawn extensive attention in multidimensional electronic systems. In this study the chemovoltaic driven reactions have enabled the microfluidity of hydrophobic galinstan into a three-dimensional (3D) semiconductor matrix. A dynamic heterointerface is developed between the atomically thin surface oxide of galinstan and the TiO₂-Ni interface. Upon the growth of Ga₂O₃ film at the Ga₂O₃-TiO₂ heterointerface, the partial reduction of the TiO₂ film was confirmed by material characterization techniques. The conductance imaging spectroscopy and electrical measurements are used to investigate the charge transfer at heterointerfaces. Concurrently, the dynamic conductance in artificial synaptic junctions is modulated to mimic the biofunctional communication characteristics of multipolar neurons, including slow and fast inhibitory and excitatory postsynaptic responses. The self-rectifying characteristics, femtojoule energy processing, tunable synaptic events, and notably the coordinated signal recognition are the main characteristics of this multisynaptic device. This novel 3D design of liquid metal-semiconductor structure opens up new opportunities for the development of bioinspired afferent systems. It further facilitates the realization of physical phenomena at liquid metal-semiconductor heterointerfaces.

Forms of explanation and understanding for neuroscience and artificial intelligence

Much of the controversy evoked by the use of deep neural networks as models of biological neural systems amount to debates over what constitutes scientific progress in neuroscience. To discuss what constitutes scientific progress, one must have a goal in mind (progress toward what?). One such long-term goal is to produce scientific explanations of intelligent capacities (e.g., object recognition, relational reasoning). I argue that the most pressing philosophical questions at the intersection of neuroscience and artificial intelligence are ultimately concerned with defining the phenomena to be explained and with what constitute valid explanations of such phenomena. I propose that a foundation in the philosophy of scientific explanation and understanding can scaffold future discussions about how an integrated science of intelligence might progress. Toward this vision, I review relevant theories of scientific explanation and discuss strategies for unifying the scientific goals of neuroscience and AI.

Natural and Artificial Intelligence: A brief introduction to the interplay between AI and neuroscience research

Neuroscience and artificial intelligence (AI) share a long history of collaboration. Advances in neuroscience, alongside huge leaps in computer processing power over the last few decades, have given rise to a new generation of in silico neural networks inspired by the architecture of the brain. These AI systems are now capable of many of the advanced perceptual and cognitive abilities of biological systems, including object recognition and decision making. Moreover, AI is now increasingly being employed as a tool for neuroscience research and is transforming our understanding of brain functions. In particular, deep learning has been used to model how convolutional layers and recurrent connections in the brain's cerebral cortex control important functions, including visual processing, memory, and motor control. Excitingly, the use of neuroscience-inspired AI also holds great promise for understanding how changes in brain networks result in psychopathologies, and could even be utilized in treatment regimes. Here we discuss recent advancements in four areas in which the relationship between neuroscience and AI has led to major advancements in the field; (1) AI models of working memory, (2) AI visual processing, (3) AI analysis of big neuroscience datasets, and (4) computational psychiatry.

Early prediction of developing spontaneous activity in cultured neuronal networks

Synchronization and bursting activity are intrinsic electrophysiological properties of in vivo and in vitro neural networks. During early development, cortical cultures exhibit a wide repertoire of synchronous bursting dynamics whose characterization may help to understand the parameters governing the transition from immature to mature networks. Here we used machine learning techniques to characterize and predict the developing spontaneous activity in mouse cortical neurons on microelectrode arrays (MEAs) during the first three weeks in vitro. Network activity at three stages of early development was defined by 18 electrophysiological features of spikes, bursts, synchrony, and connectivity. The variability of neuronal network activity during early development was investigated by applying k-means and self-organizing map (SOM) clustering analysis to features of bursts and synchrony. These electrophysiological features were predicted at the third week in vitro with high accuracy from those at earlier times using three machine learning models: Multivariate Adaptive Regression Splines, Support Vector Machines, and Random Forest. Our results indicate that initial patterns of electrical activity during the first week in vitro may already predetermine the final development of the neuronal network activity. The methodological approach used here may be applied to explore the biological mechanisms underlying the complex dynamics of spontaneous activity in developing neuronal cultures.

Neural population geometry: An approach for understanding biological and artificial neural networks

Advances in experimental neuroscience have transformed our ability to explore the structure and function of neural circuits. At the same time, advances in machine learning have unleashed the remarkable computational power of artificial neural networks (ANNs). While these two fields have different tools and applications, they present a similar challenge: namely, understanding how information is embedded and processed through high-dimensional representations to solve complex tasks. One approach to addressing this challenge is to utilize mathematical and computational tools to analyze the geometry of these high-dimensional representations, i.e., neural population geometry. We review examples of geometrical approaches providing insight into the function of biological and artificial neural networks: representation untangling in perception, a geometric theory of classification capacity, disentanglement, and abstraction in cognitive systems, topological representations underlying cognitive maps, dynamic untangling in motor systems, and a dynamical approach to cognition. Together, these findings illustrate an exciting trend at the intersection of machine learning, neuroscience, and geometry, in which neural population geometry provides a useful population-level mechanistic descriptor underlying task implementation. Importantly, geometric descriptions are applicable across sensory modalities, brain regions, network architectures, and timescales. Thus, neural population geometry has the potential to unify our understanding of structure and function in biological and artificial neural networks, bridging the gap between single neurons, population activities, and behavior.

Convolutional Neural Networks as a Model of the Visual System: Past, Present, and Future

Convolutional neural networks (CNNs) were inspired by early findings in the study of biological vision. They have since become successful tools in computer vision and state-of-the-art models of both neural activity and behavior on visual tasks. This review highlights what, in the context of CNNs, it means to be a good model in computational neuroscience and the various ways models can provide insight. Specifically, it covers the origins of CNNs and the methods by which we validate them as models of biological vision. It then goes on to elaborate on what we can learn about biological vision by understanding and experimenting on CNNs and discusses emerging opportunities for the use of CNNs in vision research beyond basic object recognition.

On the hierarchical design of biochemical-based digital computations

The understanding of the biochemical processes underpinning various biological systems has significantly increased in recent decades and has even prompted reverse engineering of certain of life's more complex processes. The most prominent example is modern computers designed to mimic neuron activity. This work forms part of growing endeavors to return advances in the theory of computation and electronics to biology. In this context, we present a set of requirements sufficient for the design of biochemical analogs of modern electronics in a hierarchical, modular fashion that mimics the design of modern computational devices. This theoretical approach is based on a simple enzymatic analog of the transistor and supported by numerical simulations of biochemical models of enzymatic networks equivalent to complex, and modular, interconnecting electronic circuitry (including clocks, Flip-Flops, adders, decoders, and multiplexers). Furthermore, the proposed approach has been implemented in the form of a Python library capable of creating and testing models of complex bio-analog digital computations based on the execution of an elementary universal logic gate. In tribute to Claude Shannon, our biochemical network materializes his example of a "password" recognition that moves the language of the modern theory of automata beyond combinatorial logic and towards sequential logic.

DENSE-GWP: AN IMPROVED PRIMARY VISUAL ENCODING MODEL BASED ON DENSE GABOR FEATURES

Brain visual encoding models based on functional magnetic resonance imaging are growing increasingly popular. The Gabor wavelet pyramid model (GWP) is a classic example, exhibiting a good prediction performance for the primary visual cortex (V1, V2, and V3). However, the local variations in the visual stimulation are quite convoluted in terms of spatial frequency, orientation, and position, posing a challenge for visual encoding models. Whether the GWP model can thoroughly extract informative and effective features from visual stimulus remains unclear. To this end, this paper proposes a dense GWP visual encoding model by ameliorating the composition of the Gabor wavelet basis from three aspects: spatial frequency, orientation, and position. The improved model named Dense-GWP model could extract denser features from the image stimulus. A regularization optimization algorithm was used to select informative and effective features, which were crucial for predicting voxel activity in the region of interest. Extensive experimental results showed that the Dense-GWP model exhibits an improved prediction performance and can therefore help further understand the human visual perception mechanism.

Related parameters of affinity and stability prediction of HLA-A*2402 restricted antigen peptides based on molecular docking

Background

Major histocompatibility complex class I (MHC-I) plays an important role in cell immune response, and stable interaction between polypeptides and MHC-I ensures efficient presentation of polypeptide-MHC-I (pMHC-I) molecular complexes to T cells. The aim of this study was to explore ways to improve the affinity and stability of the p-Human Leukocyte Antigen (HLA)-A*2402 complex.

Methods

The peptide sequences of the restricted antigen peptides for HLA-A*2402 and the results of the in vitro competitive binding test were retrieved from the literature. The affinity values were predicted using NetMHCpan v4.1 server, and the stability values were predicted using the NetMHCstab v1.0 server. Auto Vina was used to dock peptides to HLA-A*2402 protein in a flexible docking manner, while Flexpepdock was employed to optimize the docking morphology. Maestro was used to analyze the intermolecular forces and the binding affinity of the complex, while MM-GBSA was used to calculate the binding free energy values.

Results

The intermolecular interactions that maintained the affinity and stability of peptide-HLA-A*2402 complex relied mainly on HB, followed by pi stack. The binding affinity values of molecular docking were associated with the predicted values of affinity and stability, the binding affinity and the binding free energy, as well as the intermolecular force pi-stack. The pi stack had a significant negative correlation with binding affinity and binding free energy. The replacement of the residues of the polypeptides that did not form pi-stack interactions with HLA-A*2402 improved the affinity and/or stability compared to before replacement.

Conclusions

The generation and increase in the number of pi-stacks between peptides and HLA-A*2402 molecules may help improve the affinity and stability of p-HLA-A*2402 complexes. The prediction of intermolecular forces and binding affinity of peptide-HLA by means of molecular docking is a supplement to the current commonly used prediction databases.

The Remarkable Robustness of Surrogate Gradient Learning for Instilling Complex Function in Spiking Neural Networks

Brains process information in spiking neural networks. Their intricate connections shape the diverse functions these networks perform. Yet how network connectivity relates to function is poorly understood, and the functional capabilities of models of spiking networks are still rudimentary. The lack of both theoretical insight and practical algorithms to find the necessary connectivity poses a major impediment to both studying information processing in the brain and building efficient neuromorphic hardware systems. The training algorithms that solve this problem for artificial neural networks typically rely on gradient descent. But doing so in spiking networks has remained challenging due to the nondifferentiable nonlinearity of spikes. To avoid this issue, one can employ surrogate gradients to discover the required connectivity. However, the choice of a surrogate is not unique, raising the question of how its implementation influences the effectiveness of the method. Here, we use numerical simulations to systematically study how essential design parameters of surrogate gradients affect learning performance on a range of classification problems. We show that surrogate gradient learning is robust to different shapes of underlying surrogate derivatives, but the choice of the derivative's scale can substantially affect learning performance. When we combine surrogate gradients with suitable activity regularization techniques, spiking networks perform robust information processing at the sparse activity limit. Our study provides a systematic account of the remarkable robustness of surrogate gradient learning and serves as a practical guide to model functional spiking neural networks.

The structure dilemma in biological and artificial neural networks

Brain research up to date has revealed that structure and function are highly related. Thus, for example, studies have repeatedly shown that the brains of patients suffering from schizophrenia or other diseases have a different connectome compared to healthy people. Apart from stochastic processes, however, an inherent logic describing how neurons connect to each other has not yet been identified. We revisited this structural dilemma by comparing and analyzing artificial and biological-based neural networks. Namely, we used feed-forward and recurrent artificial neural networks as well as networks based on the structure of the micro-connectome of C. elegans and of the human macro-connectome. We trained these diverse networks, which markedly differ in their architecture, initialization and pruning technique, and we found remarkable parallels between biological-based and artificial neural networks, as we were additionally able to show that the dilemma is also present in artificial neural networks. Our findings show that structure contains all the information, but that this structure is not exclusive. Indeed, the same structure was able to solve completely different problems with only minimal adjustments. We particularly put interest on the influence of weights and the neuron offset value, as they show a different adaption behaviour. Our findings open up new questions in the fields of artificial and biological information processing research.

Five points to check when comparing visual perception in humans and machines

With the rise of machines to human-level performance in complex recognition tasks, a growing amount of work is directed toward comparing information processing in humans and machines. These studies are an exciting chance to learn about one system by studying the other. Here, we propose ideas on how to design, conduct, and interpret experiments such that they adequately support the investigation of mechanisms when comparing human and machine perception. We demonstrate and apply these ideas through three case studies. The first case study shows how human bias can affect the interpretation of results and that several analytic tools can help to overcome this human reference point. In the second case study, we highlight the difference between necessary and sufficient mechanisms in visual reasoning tasks. Thereby, we show that contrary to previous suggestions, feedback mechanisms might not be necessary for the tasks in question. The third case study highlights the importance of aligning experimental conditions. We find that a previously observed difference in object recognition does not hold when adapting the experiment to make conditions more equitable between humans and machines. In presenting a checklist for comparative studies of visual reasoning in humans and machines, we hope to highlight how to overcome potential pitfalls in design and inference.

Probing the structure-function relationship with neural networks constructed by solving a system of linear equations

Neural network models are an invaluable tool to understand brain function since they allow us to connect the cellular and circuit levels with behaviour. Neural networks usually comprise a huge number of parameters, which must be chosen carefully such that networks reproduce anatomical, behavioural, and neurophysiological data. These parameters are usually fitted with off-the-shelf optimization algorithms that iteratively change network parameters and simulate the network to evaluate its performance and improve fitting. Here we propose to invert the fitting process by proceeding from the network dynamics towards network parameters. Firing state transitions are chosen according to the transition graph associated with the solution of a task. Then, a system of linear equations is constructed from the network firing states and membrane potentials, in a way that guarantees the consistency of the system. This allows us to uncouple the dynamical features of the model, like its neurons firing rate and correlation, from the structural features, and the task-solving algorithm implemented by the network. We employed our method to probe the structure–function relationship in a sequence memory task. The networks obtained showed connectivity and firing statistics that recapitulated experimental observations. We argue that the proposed method is a complementary and needed alternative to the way neural networks are constructed to model brain function.

Lessons From Deep Neural Networks for Studying the Coding Principles of Biological Neural Networks

One of the central goals in systems neuroscience is to understand how information is encoded in the brain, and the standard approach is to identify the relation between a stimulus and a neural response. However, the feature of a stimulus is typically defined by the researcher's hypothesis, which may cause biases in the research conclusion. To demonstrate potential biases, we simulate four likely scenarios using deep neural networks trained on the image classification dataset CIFAR-10 and demonstrate the possibility of selecting suboptimal/irrelevant features or overestimating the network feature representation/noise correlation. Additionally, we present studies investigating neural coding principles in biological neural networks to which our points can be applied. This study aims to not only highlight the importance of careful assumptions and interpretations regarding the neural response to stimulus features but also suggest that the comparative study between deep and biological neural networks from the perspective of machine learning can be an effective strategy for understanding the coding principles of the brain.

Will We Ever Have Conscious Machines?

The question of whether artificial beings or machines could become self-aware or conscious has been a philosophical question for centuries. The main problem is that self-awareness cannot be observed from an outside perspective and the distinction of being really self-aware or merely a clever imitation cannot be answered without access to knowledge about the mechanism's inner workings. We investigate common machine learning approaches with respect to their potential ability to become self-aware. We realize that many important algorithmic steps toward machines with a core consciousness have already been taken.

Predicting the fMRI Signal Fluctuation with Recurrent Neural Networks Trained on Vascular Network Dynamics

Resting-state functional MRI (rs-fMRI) studies have revealed specific low-frequency hemodynamic signal fluctuations (<0.1 Hz) in the brain, which could be related to neuronal oscillations through the neurovascular coupling mechanism. Given the vascular origin of the fMRI signal, it remains challenging to separate the neural correlates of global rs-fMRI signal fluctuations from other confounding sources. However, the slow-oscillation detected from individual vessels by single-vessel fMRI presents strong correlation to neural oscillations. Here, we use recurrent neural networks (RNNs) to predict the future temporal evolution of the rs-fMRI slow oscillation from both rodent and human brains. The RNNs trained with vessel-specific rs-fMRI signals encode the unique brain oscillatory dynamic feature, presenting more effective prediction than the conventional autoregressive model. This RNN-based predictive modeling of rs-fMRI datasets from the Human Connectome Project (HCP) reveals brain state-specific characteristics, demonstrating an inverse relationship between the global rs-fMRI signal fluctuation with the internal default-mode network (DMN) correlation. The RNN prediction method presents a unique data-driven encoding scheme to specify potential brain state differences based on the global fMRI signal fluctuation, but not solely dependent on the global variance.

An Integrative Remote Sensing Application of Stacked Autoencoder for Atmospheric Correction and Cyanobacteria Estimation Using Hyperspectral Imagery

Hyperspectral image sensing can be used to effectively detect the distribution of harmful cyanobacteria. To accomplish this, physical- and/or model-based simulations have been conducted to perform an atmospheric correction (AC) and an estimation of pigments, including phycocyanin (PC) and chlorophyll-a (Chl-a), in cyanobacteria. However, such simulations were undesirable in certain cases, due to the difficulty of representing dynamically changing aerosol and water vapor in the atmosphere and the optical complexity of inland water. Thus, this study was focused on the development of a deep neural network model for AC and cyanobacteria estimation, without considering the physical formulation. The stacked autoencoder (SAE) network was adopted for the feature extraction and dimensionality reduction of hyperspectral imagery. The artificial neural network (ANN) and support vector regression (SVR) were sequentially applied to achieve AC and estimate cyanobacteria concentrations (i.e., SAE-ANN and SAE-SVR). Further, the ANN and SVR models without SAE were compared with SAE-ANN and SAE-SVR models for the performance evaluations. In terms of AC performance, both SAE-ANN and SAE-SVR displayed reasonable accuracy with the Nash–Sutcliffe efficiency (NSE) > 0.7. For PC and Chl-a estimation, the SAE-ANN model showed the best performance, by yielding NSE values > 0.79 and > 0.77, respectively. SAE, with fine tuning operators, improved the accuracy of the original ANN and SVR estimations, in terms of both AC and cyanobacteria estimation. This is primarily attributed to the high-level feature extraction of SAE, which can represent the spatial features of cyanobacteria. Therefore, this study demonstrated that the deep neural network has a strong potential to realize an integrative remote sensing application.

Separability and geometry of object manifolds in deep neural networks

Stimuli are represented in the brain by the collective population responses of sensory neurons, and an object presented under varying conditions gives rise to a collection of neural population responses called an ‘object manifold’. Changes in the object representation along a hierarchical sensory system are associated with changes in the geometry of those manifolds, and recent theoretical progress connects this geometry with ‘classification capacity’, a quantitative measure of the ability to support object classification. Deep neural networks trained on object classification tasks are a natural testbed for the applicability of this relation. We show how classification capacity improves along the hierarchies of deep neural networks with different architectures. We demonstrate that changes in the geometry of the associated object manifolds underlie this improved capacity, and shed light on the functional roles different levels in the hierarchy play to achieve it, through orchestrated reduction of manifolds’ radius, dimensionality and inter-manifold correlations.

Neural activity space or manifold that represents object information changes across the layers of a deep neural network. Here the authors present a theoretical account of the relationship between the geometry of the manifolds and the classification capacity of the neural networks.

Zebrafish Neuroscience: Using Artificial Neural Networks to Help Understand Brains

Brains are notoriously hard to understand, and neuroscientists need all the tools they can get their hands on to have a realistic shot at it. Advances in machine learning are proving instrumental, illustrated by their recent use to shed light on navigational strategies implemented by zebrafish brains.

Decoding and encoding (de)mixed population responses

A central tenet of neuroscience is that the brain works through large populations of interacting neurons. With recent advances in recording techniques, the inner working of these populations has come into full view. Analyzing the resulting large-scale data sets is challenging because of the often complex and 'mixed' dependency of neural activities on experimental parameters, such as stimuli, decisions, or motor responses. Here we review recent insights gained from analyzing these data with dimensionality reduction methods that 'demix' these dependencies. We demonstrate that the mappings from (carefully chosen) experimental parameters to population activities appear to be typical and stable across tasks, brain areas, and animals, and are often identifiable by linear methods. By considering when and why dimensionality reduction and demixing work well, we argue for a view of population coding in which populations represent (demixed) latent signals, corresponding to stimuli, decisions, motor responses, and so on. These latent signals are encoded into neural population activity via non-linear mappings and decoded via linear readouts. We explain how such a scheme can facilitate the propagation of information across cortical areas, and we review neural network architectures that can reproduce the encoding and decoding of latent signals in population activities. These architectures promise a link from the biophysics of single neurons to the activities of neural populations.