Recurrent neural network

Forming an episodic memory requires binding together disparate elements that co-occur in a single experience. One model of this process is that neurons representing different components of a memory bind to an "index"--- a subset of neurons unique to that memory. Evidence for this model has recently been found in chickadees, specialist birds that use hippocampal memory to store and recall locations of cached food. Chickadee hippocampus produces sparse, high-dimensional patterns ("barcodes") that uniquely specify each caching event. Unexpectedly, the same neurons that participate in barcodes als

Flexible behaviors are critical for survival in diverse and dynamic environments. However, the synaptic mechanisms that underlie flexible behaviors are largely unknown. Neuronal responses during behavior are diverse, ranging from reliable ‘classical’ responses (e.g. pure tone tuning in auditory cortex) to irregular ‘non-classically responsive’ firing. Previous work has shown that both classically and non-classically responsive cortical neurons contain significant information about sensory stimuli and behavioral decisions and are necessary for task performance suggesting that non-classically re

Local field potentials (LFPs) of different brain regions are contingent on their individual state and functions. While features of LFPs have been used to distinguish brain regions, no study has attempted to determine the precision with which LFPs can infer the recording positions across the whole brain. To tackle this problem, we hypothesize that tiny differences in LFPs from brain regions and their subregions are separable by using state-of-the-art deep learning methods. We implemented four models to best learn the invariant representations and overcome the predictable noise sources from elec

…motor plan, we trained a modular recurrent neural network model to output similar hand trajectories as macaque monkeys in a manual interception task. In the network modules designed to model the dorsal premotor (PMd) and primary motor (M1) cortices, we observed that rotational neural dynamics during motor preparation transitioned into a subspace encoding hand positions during exec

…multi-agent dynamics and accompanying neural activity in natural settings remains challenging. Here we present an in silico study of social interaction during foraging using recurrent neural network (RNN) agents trained using multi-agent reinforcement learning (MARL). Our framework explores how different group sizes, food availability, and sensory capabilities shape emergent foraging strategies in grops of 2 to 10 agents. These group sizes are especially relevant to experimental social ne

…information of multiple agents shapes neural representations, we train

…this issue by proposing a novel Recurrent Neural Network (RNN) model that enables the spontaneous organization of memory into chunks within WM. Our model leverages specialized “chunking neurons” and long-term

…look to population-level features of neural activity and the distributed computations they enable. Here, we focus on two such features: 1) a continuous neural manifold that maps spatial and task structure, 2) spatial representations that “sweep” ahead of the animal as it moves through the en

…While much is known about the brain networks underlying visual search, our understanding of the neural computations driving this behavior is limited. Simulating this behavior in-silico can provide a tool for developing and testing hypotheses about these neural computations. To address this gap, we trained an image-computable artificial neural networ

Dynamical systems theory provides a promising framework for understanding neural activity. LFADS approximates neural recordings using a recurrent neural network (RNN) generator, but analyzing its dynamics has been challenging. We propose a simple modification to LFADS, adding a per-trial inferred bias, that dramatically improves interpretability by replacing complex multi-stable dynamics with simpler contextual dynamics. Both models were applied to neural activity recorded from macaque motor cortex during reaching movements. While the kinematic decoding was similar, they implemented contrasti

Stable neural representations are essential for consistent behavior and learning, but recent findings on representational drift challenge our understanding of how individual neurons contribute to stable brain computations over time. While representational drift has been observed across multiple brain regions, its relationship to structural connectivity and homeostatic statistics remains unclear [1]. To address this, we explored how drift interacts with structural recurrency and noise in a recurrent neural network (RNN) model. By simulating synaptic-level noise through random connectivity manip

The mammalian olfactory system shows an exceptional ability for rapid and accurate decoding of both the identity and concentration of odorants. In a given scene, only a handful out of millions of possible odorants are present. Yet, the mammals can effectively distinguish these odorants using the dimensionally-reduced representation provided by olfactory receptor neurons. To account for this capability, normative compressed sensing models have been proposed by previous studies. However, these models do not map cleanly onto the unique anatomy and physiology of the olfactory bulb, nor have they d

…rest – composed of an input layer and recurrent neural network – has fixed connections. The benefit of restricting training to the output layer is that it does not require the computationally expensive and biologically-implausible backpropagation typically used to train deep-learning networks. However, while reservoirs can do tasks well, their performance is unreliable; depending on the random initialisation of fixed weights within the network, their capabilities vary wildly, si

Noradrenaline (NA) is a ubiquitous neuromodulator largely produced by locus coeruleus (LC) neurons that impacts multiple cognitive processes. Non-invasive methods are used to track LC-NA activity. For instance, pupillometry is shown to be linked to arousal state, LC-NA, and other neuromodulator activity. However, given that pupillometry is a relatively slow process, the extent by which LC-NA can be predicted by pupil is limited. Since movements are linked with increased LC activity, orofacial movements could potentially be used as an additional psychophysiological marker for arousal and LC-N

…We model motor cortex as a recurrent neural network (RNN) and compare two potential output scenario

…propose a generative model that uses a recurrent neural network to predict upper limb muscle activity during various simple and complex everyday movements. By identifying movement primitives within the si

…While much is known about the brain networks underlying visual search, our understanding of the neural computations driving this behavior is limited, resulting in challenges in simulating such behavior in-silico. To address this gap, we developed an image-computable general-architecture visual search model (GVSM). We trained an artificial neural network to perform naturalistic visual search [Fig. A-B] using the same stimulus configuration as human psychophysics ex

…novel stimuli [1, 2], indicating that neural circuits operate in distinct processing modes depending on stimulus familiarity. In layer 2/3 of mouse primary visual cortex (V1) specifically, populations of pyramidal and vasoactive-intestinal-peptide-expressing (VIP) cells show pronounced responses to novel visual stimuli (novelty responses) and weaker responses to familiar ones. Although experimental and the

Efficient, physiologically-detailed musculoskeletal simulators and powerful learning algorithms provide new computational tools to tackle the grand challenge of understanding biological motor control. Our winning solution for the first NeurIPS MyoChallenge leverages an approach mirroring human learning and showcases reinforcement and curriculum learning as mechanisms to find motor control policies in complex object manipulation tasks. The training procedure, which we call static-to-dynamic stability (SDS), defines a curriculum of progressively more complex state-space neighborhoods, which guid

…analysis of synaptic remodeling in a recurrent neural network.In our experiment, mice were trained to discriminate between two stimuli in a go/no-go task, followed by a retention test after intervals of either 2 or 16 days. This phase was succeeded by reversal learning, where the stimulus contingencies were switched. We observed that mic

Stable neural activity plays a pivotal role in various cognitive, behavioral, and physiological functions, including information processing, memory consolidation, behavioral regulation, and attentional mechanisms. The orchestration of such stability is governed by the intricate interplay of synaptic connections within neural populations. Extensive theoretical and empirical investigations have underscored the influential role of synaptic weights in shaping both the collective dynamics and the fluctuations of neural activity. A salient inquiry arising from this understanding pertains to the mech

…OCD symptoms and brain attractor network states, pointing to a connection with glutamatergic activity. In our investigation, conducted over three time points with 7-week intervals, we collected a comprehensive dataset comprising resting-state and task-based fMRI signals, DTI data, and metabolic details, particularly focusing

In the classical views of systems memory consolidation, memories are transferred from the hippocampus to cortex. But with the advance of recording techniques, it appears that many memories are learned and represented in multiple brain regions simultaneously. Moreover, representations change dynamically while the behavior of animals remains stable. Here, we propose a population perspective on memory consolidation that reproduces many features of this “representational drift”. We develop an analytically tractable model for the dynamics of memory engrams that is structurally similar to a recurren

Cognitive flexibility is necessary for efficient disengagement from a previous task or reconfiguring and implementing new responses to a novel demand. The orbitofrontal cortex is associated with value encoding and is crucial for flexibility to switch behaviours in front of unexpected outcomes. However, the neuronal mechanisms behind such flexibility to modify the encoded value are unknown. We aim to determine the dynamical mechanisms of orbitofrontal neuronal circuits involved in value-based decisions across different environmental contingencies. For this purpose, we perform large-scale electr