Gradient-free training of recurrent neural networks using random perturbations.

Recurrent neural networks (RNNs) hold immense potential for computations due to their Turing completeness and sequential processing capabilities, yet existing methods for their training encounter efficiency challenges. Backpropagation through time (BPTT), the prevailing method, extends the backpropagation (BP) algorithm by unrolling the RNN over time. However, this approach suffers from significant drawbacks, including the need to interleave forward and backward phases and store exact gradient information. Furthermore, BPTT has been shown to struggle to propagate gradient information for long sequences, leading to vanishing gradients. An alternative strategy to using gradient-based methods like BPTT involves stochastically approximating gradients through perturbation-based methods. This learning approach is exceptionally simple, necessitating only forward passes in the network and a global reinforcement signal as feedback. Despite its simplicity, the random nature of its updates typically leads to inefficient optimization, limiting its effectiveness in training neural networks. In this study, we present a new approach to perturbation-based learning in RNNs whose performance is competitive with BPTT, while maintaining the inherent advantages over gradient-based learning. To this end, we extend the recently introduced activity-based node perturbation (ANP) method to operate in the time domain, leading to more efficient learning and generalization. We subsequently conduct a range of experiments to validate our approach. Our results show similar performance, convergence time and scalability when compared to BPTT, strongly outperforming standard node perturbation and weight perturbation methods. These findings suggest that perturbation-based learning methods offer a versatile alternative to gradient-based methods for training RNNs which can be ideally suited for neuromorphic computing applications.

Brain2GAN: Feature-disentangled neural encoding and decoding of visual perception in the primate brain.

A challenging goal of neural coding is to characterize the neural representations underlying visual perception. To this end, multi-unit activity (MUA) of macaque visual cortex was recorded in a passive fixation task upon presentation of faces and natural images. We analyzed the relationship between MUA and latent representations of state-of-the-art deep generative models, including the conventional and feature-disentangled representations of generative adversarial networks (GANs) (i.e., z- and w-latents of StyleGAN, respectively) and language-contrastive representations of latent diffusion networks (i.e., CLIP-latents of Stable Diffusion). A mass univariate neural encoding analysis of the latent representations showed that feature-disentangled w representations outperform both z and CLIP representations in explaining neural responses. Further, w-latent features were found to be positioned at the higher end of the complexity gradient which indicates that they capture visual information relevant to high-level neural activity. Subsequently, a multivariate neural decoding analysis of the feature-disentangled representations resulted in state-of-the-art spatiotemporal reconstructions of visual perception. Taken together, our results not only highlight the important role of feature-disentanglement in shaping high-level neural representations underlying visual perception but also serve as an important benchmark for the future of neural coding.

Gaze-contingent processing improves mobility, scene recognition and visual search in simulated head-steered prosthetic vision.

Objective.The enabling technology of visual prosthetics for the blind is making rapid progress. However, there are still uncertainties regarding the functional outcomes, which can depend on many design choices in the development. In visual prostheses with a head-mounted camera, a particularly challenging question is how to deal with the gaze-locked visual percept associated with spatial updating conflicts in the brain. The current study investigates a recently proposed compensation strategy based on gaze-contingent image processing with eye-tracking. Gaze-contingent processing is expected to reinforce natural-like visual scanning and reestablished spatial updating based on eye movements. The beneficial effects remain to be investigated for daily life activities in complex visual environments.Approach.The current study evaluates the benefits of gaze-contingent processing versus gaze-locked and gaze-ignored simulations in the context of mobility, scene recognition and visual search, using a virtual reality simulated prosthetic vision paradigm with sighted subjects.Main results.Compared to gaze-locked vision, gaze-contingent processing was consistently found to improve the speed in all experimental tasks, as well as the subjective quality of vision. Similar or further improvements were found in a control condition that ignores gaze-dependent effects, a simulation that is unattainable in the clinical reality.Significance.Our results suggest that gaze-locked vision and spatial updating conflicts can be debilitating for complex visually-guided activities of daily living such as mobility and orientation. Therefore, for prospective users of head-steered prostheses with an unimpaired oculomotor system, the inclusion of a compensatory eye-tracking system is strongly endorsed.

Towards biologically plausible phosphene simulation for the differentiable optimization of visual cortical prostheses.

Blindness affects millions of people around the world. A promising solution to restoring a form of vision for some individuals are cortical visual prostheses, which bypass part of the impaired visual pathway by converting camera input to electrical stimulation of the visual system. The artificially induced visual percept (a pattern of localized light flashes, or 'phosphenes') has limited resolution, and a great portion of the field's research is devoted to optimizing the efficacy, efficiency, and practical usefulness of the encoding of visual information. A commonly exploited method is non-invasive functional evaluation in sighted subjects or with computational models by using simulated prosthetic vision (SPV) pipelines. An important challenge in this approach is to balance enhanced perceptual realism, biologically plausibility, and real-time performance in the simulation of cortical prosthetic vision. We present a biologically plausible, PyTorch-based phosphene simulator that can run in real-time and uses differentiable operations to allow for gradient-based computational optimization of phosphene encoding models. The simulator integrates a wide range of clinical results with neurophysiological evidence in humans and non-human primates. The pipeline includes a model of the retinotopic organization and cortical magnification of the visual cortex. Moreover, the quantitative effects of stimulation parameters and temporal dynamics on phosphene characteristics are incorporated. Our results demonstrate the simulator's suitability for both computational applications such as end-to-end deep learning-based prosthetic vision optimization as well as behavioral experiments. The modular and open-source software provides a flexible simulation framework for computational, clinical, and behavioral neuroscientists working on visual neuroprosthetics.

Deep neural networks are not a single hypothesis but a language for expressing computational hypotheses.

An ideal vision model accounts for behavior and neurophysiology in both naturalistic conditions and designed lab experiments. Unlike psychological theories, artificial neural networks (ANNs) actually perform visual tasks and generate testable predictions for arbitrary inputs. These advantages enable ANNs to engage the entire spectrum of the evidence. Failures of particular models drive progress in a vibrant ANN research program of human vision.

PhenoScore quantifies phenotypic variation for rare genetic diseases by combining facial analysis with other clinical features using a machine-learning framework.

Several molecular and phenotypic algorithms exist that establish genotype-phenotype correlations, including facial recognition tools. However, no unified framework that investigates both facial data and other phenotypic data directly from individuals exists. We developed PhenoScore: an open-source, artificial intelligence-based phenomics framework, combining facial recognition technology with Human Phenotype Ontology data analysis to quantify phenotypic similarity. Here we show PhenoScore's ability to recognize distinct phenotypic entities by establishing recognizable phenotypes for 37 of 40 investigated syndromes against clinical features observed in individuals with other neurodevelopmental disorders and show it is an improvement on existing approaches. PhenoScore provides predictions for individuals with variants of unknown significance and enables sophisticated genotype-phenotype studies by testing hypotheses on possible phenotypic (sub)groups. PhenoScore confirmed previously known phenotypic subgroups caused by variants in the same gene for SATB1, SETBP1 and DEAF1 and provides objective clinical evidence for two distinct ADNP-related phenotypes, already established functionally.

Direct speech reconstruction from sensorimotor brain activity with optimized deep learning models.

Objective.Development of brain-computer interface (BCI) technology is key for enabling communication in individuals who have lost the faculty of speech due to severe motor paralysis. A BCI control strategy that is gaining attention employs speech decoding from neural data. Recent studies have shown that a combination of direct neural recordings and advanced computational models can provide promising results. Understanding which decoding strategies deliver best and directly applicable results is crucial for advancing the field.Approach.In this paper, we optimized and validated a decoding approach based on speech reconstruction directly from high-density electrocorticography recordings from sensorimotor cortex during a speech production task.Main results.We show that (1) dedicated machine learning optimization of reconstruction models is key for achieving the best reconstruction performance; (2) individual word decoding in reconstructed speech achieves 92%-100% accuracy (chance level is 8%); (3) direct reconstruction from sensorimotor brain activity produces intelligible speech.Significance.These results underline the need for model optimization in achieving best speech decoding results and highlight the potential that reconstruction-based speech decoding from sensorimotor cortex can offer for development of next-generation BCI technology for communication.

Disentangling rodent behaviors to improve automated behavior recognition.

Automated observation and analysis of behavior is important to facilitate progress in many fields of science. Recent developments in deep learning have enabled progress in object detection and tracking, but rodent behavior recognition struggles to exceed 75-80% accuracy for ethologically relevant behaviors. We investigate the main reasons why and distinguish three aspects of behavior dynamics that are difficult to automate. We isolate these aspects in an artificial dataset and reproduce effects with the state-of-the-art behavior recognition models. Having an endless amount of labeled training data with minimal input noise and representative dynamics will enable research to optimize behavior recognition architectures and get closer to human-like recognition performance for behaviors with challenging dynamics.

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.

An in-silico framework for modeling optimal control of neural systems.

Introduction: Brain-machine interfaces have reached an unprecedented capacity to measure and drive activity in the brain, allowing restoration of impaired sensory, cognitive or motor function. Classical control theory is pushed to its limit when aiming to design control laws that are suitable for large-scale, complex neural systems. This work proposes a scalable, data-driven, unified approach to study brain-machine-environment interaction using established tools from dynamical systems, optimal control theory, and deep learning. Methods: To unify the methodology, we define the environment, neural system, and prosthesis in terms of differential equations with learnable parameters, which effectively reduce to recurrent neural networks in the discrete-time case. Drawing on tools from optimal control, we describe three ways to train the system: Direct optimization of an objective function, oracle-based learning, and reinforcement learning. These approaches are adapted to different assumptions about knowledge of system equations, linearity, differentiability, and observability. Results: We apply the proposed framework to train an in-silico neural system to perform tasks in a linear and a nonlinear environment, namely particle stabilization and pole balancing. After training, this model is perturbed to simulate impairment of sensor and motor function. We show how a prosthetic controller can be trained to restore the behavior of the neural system under increasing levels of perturbation. Discussion: We expect that the proposed framework will enable rapid and flexible synthesis of control algorithms for neural prostheses that reduce the need for in-vivo testing. We further highlight implications for sparse placement of prosthetic sensor and actuator components.

Brain2Pix: Fully convolutional naturalistic video frame reconstruction from brain activity.

Reconstructing complex and dynamic visual perception from brain activity remains a major challenge in machine learning applications to neuroscience. Here, we present a new method for reconstructing naturalistic images and videos from very large single-participant functional magnetic resonance imaging data that leverages the recent success of image-to-image transformation networks. This is achieved by exploiting spatial information obtained from retinotopic mappings across the visual system. More specifically, we first determine what position each voxel in a particular region of interest would represent in the visual field based on its corresponding receptive field location. Then, the 2D image representation of the brain activity on the visual field is passed to a fully convolutional image-to-image network trained to recover the original stimuli using VGG feature loss with an adversarial regularizer. In our experiments, we show that our method offers a significant improvement over existing video reconstruction techniques.

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.

Optimization of Neuroprosthetic Vision via End-to-End Deep Reinforcement Learning.

Visual neuroprostheses are a promising approach to restore basic sight in visually impaired people. A major challenge is to condense the sensory information contained in a complex environment into meaningful stimulation patterns at low spatial and temporal resolution. Previous approaches considered task-agnostic feature extractors such as edge detectors or semantic segmentation, which are likely suboptimal for specific tasks in complex dynamic environments. As an alternative approach, we propose to optimize stimulation patterns by end-to-end training of a feature extractor using deep reinforcement learning agents in virtual environments. We present a task-oriented evaluation framework to compare different stimulus generation mechanisms, such as static edge-based and adaptive end-to-end approaches like the one introduced here. Our experiments in Atari games show that stimulation patterns obtained via task-dependent end-to-end optimized reinforcement learning result in equivalent or improved performance compared to fixed feature extractors on high difficulty levels. These findings signify the relevance of adaptive reinforcement learning for neuroprosthetic vision in complex environments.

A partially nested cortical hierarchy of neural states underlies event segmentation in the human brain.

A fundamental aspect of human experience is that it is segmented into discrete events. This may be underpinned by transitions between distinct neural states. Using an innovative data-driven state segmentation method, we investigate how neural states are organized across the cortical hierarchy and where in the cortex neural state boundaries and perceived event boundaries overlap. Our results show that neural state boundaries are organized in a temporal cortical hierarchy, with short states in primary sensory regions, and long states in lateral and medial prefrontal cortex. State boundaries are shared within and between groups of brain regions that resemble well-known functional networks. Perceived event boundaries overlap with neural state boundaries across large parts of the cortical hierarchy, particularly when those state boundaries demarcate a strong transition or are shared between brain regions. Taken together, these findings suggest that a partially nested cortical hierarchy of neural states forms the basis of event segmentation.

Towards Naturalistic Speech Decoding from Intracranial Brain Data.

Speech decoding from brain activity can enable development of brain-computer interfaces (BCIs) to restore naturalistic communication in paralyzed patients. Previous work has focused on development of decoding models from isolated speech data with a clean background and multiple repetitions of the material. In this study, we describe a novel approach to speech decoding that relies on a generative adversarial neural network (GAN) to reconstruct speech from brain data recorded during a naturalistic speech listening task (watching a movie). We compared the GAN-based approach, where reconstruction was done from the compressed latent representation of sound decoded from the brain, with several baseline models that reconstructed sound spectrogram directly. We show that the novel approach provides more accurate reconstructions compared to the baselines. These results underscore the potential of GAN models for speech decoding in naturalistic noisy environments and further advancing of BCIs for naturalistic communication. Clinical Relevance - This study presents a novel speech decoding paradigm that combines advances in deep learning, speech synthesis and neural engineering, and has the potential to advance the field of BCI for severely paralyzed individuals.

End-to-end optimization of prosthetic vision.

Neural prosthetics may provide a promising solution to restore visual perception in some forms of blindness. The restored prosthetic percept is rudimentary compared to normal vision and can be optimized with a variety of image preprocessing techniques to maximize relevant information transfer. Extracting the most useful features from a visual scene is a nontrivial task and optimal preprocessing choices strongly depend on the context. Despite rapid advancements in deep learning, research currently faces a difficult challenge in finding a general and automated preprocessing strategy that can be tailored to specific tasks or user requirements. In this paper, we present a novel deep learning approach that explicitly addresses this issue by optimizing the entire process of phosphene generation in an end-to-end fashion. The proposed model is based on a deep auto-encoder architecture and includes a highly adjustable simulation module of prosthetic vision. In computational validation experiments, we show that such an approach is able to automatically find a task-specific stimulation protocol. The results of these proof-of-principle experiments illustrate the potential of end-to-end optimization for prosthetic vision. The presented approach is highly modular and our approach could be extended to automated dynamic optimization of prosthetic vision for everyday tasks, given any specific constraints, accommodating individual requirements of the end-user.

DeNovoCNN: a deep learning approach to de novo variant calling in next generation sequencing data.

De novo mutations (DNMs) are an important cause of genetic disorders. The accurate identification of DNMs from sequencing data is therefore fundamental to rare disease research and diagnostics. Unfortunately, identifying reliable DNMs remains a major challenge due to sequence errors, uneven coverage, and mapping artifacts. Here, we developed a deep convolutional neural network (CNN) DNM caller (DeNovoCNN), that encodes the alignment of sequence reads for a trio as 160$ \times$164 resolution images. DeNovoCNN was trained on DNMs of 5616 whole exome sequencing (WES) trios achieving total 96.74% recall and 96.55% precision on the test dataset. We find that DeNovoCNN has increased recall/sensitivity and precision compared to existing DNM calling approaches (GATK, DeNovoGear, DeepTrio, Samtools) based on the Genome in a Bottle reference dataset and independent WES and WGS trios. Validations of DNMs based on Sanger and PacBio HiFi sequencing confirm that DeNovoCNN outperforms existing methods. Most importantly, our results suggest that DeNovoCNN is likely robust against different exome sequencing and analyses approaches, thereby allowing the application on other datasets. DeNovoCNN is freely available as a Docker container and can be run on existing alignment (BAM/CRAM) and variant calling (VCF) files from WES and WGS without a need for variant recalling.

Bayesian model averaging for nonparametric discontinuity design.

Quasi-experimental research designs, such as regression discontinuity and interrupted time series, allow for causal inference in the absence of a randomized controlled trial, at the cost of additional assumptions. In this paper, we provide a framework for discontinuity-based designs using Bayesian model averaging and Gaussian process regression, which we refer to as 'Bayesian nonparametric discontinuity design', or BNDD for short. BNDD addresses the two major shortcomings in most implementations of such designs: overconfidence due to implicit conditioning on the alleged effect, and model misspecification due to reliance on overly simplistic regression models. With the appropriate Gaussian process covariance function, our approach can detect discontinuities of any order, and in spectral features. We demonstrate the usage of BNDD in simulations, and apply the framework to determine the effect of running for political positions on longevity, of the effect of an alleged historical phantom border in the Netherlands on Dutch voting behaviour, and of Kundalini Yoga meditation on heart rate.

A 10-hour within-participant magnetoencephalography narrative dataset to test models of language comprehension.

Recently, cognitive neuroscientists have increasingly studied the brain responses to narratives. At the same time, we are witnessing exciting developments in natural language processing where large-scale neural network models can be used to instantiate cognitive hypotheses in narrative processing. Yet, they learn from text alone and we lack ways of incorporating biological constraints during training. To mitigate this gap, we provide a narrative comprehension magnetoencephalography (MEG) data resource that can be used to train neural network models directly on brain data. We recorded from 3 participants, 10 separate recording hour-long sessions each, while they listened to audiobooks in English. After story listening, participants answered short questions about their experience. To minimize head movement, the participants wore MEG-compatible head casts, which immobilized their head position during recording. We report a basic evoked-response analysis showing that the responses accurately localize to primary auditory areas. The responses are robust and conserved across 10 sessions for every participant. We also provide usage notes and briefly outline possible future uses of the resource.

Real-world indoor mobility with simulated prosthetic vision: The benefits and feasibility of contour-based scene simplification at different phosphene resolutions.

Neuroprosthetic implants are a promising technology for restoring some form of vision in people with visual impairments via electrical neurostimulation in the visual pathway. Although an artificially generated prosthetic percept is relatively limited compared with normal vision, it may provide some elementary perception of the surroundings, re-enabling daily living functionality. For mobility in particular, various studies have investigated the benefits of visual neuroprosthetics in a simulated prosthetic vision paradigm with varying outcomes. The previous literature suggests that scene simplification via image processing, and particularly contour extraction, may potentially improve the mobility performance in a virtual environment. In the current simulation study with sighted participants, we explore both the theoretically attainable benefits of strict scene simplification in an indoor environment by controlling the environmental complexity, as well as the practically achieved improvement with a deep learning-based surface boundary detection implementation compared with traditional edge detection. A simulated electrode resolution of 26 × 26 was found to provide sufficient information for mobility in a simple environment. Our results suggest that, for a lower number of implanted electrodes, the removal of background textures and within-surface gradients may be beneficial in theory. However, the deep learning-based implementation for surface boundary detection did not improve mobility performance in the current study. Furthermore, our findings indicate that, for a greater number of electrodes, the removal of within-surface gradients and background textures may deteriorate, rather than improve, mobility. Therefore, finding a balanced amount of scene simplification requires a careful tradeoff between informativity and interpretability that may depend on the number of implanted electrodes.