A neural network trained for prediction mimics diverse features of biological neurons and perception

PrediRep: Modeling hierarchical predictive coding with an unsupervised deep learning network

Hierarchical predictive coding (hPC) provides a compelling framework for understanding how the cortex predicts future sensory inputs by minimizing prediction errors through an internal generative model of the external world. Existing deep learning models inspired by hPC incorporate architectural choices that deviate from core hPC principles, potentially limiting their utility for neuroscientific investigations. We introduce PrediRep (Predicting Representations), a novel deep learning network that adheres more closely to architectural principles of hPC. We validate PrediRep by comparing its functional alignment with hPC to that of existing models after being trained on a next-frame prediction task. Our findings demonstrate that PrediRep, particularly when trained with an all-level loss function (PrediRepAll), exhibits high functional alignment with hPC. In contrast to other contemporary deep learning networks inspired by hPC, it consistently processes input-relevant information at higher hierarchical levels and maintains active representations and accurate predictions across all hierarchical levels. Although PrediRep was designed primarily to serve as a model suitable for neuroscientific research rather than to optimize performance, it nevertheless achieves competitive performance in next-frame prediction while utilizing significantly fewer trainable parameters than alternative models. Our results underscore that even minor architectural deviations from neuroscientific theories like hPC can lead to significant functional discrepancies. By faithfully adhering to hPC principles, PrediRep provides a more accurate tool for in silico exploration of cortical phenomena. PrediRep's lightweight and biologically plausible design makes it well-suited for future studies aiming to investigate the neural underpinnings of predictive coding and to derive empirically testable predictions.

Design of Multi-Cancer VOCs Profiling Platform via a Deep Learning-Assisted Sensing Library Screening Strategy

The efficiency of sensor arrays in parallel discrimination of multianalytes is fundamentally influenced by the quantity and performance of the sensor elements. The advent of combinational design has notably accelerated the generation of chemical libraries, offering numerous candidates for the development of robust sensor arrays. However, screening elements with superior cross-responsiveness remains challenging, impeding the development of high-performance sensor arrays. Herein, we propose a new deep learning-assisted, two-step screening strategy to identify the optimal combination of minimal sensor elements, using a designed volatile organic compounds (VOCs)-targeted sensor library. 400 sensing elements constructed by pairing 20 ionizable cationic elements and 20 anionic dyes in the sensor library were employed for various VOCs, generating plentiful color variation data. By employing a feedforward neural network─random forest-recursive feature elimination (FRR) algorithm, sensing elements were effectively screened, resulting in the rapidly producing 8-element and 10-element arrays for two VOC models, both achieving 100% discrimination accuracy. Furthermore, a smartphone-based point-of-care testing (POCT) platform achieved cancer discrimination in a simulated cancer VOC model, using image-based deep learning, demonstrating the rationality and practicality of deep learning in the assembly of sensor elements for parallel sensing platforms.

Brain-like border ownership signals support prediction of natural videos

To make sense of visual scenes, the brain must segment foreground from background. This is thought to be facilitated by neurons that signal border ownership (BOS), which indicate which side of a border in their receptive field is owned by an object. How these signals emerge without a teaching signal of what is foreground remains unclear. Here we find that many units in PredNet, a self-supervised deep neural network trained to predict future frames in natural videos, are selective for BOS. They share key properties with BOS neurons in the brain, including robustness to object transformations and hysteresis. Ablation revealed that BOS units contribute more to prediction than other units for videos with moving objects. Our findings suggest that BOS neurons might emerge due to an evolutionary or developmental pressure to predict future input in natural, complex dynamic environments, even without an explicit requirement to segment foreground from background.

Pre-training artificial neural networks with spontaneous retinal activity improves motion prediction in natural scenes

The ability to process visual stimuli rich with motion represents an essential skill for animal survival and is largely already present at the onset of vision. Although the exact mechanisms underlying its maturation remain elusive, spontaneous activity patterns in the retina, known as retinal waves, have been shown to contribute to this developmental process. Retinal waves exhibit complex spatio-temporal statistics and contribute to the establishment of circuit connectivity and function in the visual system, including the formation of retinotopic maps and the refinement of receptive fields in downstream areas such as the thalamus and visual cortex. Recent work in mice has shown that retinal waves have statistical features matching those of natural visual stimuli, such as optic flow, suggesting that they could prime the visual system for motion processing upon vision onset. Motivated by these findings, we examined whether artificial neural network (ANN) models trained on natural movies show improved performance if pre-trained with retinal waves. We employed the spatio-temporally complex task of next-frame prediction, in which the ANN was trained to predict the next frame based on preceding input frames of a movie. We found that pre-training ANNs with retinal waves enhances the processing of real-world visual stimuli and accelerates learning. Strikingly, when we merely replaced the initial training epochs on naturalistic stimuli with retinal waves, keeping the total training time the same, we still found that an ANN trained on retinal waves temporarily outperforms one trained solely on natural movies. Similar to observations made in biological systems, we also found that pre-training with spontaneous activity refines the receptive field of ANN neurons. Overall, our work sheds light on the functional role of spatio-temporally patterned spontaneous activity in the processing of motion in natural scenes, suggesting it acts as a training signal to prepare the developing visual system for adult visual processing.

A spatiotemporal style transfer algorithm for dynamic visual stimulus generation

Understanding how visual information is encoded in biological and artificial systems often requires the generation of appropriate stimuli to test specific hypotheses, but available methods for video generation are scarce. Here we introduce the spatiotemporal style transfer (STST) algorithm, a dynamic visual stimulus generation framework that allows the manipulation and synthesis of video stimuli for vision research. We show how stimuli can be generated that match the low-level spatiotemporal features of their natural counterparts, but lack their high-level semantic features, providing a useful tool to study object recognition. We used these stimuli to probe PredNet, a predictive coding deep network, and found that its next-frame predictions were not disrupted by the omission of high-level information, with human observers also confirming the preservation of low-level features and lack of high-level information in the generated stimuli. We also introduce a procedure for the independent spatiotemporal factorization of dynamic stimuli. Testing such factorized stimuli on humans and deep vision models suggests a spatial bias in how humans and deep vision models encode dynamic visual information. These results showcase potential applications of the STST algorithm as a versatile tool for dynamic stimulus generation in vision science.

Learning to segment self-generated from externally caused optic flow through sensorimotor mismatch circuits

Efficient sensory detection requires the capacity to ignore task-irrelevant information, for example when optic flow patterns created by egomotion need to be disentangled from object perception. To investigate how this is achieved in the visual system, predictive coding with sensorimotor mismatch detection is an attractive starting point. Indeed, experimental evidence for sensorimotor mismatch signals in early visual areas exists, but it is not understood how they are integrated into cortical networks that perform input segmentation and categorization. Our model advances a biologically plausible solution by extending predictive coding models with the ability to distinguish self-generated from externally caused optic flow. We first show that a simple three neuron circuit produces experience-dependent sensorimotor mismatch responses, in agreement with calcium imaging data from mice. This microcircuit is then integrated into a neural network with two generative streams. The motor-to-visual stream consists of parallel microcircuits between motor and visual areas and learns to spatially predict optic flow resulting from self-motion. The second stream bidirectionally connects a motion-selective higher visual area (mHVA) to V1, assigning a crucial role to the abundant feedback connections to V1: the maintenance of a generative model of externally caused optic flow. In the model, area mHVA learns to segment moving objects from the background, and facilitates object categorization. Based on shared neurocomputational principles across species, the model also maps onto primate visual cortex. Our work extends Hebbian predictive coding to sensorimotor settings, in which the agent actively moves - and learns to predict the consequences of its own movements.

Adaptation optimizes sensory encoding for future stimuli

Sensory neurons continually adapt their response characteristics according to recent stimulus history. However, it is unclear how such a reactive process can benefit the organism. Here, we test the hypothesis that adaptation actually acts proactively in the sense that it optimally adjusts sensory encoding for future stimuli. We first quantified human subjects' ability to discriminate visual orientation under different adaptation conditions. Using an information theoretic analysis, we found that adaptation leads to a reallocation of coding resources such that encoding accuracy peaks at the mean orientation of the adaptor while total coding capacity remains constant. We then asked whether this characteristic change in encoding accuracy is predicted by the temporal statistics of natural visual input. Analyzing the retinal input of freely behaving human subjects showed that the distribution of local visual orientations in the retinal input stream indeed peaks at the mean orientation of the preceding input history (i.e., the adaptor). We further tested our hypothesis by analyzing the internal sensory representations of a recurrent neural network trained to predict the next frame of natural scene videos (PredNet). Simulating our human adaptation experiment with PredNet, we found that the network exhibited the same change in encoding accuracy as observed in human subjects. Taken together, our results suggest that adaptation-induced changes in encoding accuracy prepare the visual system for future stimuli.

Poisson Variational Autoencoder

Variational autoencoders (VAEs) employ Bayesian inference to interpret sensory inputs, mirroring processes that occur in primate vision across both ventral [1] and dorsal [2] pathways. Despite their success, traditional VAEs rely on continuous latent variables, which deviates sharply from the discrete nature of biological neurons. Here, we developed the Poisson VAE ( 𝒫 -VAE), a novel architecture that combines principles of predictive coding with a VAE that encodes inputs into discrete spike counts. Combining Poisson-distributed latent variables with predictive coding introduces a metabolic cost term in the model loss function, suggesting a relationship with sparse coding which we verify empirically. Additionally, we analyze the geometry of learned representations, contrasting the 𝒫 -VAE to alternative VAE models. We find that the 𝒫 -VAE encodes its inputs in relatively higher dimensions, facilitating linear separability of categories in a downstream classification task with a much better (5×) sample efficiency. Our work provides an interpretable computational framework to study brain-like sensory processing and paves the way for a deeper understanding of perception as an inferential process.

Modeling human activity comprehension at human scale: prediction, segmentation, and categorization

Humans form sequences of event models-representations of the current situation-to predict how activity will unfold. Multiple mechanisms have been proposed for how the cognitive system determines when to segment the stream of behavior and switch from one active event model to another. Here, we constructed a computational model that learns knowledge about event classes (event schemas), by combining recurrent neural networks for short-term dynamics with Bayesian inference over event classes for event-to-event transitions. This architecture represents event schemas and uses them to construct a series of event models. This architecture was trained on one pass through 18 h of naturalistic human activities. Another 3.5 h of activities were used to test each variant for agreement with human segmentation and categorization. The architecture was able to learn to predict human activity, and it developed segmentation and categorization approaching human-like performance. We then compared two variants of this architecture designed to better emulate human event segmentation: one transitioned when the active event model produced high uncertainty in its prediction and the other transitioned when the active event model produced a large prediction error. The two variants learned to segment and categorize events, and the prediction uncertainty variant provided a somewhat closer match to human segmentation and categorization-despite being given no feedback about segmentation or categorization. These results suggest that event model transitioning based on prediction uncertainty or prediction error can reproduce two important features of human event comprehension.

Brain-like border ownership signals support prediction of natural videos

To make sense of visual scenes, the brain must segment foreground from background. This is thought to be facilitated by neurons in the primate visual system that encode border ownership (BOS), i.e. whether a local border is part of an object on one or the other side of the border. It is unclear how these signals emerge in neural networks without a teaching signal of what is foreground and background. In this study, we investigated whether BOS signals exist in PredNet, a self-supervised artificial neural network trained to predict the next image frame of natural video sequences. We found that a significant number of units in PredNet are selective for BOS. Moreover these units share several other properties with the BOS neurons in the brain, including robustness to scene variations that constitute common object transformations in natural videos, and hysteresis of BOS signals. Finally, we performed ablation experiments and found that BOS units contribute more to prediction than non-BOS units for videos with moving objects. Our findings indicate that BOS units are especially useful to predict future input in natural videos, even when networks are not required to segment foreground from background. This suggests that BOS neurons in the brain might be the result of evolutionary or developmental pressure to predict future input in natural, complex dynamic visual environments.

Predictive sequence learning in the hippocampal formation

The hippocampus receives sequences of sensory inputs from the cortex during exploration and encodes the sequences with millisecond precision. We developed a predictive autoencoder model of the hippocampus including the trisynaptic and monosynaptic circuits from the entorhinal cortex (EC). CA3 was trained as a self-supervised recurrent neural network to predict its next input. We confirmed that CA3 is predicting ahead by analyzing the spike coupling between simultaneously recorded neurons in the dentate gyrus, CA3, and CA1 of the mouse hippocampus. In the model, CA1 neurons signal prediction errors by comparing CA3 predictions to the next direct EC input. The model exhibits the rapid appearance and slow fading of CA1 place cells and displays replay and phase precession from CA3. The model could be learned in a biologically plausible way with error-encoding neurons. Similarities between the hippocampal and thalamocortical circuits suggest that such computation motif could also underlie self-supervised sequence learning in the cortex.

Artificial intelligence for screening and diagnosis of amyotrophic lateral sclerosis: a systematic review and meta-analysis

INTRODUCTION

Amyotrophic lateral sclerosis (ALS) is a rare and fatal neurological disease that leads to progressive motor function degeneration. Diagnosing ALS is challenging due to the absence of a specific detection test. The use of artificial intelligence (AI) can assist in the investigation and treatment of ALS.

METHODS

We searched seven databases for literature on the application of AI in the early diagnosis and screening of ALS in humans. The findings were summarized using random-effects summary receiver operating characteristic curve. The risk of bias (RoB) analysis was carried out using QUADAS-2 or QUADAS-C tools.

RESULTS

In the 34 analyzed studies, a meta-prevalence of 47% for ALS was noted. For ALS detection, the pooled sensitivity of AI models was 94.3% (95% CI - 63.2% to 99.4%) with a pooled specificity of 98.9% (95% CI - 92.4% to 99.9%). For ALS classification, the pooled sensitivity of AI models was 90.9% (95% CI - 86.5% to 93.9%) with a pooled specificity of 92.3% (95% CI - 84.8% to 96.3%). Based on type of input for classification, the pooled sensitivity of AI models for gait, electromyography, and magnetic resonance signals was 91.2%, 92.6%, and 82.2%, respectively. The pooled specificity for gait, electromyography, and magnetic resonance signals was 94.1%, 96.5%, and 77.3%, respectively.

CONCLUSIONS

Although AI can play a significant role in the screening and diagnosis of ALS due to its high sensitivities and specificities, concerns remain regarding quality of evidence reported in the literature.

PreCNet: Next-Frame Video Prediction Based on Predictive Coding

Predictive coding, currently a highly influential theory in neuroscience, has not been widely adopted in machine learning yet. In this work, we transform the seminal model of Rao and Ballard (1999) into a modern deep learning framework while remaining maximally faithful to the original schema. The resulting network we propose (PreCNet) is tested on a widely used next-frame video prediction benchmark, which consists of images from an urban environment recorded from a car-mounted camera, and achieves state-of-the-art performance. Performance on all measures (MSE, PSNR, and SSIM) was further improved when a larger training set (2M images from BDD100k) pointed to the limitations of the KITTI training set. This work demonstrates that an architecture carefully based on a neuroscience model, without being explicitly tailored to the task at hand, can exhibit exceptional performance.

Predictive coding with spiking neurons and feedforward gist signaling

Predictive coding (PC) is an influential theory in neuroscience, which suggests the existence of a cortical architecture that is constantly generating and updating predictive representations of sensory inputs. Owing to its hierarchical and generative nature, PC has inspired many computational models of perception in the literature. However, the biological plausibility of existing models has not been sufficiently explored due to their use of artificial neurons that approximate neural activity with firing rates in the continuous time domain and propagate signals synchronously. Therefore, we developed a spiking neural network for predictive coding (SNN-PC), in which neurons communicate using event-driven and asynchronous spikes. Adopting the hierarchical structure and Hebbian learning algorithms from previous PC neural network models, SNN-PC introduces two novel features: (1) a fast feedforward sweep from the input to higher areas, which generates a spatially reduced and abstract representation of input (i.e., a neural code for the gist of a scene) and provides a neurobiological alternative to an arbitrary choice of priors; and (2) a separation of positive and negative error-computing neurons, which counters the biological implausibility of a bi-directional error neuron with a very high baseline firing rate. After training with the MNIST handwritten digit dataset, SNN-PC developed hierarchical internal representations and was able to reconstruct samples it had not seen during training. SNN-PC suggests biologically plausible mechanisms by which the brain may perform perceptual inference and learning in an unsupervised manner. In addition, it may be used in neuromorphic applications that can utilize its energy-efficient, event-driven, local learning, and parallel information processing nature.

The cognitive neuroscience of self-awareness: Current framework, clinical implications, and future research directions

Self-awareness, the ability to take oneself as the object of awareness, has been an enigma for our species, with different answers to this question being provided by religion, philosophy, and, more recently, science. The current review aims to discuss the neurocognitive mechanisms underlying self-awareness. The multidimensional nature of self-awareness will be explored, suggesting how it can be thought of as an emergent property observed in different cognitive complexity levels, within a predictive coding approach. A presentation of alterations of self-awareness in neuropsychiatric conditions will ground a discussion on alternative frameworks to understand this phenomenon, in health and psychopathology, with future research directions being indicated to fill current gaps in the literature. This article is categorized under: Philosophy > Consciousness Psychology > Brain Function and Dysfunction Neuroscience > Cognition.

Dynamic predictive coding: A model of hierarchical sequence learning and prediction in the neocortex

We introduce dynamic predictive coding, a hierarchical model of spatiotemporal prediction and sequence learning in the neocortex. The model assumes that higher cortical levels modulate the temporal dynamics of lower levels, correcting their predictions of dynamics using prediction errors. As a result, lower levels form representations that encode sequences at shorter timescales (e.g., a single step) while higher levels form representations that encode sequences at longer timescales (e.g., an entire sequence). We tested this model using a two-level neural network, where the top-down modulation creates low-dimensional combinations of a set of learned temporal dynamics to explain input sequences. When trained on natural videos, the lower-level model neurons developed space-time receptive fields similar to those of simple cells in the primary visual cortex while the higher-level responses spanned longer timescales, mimicking temporal response hierarchies in the cortex. Additionally, the network's hierarchical sequence representation exhibited both predictive and postdictive effects resembling those observed in visual motion processing in humans (e.g., in the flash-lag illusion). When coupled with an associative memory emulating the role of the hippocampus, the model allowed episodic memories to be stored and retrieved, supporting cue-triggered recall of an input sequence similar to activity recall in the visual cortex. When extended to three hierarchical levels, the model learned progressively more abstract temporal representations along the hierarchy. Taken together, our results suggest that cortical processing and learning of sequences can be interpreted as dynamic predictive coding based on a hierarchical spatiotemporal generative model of the visual world.

Neural mechanisms of visual motion extrapolation

Because neural processing takes time, the brain only has delayed access to sensory information. When localising moving objects this is problematic, as an object will have moved on by the time its position has been determined. Here, we consider predictive motion extrapolation as a fundamental delay-compensation strategy. From a population-coding perspective, we outline how extrapolation can be achieved by a forwards shift in the population-level activity distribution. We identify general mechanisms underlying such shifts, involving various asymmetries which facilitate the targeted 'enhancement' and/or 'dampening' of population-level activity. We classify these on the basis of their potential implementation (intra- vs inter-regional processes) and consider specific examples in different visual regions. We consider how motion extrapolation can be achieved during inter-regional signaling, and how asymmetric connectivity patterns which support extrapolation can emerge spontaneously from local synaptic learning rules. Finally, we consider how more abstract 'model-based' predictive strategies might be implemented. Overall, we present an integrative framework for understanding how the brain determines the real-time position of moving objects, despite neural delays.

Reconstructing visual illusory experiences from human brain activity

Visual illusions provide valuable insights into the brain's interpretation of the world given sensory inputs. However, the precise manner in which brain activity translates into illusory experiences remains largely unknown. Here, we leverage a brain decoding technique combined with deep neural network (DNN) representations to reconstruct illusory percepts as images from brain activity. The reconstruction model was trained on natural images to establish a link between brain activity and perceptual features and then tested on two types of illusions: illusory lines and neon color spreading. Reconstructions revealed lines and colors consistent with illusory experiences, which varied across the source visual cortical areas. This framework offers a way to materialize subjective experiences, shedding light on the brain's internal representations of the world.

Predictive learning by a burst-dependent learning rule

Humans and other animals are able to quickly generalize latent dynamics of spatiotemporal sequences, often from a minimal number of previous experiences. Additionally, internal representations of external stimuli must remain stable, even in the presence of sensory noise, in order to be useful for informing behavior. In contrast, typical machine learning approaches require many thousands of samples, and generalize poorly to unexperienced examples, or fail completely to predict at long timescales. Here, we propose a novel neural network module which incorporates hierarchy and recurrent feedback terms, constituting a simplified model of neocortical microcircuits. This microcircuit predicts spatiotemporal trajectories at the input layer using a temporal error minimization algorithm. We show that this module is able to predict with higher accuracy into the future compared to traditional models. Investigating this model we find that successive predictive models learn representations which are increasingly removed from the raw sensory space, namely as successive temporal derivatives of the positional information. Next, we introduce a spiking neural network model which implements the rate-model through the use of a recently proposed biological learning rule utilizing dual-compartment neurons. We show that this network performs well on the same tasks as the mean-field models, by developing intrinsic dynamics that follow the dynamics of the external stimulus, while coordinating transmission of higher-order dynamics. Taken as a whole, these findings suggest that hierarchical temporal abstraction of sequences, rather than feed-forward reconstruction, may be responsible for the ability of neural systems to quickly adapt to novel situations.

Hierarchical temporal prediction captures motion processing along the visual pathway

Visual neurons respond selectively to features that become increasingly complex from the eyes to the cortex. Retinal neurons prefer flashing spots of light, primary visual cortical (V1) neurons prefer moving bars, and those in higher cortical areas favor complex features like moving textures. Previously, we showed that V1 simple cell tuning can be accounted for by a basic model implementing temporal prediction - representing features that predict future sensory input from past input (Singer et al., 2018). Here, we show that hierarchical application of temporal prediction can capture how tuning properties change across at least two levels of the visual system. This suggests that the brain does not efficiently represent all incoming information; instead, it selectively represents sensory inputs that help in predicting the future. When applied hierarchically, temporal prediction extracts time-varying features that depend on increasingly high-level statistics of the sensory input.

Artificial-Neural-Network-Driven Innovations in Time-Varying Process Diagnosis of Low-K Oxide Deposition

To address the challenges in real-time process diagnosis within the semiconductor manufacturing industry, this paper presents a novel machine learning approach for analyzing the time-varying 10th harmonics during the deposition of low-k oxide (SiOF) on a 600 Å undoped silicate glass thin liner using a high-density plasma chemical vapor deposition system. The 10th harmonics, which are high-frequency components 10 times the fundamental frequency, are generated in the plasma sheath because of their nonlinear nature. An artificial neural network with a three-hidden-layer architecture was applied and optimized using k-fold cross-validation to analyze the harmonics generated in the plasma sheath during the deposition process. The model exhibited a binary cross-entropy loss of 0.1277 and achieved an accuracy of 0.9461. This approach enables the accurate prediction of process performance, resulting in significant cost reduction and enhancement of semiconductor manufacturing processes. This model has the potential to improve defect control and yield, thereby benefiting the semiconductor industry. Despite the limitations imposed by the limited dataset, the model demonstrated promising results, and further performance improvements are anticipated with the inclusion of additional data in future studies.

Local minimization of prediction errors drives learning of invariant object representations in a generative network model of visual perception

The ventral visual processing hierarchy of the cortex needs to fulfill at least two key functions: perceived objects must be mapped to high-level representations invariantly of the precise viewing conditions, and a generative model must be learned that allows, for instance, to fill in occluded information guided by visual experience. Here, we show how a multilayered predictive coding network can learn to recognize objects from the bottom up and to generate specific representations via a top-down pathway through a single learning rule: the local minimization of prediction errors. Trained on sequences of continuously transformed objects, neurons in the highest network area become tuned to object identity invariant of precise position, comparable to inferotemporal neurons in macaques. Drawing on this, the dynamic properties of invariant object representations reproduce experimentally observed hierarchies of timescales from low to high levels of the ventral processing stream. The predicted faster decorrelation of error-neuron activity compared to representation neurons is of relevance for the experimental search for neural correlates of prediction errors. Lastly, the generative capacity of the network is confirmed by reconstructing specific object images, robust to partial occlusion of the inputs. By learning invariance from temporal continuity within a generative model, the approach generalizes the predictive coding framework to dynamic inputs in a more biologically plausible way than self-supervised networks with non-local error-backpropagation. This was achieved simply by shifting the training paradigm to dynamic inputs, with little change in architecture and learning rule from static input-reconstructing Hebbian predictive coding networks.

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.

Learning cortical hierarchies with temporal Hebbian updates

A key driver of mammalian intelligence is the ability to represent incoming sensory information across multiple abstraction levels. For example, in the visual ventral stream, incoming signals are first represented as low-level edge filters and then transformed into high-level object representations. Similar hierarchical structures routinely emerge in artificial neural networks (ANNs) trained for object recognition tasks, suggesting that similar structures may underlie biological neural networks. However, the classical ANN training algorithm, backpropagation, is considered biologically implausible, and thus alternative biologically plausible training methods have been developed such as Equilibrium Propagation, Deep Feedback Control, Supervised Predictive Coding, and Dendritic Error Backpropagation. Several of those models propose that local errors are calculated for each neuron by comparing apical and somatic activities. Notwithstanding, from a neuroscience perspective, it is not clear how a neuron could compare compartmental signals. Here, we propose a solution to this problem in that we let the apical feedback signal change the postsynaptic firing rate and combine this with a differential Hebbian update, a rate-based version of classical spiking time-dependent plasticity (STDP). We prove that weight updates of this form minimize two alternative loss functions that we prove to be equivalent to the error-based losses used in machine learning: the inference latency and the amount of top-down feedback necessary. Moreover, we show that the use of differential Hebbian updates works similarly well in other feedback-based deep learning frameworks such as Predictive Coding or Equilibrium Propagation. Finally, our work removes a key requirement of biologically plausible models for deep learning and proposes a learning mechanism that would explain how temporal Hebbian learning rules can implement supervised hierarchical learning.

Process optimization of line patterns in extreme ultraviolet lithography using machine learning and a simulated annealing algorithm

Resolution, line edge/width roughness, and sensitivity (RLS) are critical indicators for evaluating the imaging performance of resists. As the technology node gradually shrinks, stricter indicator control is required for high-resolution imaging. However, current research can improve only part of the RLS indicators of resists for line patterns, and it is difficult to improve the overall imaging performance of resists in extreme ultraviolet lithography. Here, we report a lithographic process optimization system of line patterns, where RLS models are first established by adopting a machine learning method, and then these models are optimized using a simulated annealing algorithm. Finally, the process parameter combination with optimal imaging quality of line patterns can be obtained. This system can control resist RLS indicators, and it exhibits high optimization accuracy, which facilitates the reduction of process optimization time and cost and accelerates the development of the lithography process.

A time-causal and time-recursive scale-covariant scale-space representation of temporal signals and past time

This article presents an overview of a theory for performing temporal smoothing on temporal signals in such a way that: (i) temporally smoothed signals at coarser temporal scales are guaranteed to constitute simplifications of corresponding temporally smoothed signals at any finer temporal scale (including the original signal) and (ii) the temporal smoothing process is both time-causal and time-recursive, in the sense that it does not require access to future information and can be performed with no other temporal memory buffer of the past than the resulting smoothed temporal scale-space representations themselves. For specific subsets of parameter settings for the classes of linear and shift-invariant temporal smoothing operators that obey this property, it is shown how temporal scale covariance can be additionally obtained, guaranteeing that if the temporal input signal is rescaled by a uniform temporal scaling factor, then also the resulting temporal scale-space representations of the rescaled temporal signal will constitute mere rescalings of the temporal scale-space representations of the original input signal, complemented by a shift along the temporal scale dimension. The resulting time-causal limit kernel that obeys this property constitutes a canonical temporal kernel for processing temporal signals in real-time scenarios when the regular Gaussian kernel cannot be used, because of its non-causal access to information from the future, and we cannot additionally require the temporal smoothing process to comprise a complementary memory of the past beyond the information contained in the temporal smoothing process itself, which in this way also serves as a multi-scale temporal memory of the past. We describe how the time-causal limit kernel relates to previously used temporal models, such as Koenderink's scale-time kernels and the ex-Gaussian kernel. We do also give an overview of how the time-causal limit kernel can be used for modelling the temporal processing in models for spatio-temporal and spectro-temporal receptive fields, and how it more generally has a high potential for modelling neural temporal response functions in a purely time-causal and time-recursive way, that can also handle phenomena at multiple temporal scales in a theoretically well-founded manner. We detail how this theory can be efficiently implemented for discrete data, in terms of a set of recursive filters coupled in cascade. Hence, the theory is generally applicable for both: (i) modelling continuous temporal phenomena over multiple temporal scales and (ii) digital processing of measured temporal signals in real time. We conclude by stating implications of the theory for modelling temporal phenomena in biological, perceptual, neural and memory processes by mathematical models, as well as implications regarding the philosophy of time and perceptual agents. Specifically, we propose that for A-type theories of time, as well as for perceptual agents, the notion of a non-infinitesimal inner temporal scale of the temporal receptive fields has to be included in representations of the present, where the inherent nonzero temporal delay of such time-causal receptive fields implies a need for incorporating predictions from the actual time-delayed present in the layers of a perceptual hierarchy, to make it possible for a representation of the perceptual present to constitute a representation of the environment with timing properties closer to the actual present.

Challenging deep learning models with image distortion based on the abutting grating illusion

Even state-of-the-art deep learning models lack fundamental abilities compared with humans. While many image distortions have been proposed to compare deep learning with humans, they depend on mathematical transformations instead of human cognitive functions. Here, we propose an image distortion based on the abutting grating illusion, which is a phenomenon discovered in humans and animals. The distortion generates illusory contour perception using line gratings abutting each other. We applied the method to MNIST, high-resolution MNIST, and "16-class-ImageNet" silhouettes. Many models, including models trained from scratch and 109 models pretrained with ImageNet or various data augmentation techniques, were tested. Our results show that abutting grating distortion is challenging even for state-of-the-art deep learning models. We discovered that DeepAugment models outperformed other pretrained models. Visualization of early layers indicates that better-performing models exhibit the endstopping property, which is consistent with neuroscience discoveries. Twenty-four human subjects classified distorted samples to validate the distortion.

Inconsistent illusory motion in predictive coding deep neural networks

Why do we perceive illusory motion in some static images? Several accounts point to eye movements, response latencies to different image elements, or interactions between image patterns and motion energy detectors. Recently PredNet, a recurrent deep neural network (DNN) based on predictive coding principles, was reported to reproduce the "Rotating Snakes" illusion, suggesting a role for predictive coding. We begin by replicating this finding, then use a series of "in silico" psychophysics and electrophysiology experiments to examine whether PredNet behaves consistently with human observers and non-human primate neural data. A pretrained PredNet predicted illusory motion for all subcomponents of the Rotating Snakes pattern, consistent with human observers. However, we found no simple response delays in internal units, unlike evidence from electrophysiological data. PredNet's detection of motion in gradients seemed dependent on contrast, but depends predominantly on luminance in humans. Finally, we examined the robustness of the illusion across ten PredNets of identical architecture, retrained on the same video data. There was large variation across network instances in whether they reproduced the Rotating Snakes illusion, and what motion, if any, they predicted for simplified variants. Unlike human observers, no network predicted motion for greyscale variants of the Rotating Snakes pattern. Our results sound a cautionary note: even when a DNN successfully reproduces some idiosyncrasy of human vision, more detailed investigation can reveal inconsistencies between humans and the network, and between different instances of the same network. These inconsistencies suggest that predictive coding does not reliably give rise to human-like illusory motion.

Process optimization of contact hole patterns via a simulated annealing algorithm in extreme ultraviolet lithography

The critical dimension (CD), roughness, and sensitivity are extremely significant indicators for evaluating the imaging performance of photoresists in extreme ultraviolet lithography. As the CD gradually shrinks, tighter indicator control is required for high fidelity imaging. However, current research primarily focuses on the optimization of one indicator of one-dimensional line patterns, and little attention has been paid to two-dimensional patterns. Here, we report an image quality optimization method of two-dimensional contact holes. This method takes horizontal and vertical contact widths, contact edge roughness, and sensitivity as evaluation indicators, and uses machine learning to establish the corresponding relationship between process parameters and each indicator. Then, the simulated annealing algorithm is applied to search for the optimal process parameters, and finally, a set of process parameters with optimum image quality is obtained. Rigorous imaging results of lithography demonstrate that this method has very high optimization accuracy and can improve the overall performance of the device, dramatically accelerating the development of the lithography process.

Understanding Human Object Vision: A Picture Is Worth a Thousand Representations

Objects are the core meaningful elements in our visual environment. Classic theories of object vision focus upon object recognition and are elegant and simple. Some of their proposals still stand, yet the simplicity is gone. Recent evolutions in behavioral paradigms, neuroscientific methods, and computational modeling have allowed vision scientists to uncover the complexity of the multidimensional representational space that underlies object vision. We review these findings and propose that the key to understanding this complexity is to relate object vision to the full repertoire of behavioral goals that underlie human behavior, running far beyond object recognition. There might be no such thing as core object recognition, and if it exists, then its importance is more limited than traditionally thought.

Reconstructing the predictive architecture of the mind and brain

Predictive processing has become an influential framework in cognitive neuroscience. However, it often lacks specificity and direct empirical support. How can we probe the nature and limits of the predictive brain? We highlight the potential of recent advances in artificial intelligence (AI) for providing a richer and more computationally explicit test of this theory of cortical function.

Deep problems with neural network models of human vision

Deep neural networks (DNNs) have had extraordinary successes in classifying photographic images of objects and are often described as the best models of biological vision. This conclusion is largely based on three sets of findings: (1) DNNs are more accurate than any other model in classifying images taken from various datasets, (2) DNNs do the best job in predicting the pattern of human errors in classifying objects taken from various behavioral datasets, and (3) DNNs do the best job in predicting brain signals in response to images taken from various brain datasets (e.g., single cell responses or fMRI data). However, these behavioral and brain datasets do not test hypotheses regarding what features are contributing to good predictions and we show that the predictions may be mediated by DNNs that share little overlap with biological vision. More problematically, we show that DNNs account for almost no results from psychological research. This contradicts the common claim that DNNs are good, let alone the best, models of human object recognition. We argue that theorists interested in developing biologically plausible models of human vision need to direct their attention to explaining psychological findings. More generally, theorists need to build models that explain the results of experiments that manipulate independent variables designed to test hypotheses rather than compete on making the best predictions. We conclude by briefly summarizing various promising modeling approaches that focus on psychological data.

Predictive coding is a consequence of energy efficiency in recurrent neural networks

Predictive coding is a promising framework for understanding brain function. It postulates that the brain continuously inhibits predictable sensory input, ensuring preferential processing of surprising elements. A central aspect of this view is its hierarchical connectivity, involving recurrent message passing between excitatory bottom-up signals and inhibitory top-down feedback. Here we use computational modeling to demonstrate that such architectural hardwiring is not necessary. Rather, predictive coding is shown to emerge as a consequence of energy efficiency. When training recurrent neural networks to minimize their energy consumption while operating in predictive environments, the networks self-organize into prediction and error units with appropriate inhibitory and excitatory interconnections and learn to inhibit predictable sensory input. Moving beyond the view of purely top-down-driven predictions, we demonstrate, via virtual lesioning experiments, that networks perform predictions on two timescales: fast lateral predictions among sensory units and slower prediction cycles that integrate evidence over time.

Neural representational geometry underlies few-shot concept learning

Understanding the neural basis of the remarkable human cognitive capacity to learn novel concepts from just one or a few sensory experiences constitutes a fundamental problem. We propose a simple, biologically plausible, mathematically tractable, and computationally powerful neural mechanism for few-shot learning of naturalistic concepts. We posit that the concepts that can be learned from few examples are defined by tightly circumscribed manifolds in the neural firing-rate space of higher-order sensory areas. We further posit that a single plastic downstream readout neuron learns to discriminate new concepts based on few examples using a simple plasticity rule. We demonstrate the computational power of our proposal by showing that it can achieve high few-shot learning accuracy on natural visual concepts using both macaque inferotemporal cortex representations and deep neural network (DNN) models of these representations and can even learn novel visual concepts specified only through linguistic descriptors. Moreover, we develop a mathematical theory of few-shot learning that links neurophysiology to predictions about behavioral outcomes by delineating several fundamental and measurable geometric properties of neural representations that can accurately predict the few-shot learning performance of naturalistic concepts across all our numerical simulations. This theory reveals, for instance, that high-dimensional manifolds enhance the ability to learn new concepts from few examples. Intriguingly, we observe striking mismatches between the geometry of manifolds in the primate visual pathway and in trained DNNs. We discuss testable predictions of our theory for psychophysics and neurophysiological experiments.

Analysis and Dynamic Monitoring of Indoor Air Quality Based on Laser-Induced Breakdown Spectroscopy and Machine Learning

The air quality of the living area influences human health to a certain extent. Therefore, it is particularly important to detect the quality of indoor air. However, traditional detection methods mainly depend on chemical analysis, which has long been criticized for its high time cost. In this research, a rapid air detection method for the indoor environment using laser-induced breakdown spectroscopy (LIBS) and machine learning was proposed. Four common scenes were simulated, including burning carbon, burning incense, spraying perfume and hot shower which often led to indoor air quality changes. Two steps of spectral measurements and algorithm analysis were used in the experiment. Moreover, the proposed method was found to be effective in distinguishing different kinds of aerosols and presenting sensitivity to the air compositions. In this paper, the signal was isolated by the forest, so the singular values were filtered out. Meanwhile, the spectra of different scenarios were analyzed via the principal component analysis (PCA), and the air environment was classified by K-Nearest Neighbor (KNN) algorithm with an accuracy of 99.2%. Moreover, based on the establishment of a high-precision quantitative detection model, a back propagation (BP) neural network was introduced to improve the robustness and accuracy of indoor environment. The results show that by taking this method, the dynamic prediction of elements concentration can be realized, and its recognition accuracy is 96.5%.

A Layered Spiking Neural System for Classification Problems

Biological brains have a natural capacity for resolving certain classification tasks. Studies on biologically plausible spiking neurons, architectures and mechanisms of artificial neural systems that closely match biological observations while giving high classification performance are gaining momentum. Spiking neural P systems (SN P systems) are a class of membrane computing models and third-generation neural networks that are based on the behavior of biological neural cells and have been used in various engineering applications. Furthermore, SN P systems are characterized by a highly flexible structure that enables the design of a machine learning algorithm by mimicking the structure and behavior of biological cells without the over-simplification present in neural networks. Based on this aspect, this paper proposes a novel type of SN P system, namely, layered SN P system (LSN P system), to solve classification problems by supervised learning. The proposed LSN P system consists of a multi-layer network containing multiple weighted fuzzy SN P systems with adaptive weight adjustment rules. The proposed system employs specific ascending dimension techniques and a selection method of output neurons for classification problems. The experimental results obtained using benchmark datasets from the UCI machine learning repository and MNIST dataset demonstrated the feasibility and effectiveness of the proposed LSN P system. More importantly, the proposed LSN P system presents the first SN P system that demonstrates sufficient performance for use in addressing real-world classification problems.

Motion illusion-like patterns extracted from photo and art images using predictive deep neural networks

In our previous study, we successfully reproduced the illusory motion perceived in the rotating snakes illusion using deep neural networks incorporating predictive coding theory. In the present study, we further examined the properties of the network using a set of 1500 images, including ordinary static images of paintings and photographs and images of various types of motion illusions. Results showed that the networks clearly classified a group of illusory images and others and reproduced illusory motions against various types of illusions similar to human perception. Notably, the networks occasionally detected anomalous motion vectors, even in ordinally static images where humans were unable to perceive any illusory motion. Additionally, illusion-like designs with repeating patterns were generated using areas where anomalous vectors were detected, and psychophysical experiments were conducted, in which illusory motion perception in the generated designs was detected. The observed inaccuracy of the networks will provide useful information for further understanding information processing associated with human vision.

A computational examination of the two-streams hypothesis: which pathway needs a longer memory?

The two visual streams hypothesis is a robust example of neural functional specialization that has inspired countless studies over the past four decades. According to one prominent version of the theory, the fundamental goal of the dorsal visual pathway is the transformation of retinal information for visually-guided motor behavior. To that end, the dorsal stream processes input using absolute (or veridical) metrics only when the movement is initiated, necessitating very little, or no, memory. Conversely, because the ventral visual pathway does not involve motor behavior (its output does not influence the real world), the ventral stream processes input using relative (or illusory) metrics and can accumulate or integrate sensory evidence over long time constants, which provides a substantial capacity for memory. In this study, we tested these relations between functional specialization, processing metrics, and memory by training identical recurrent neural networks to perform either a viewpoint-invariant object classification task or an orientation/size determination task. The former task relies on relative metrics, benefits from accumulating sensory evidence, and is usually attributed to the ventral stream. The latter task relies on absolute metrics, can be computed accurately in the moment, and is usually attributed to the dorsal stream. To quantify the amount of memory required for each task, we chose two types of neural network models. Using a long-short-term memory (LSTM) recurrent network, we found that viewpoint-invariant object categorization (object task) required a longer memory than orientation/size determination (orientation task). Additionally, to dissect this memory effect, we considered factors that contributed to longer memory in object tasks. First, we used two different sets of objects, one with self-occlusion of features and one without. Second, we defined object classes either strictly by visual feature similarity or (more liberally) by semantic label. The models required greater memory when features were self-occluded and when object classes were defined by visual feature similarity, showing that self-occlusion and visual similarity among object task samples are contributing to having a long memory. The same set of tasks modeled using modified leaky-integrator echo state recurrent networks (LiESN), however, did not replicate the results, except under some conditions. This may be because LiESNs cannot perform fine-grained memory adjustments due to their network-wide memory coefficient and fixed recurrent weights. In sum, the LSTM simulations suggest that longer memory is advantageous for performing viewpoint-invariant object classification (a putative ventral stream function) because it allows for interpolation of features across viewpoints. The results further suggest that orientation/size determination (a putative dorsal stream function) does not benefit from longer memory. These findings are consistent with the two visual streams theory of functional specialization.

SUPPLEMENTARY INFORMATION

The online version contains supplementary material available at 10.1007/s11571-021-09703-z.

A self-supervised domain-general learning framework for human ventral stream representation

Anterior regions of the ventral visual stream encode substantial information about object categories. Are top-down category-level forces critical for arriving at this representation, or can this representation be formed purely through domain-general learning of natural image structure? Here we present a fully self-supervised model which learns to represent individual images, rather than categories, such that views of the same image are embedded nearby in a low-dimensional feature space, distinctly from other recently encountered views. We find that category information implicitly emerges in the local similarity structure of this feature space. Further, these models learn hierarchical features which capture the structure of brain responses across the human ventral visual stream, on par with category-supervised models. These results provide computational support for a domain-general framework guiding the formation of visual representation, where the proximate goal is not explicitly about category information, but is instead to learn unique, compressed descriptions of the visual world.

ImageNet-trained deep neural networks exhibit illusion-like response to the Scintillating grid

Deep neural network (DNN) models for computer vision are capable of human-level object recognition. Consequently, similarities between DNN and human vision are of interest. Here, we characterize DNN representations of Scintillating grid visual illusion images in which white disks are perceived to be partially black. Specifically, we use VGG-19 and ResNet-101 DNN models that were trained for image classification and consider the representational dissimilarity (\(L^1\) distance in the penultimate layer) between pairs of images: one with white Scintillating grid disks and the other with disks of decreasing luminance levels. Results showed a nonmonotonic relation, such that decreasing disk luminance led to an increase and subsequently a decrease in representational dissimilarity. That is, the Scintillating grid image with white disks was closer, in terms of the representation, to images with black disks than images with gray disks. In control nonillusion images, such nonmonotonicity was rare. These results suggest that nonmonotonicity in a deep computational representation is a potential test for illusion-like response geometry in DNN models.

Learning About the World by Learning About Images

One of the deepest insights in neuroscience is that sensory encoding should take advantage of statistical regularities. Humans’ visual experience contains many redundancies: Scenes mostly stay the same from moment to moment, and nearby image locations usually have similar colors. A visual system that knows which regularities shape natural images can exploit them to encode scenes compactly or guess what will happen next. Although these principles have been appreciated for more than 60 years, until recently it has been possible to convert them into explicit models only for the earliest stages of visual processing. But recent advances in unsupervised deep learning have changed that. Neural networks can be taught to compress images or make predictions in space or time. In the process, they learn the statistical regularities that structure images, which in turn often reflect physical objects and processes in the outside world. The astonishing accomplishments of unsupervised deep learning reaffirm the importance of learning statistical regularities for sensory coding and provide a coherent framework for how knowledge of the outside world gets into visual cortex.

Unsupervised neural network models of the ventral visual stream

Deep neural networks currently provide the best quantitative models of the response patterns of neurons throughout the primate ventral visual stream. However, such networks have remained implausible as a model of the development of the ventral stream, in part because they are trained with supervised methods requiring many more labels than are accessible to infants during development. Here, we report that recent rapid progress in unsupervised learning has largely closed this gap. We find that neural network models learned with deep unsupervised contrastive embedding methods achieve neural prediction accuracy in multiple ventral visual cortical areas that equals or exceeds that of models derived using today's best supervised methods and that the mapping of these neural network models' hidden layers is neuroanatomically consistent across the ventral stream. Strikingly, we find that these methods produce brain-like representations even when trained solely with real human child developmental data collected from head-mounted cameras, despite the fact that these datasets are noisy and limited. We also find that semisupervised deep contrastive embeddings can leverage small numbers of labeled examples to produce representations with substantially improved error-pattern consistency to human behavior. Taken together, these results illustrate a use of unsupervised learning to provide a quantitative model of a multiarea cortical brain system and present a strong candidate for a biologically plausible computational theory of primate sensory learning.

Integration of AI and traditional medicine in drug discovery

AI integration in plant-based traditional medicine could be used to overcome drug discovery challenges.

DNNBrain: A Unifying Toolbox for Mapping Deep Neural Networks and Brains

Deep neural networks (DNNs) have attained human-level performance on dozens of challenging tasks via an end-to-end deep learning strategy. Deep learning allows data representations that have multiple levels of abstraction; however, it does not explicitly provide any insights into the internal operations of DNNs. Deep learning's success is appealing to neuroscientists not only as a method for applying DNNs to model biological neural systems but also as a means of adopting concepts and methods from cognitive neuroscience to understand the internal representations of DNNs. Although general deep learning frameworks, such as PyTorch and TensorFlow, could be used to allow such cross-disciplinary investigations, the use of these frameworks typically requires high-level programming expertise and comprehensive mathematical knowledge. A toolbox specifically designed as a mechanism for cognitive neuroscientists to map both DNNs and brains is urgently needed. Here, we present DNNBrain, a Python-based toolbox designed for exploring the internal representations of DNNs as well as brains. Through the integration of DNN software packages and well-established brain imaging tools, DNNBrain provides application programming and command line interfaces for a variety of research scenarios. These include extracting DNN activation, probing and visualizing DNN representations, and mapping DNN representations onto the brain. We expect that our toolbox will accelerate scientific research by both applying DNNs to model biological neural systems and utilizing paradigms of cognitive neuroscience to unveil the black box of DNNs.

Incorporating intrinsic suppression in deep neural networks captures dynamics of adaptation in neurophysiology and perception

Adaptation is a fundamental property of sensory systems that can change subjective experiences in the context of recent information. Adaptation has been postulated to arise from recurrent circuit mechanisms or as a consequence of neuronally intrinsic suppression. However, it is unclear whether intrinsic suppression by itself can account for effects beyond reduced responses. Here, we test the hypothesis that complex adaptation phenomena can emerge from intrinsic suppression cascading through a feedforward model of visual processing. A deep convolutional neural network with intrinsic suppression captured neural signatures of adaptation including novelty detection, enhancement, and tuning curve shifts, while producing aftereffects consistent with human perception. When adaptation was trained in a task where repeated input affects recognition performance, an intrinsic mechanism generalized better than a recurrent neural network. Our results demonstrate that feedforward propagation of intrinsic suppression changes the functional state of the network, reproducing key neurophysiological and perceptual properties of adaptation.