Working memory

…of continuous angular variables for working memory, motor control, and navigation. Theories have proposed that such representations are maintained by a class of recurrently connected networks called ring attractor networks. These networks rely on large numbers of neurons to maintain continuous and stable representations and to accurately integrate incoming signals. The head direction system of the fruit fly, however, seems to achieve these properties with a remarkably small network. These findings challenge our understanding of ring attr

Xiao-Jing Wang surveys theoretical neuroscience through models of cognition, working memory and decision-making. The discussion connects his textbook with the field’s history and current research, including inhibitory-cell diversity, large-scale cortical dynamics and an explanatory approach that links mechanisms across levels rather than treating them separately.

Understanding the computational principles of the brain and replicating them on neuromorphic hardware and modern deep learning architectures is crucial for advancing neuro-inspired AI (NeuroAI). Here, we develop an experimentally-constrained biophysical network model of neocortical circuit motifs, focusing on layers 2-3 of the primary visual cortex (V1). We investigate the role of four major cortical interneuron classes in a competitive-cooperative computational primitive and validate these circuit motifs implemented soft winner-take-all (sWTA) computation for gain modulation, signal restorati

Though continuous attractors are believed to underlie working memory in the brain, learning to remember over long timescales is fundamentally challenging for recurrent neural networks (RNNs). While much prior work has explored why RNNs struggle to learn long timescales and how to mitigate this, we still lack a clear understanding of the dynamics involved when RNNs learn long timescales via gradient descent. Here we build a mathematical theory of the learning dynamics of linear RNNs trained to integrate white noise. We show that when the initial recurrent weights are small, the dynamics of lea

…to coordinate when and which pieces of working memory (WM) information are operated on in the specific task context. For that, fundamental WM flexibility is desirable so that a given task learned using a specific set of WM items can be immediately performed on novel items too without any extra training (generalization in the spirit of zero-shot learning). Here we apply the computational paradigm of Spatial computing as a candidate mechanism for such generalization in a Delayed Match-To-Sample Task (DMTS), using an attracto

Hierarchical working memory refers to the human ability to group information on-the-fly into larger chunks, thereby overcoming the limited capacity of Working Memory (WM), as first suggested by George A. Miller in his 1956 paper. Although chunking has been explored experimentally in numerous studies since then, its neuronal underpinnings remain unknown to this day. In our work, we address 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

…of complex cognitive functions like working memory (WM) at the circuit level. While single-neuron analyses, primarily within the medial temporal lobes, have elucidated how individual neurons contribute to memory computation, the role of local field potential (LFP) dynamics in complementing single-unit activity during WM tasks remains poorly understood. To address this, we developed biologically plausible spiking recurrent neural networks (RNNs) trained on a delayed match-to-sample (DMS) task which in

Understanding how humans form knowledge structures from a stream of information is a question with implications for both human cognition and artificial intelligence. Structural learning is affected by the training regime, which can be either blocked into trials of similar forms or interleaved into trials of different forms. Previous studies show mixed results with some showing that blocked regimes improve different types of learning [1, 2, 3], while others show advantages of blocked regimes [4, 5, 6]. Additionally, few studies investigated the effect of regime on structural learning specifical

…of other relevant information held in working memory. Despite this interdependence, the neural basis of working memory and perceptual decision-making have largely been studied separately. Thus, how the brain flexibly combines sensory information with information in memory to form decisions remains unknown. To study this interaction, we developed a novel behavioral task in which monkeys hold a set of color images in working memory as context,

How does the brain gain insight from limited novel experience? Humans and animals have a striking ability to learn relationships between items of experience, enabling structured generalization and rapid assimilation of new information. A fundamental type of such relational learning is order learning, which enables transitive inference: if A>B and B>C, then A>C. Remarkably, both humans and animals can quickly and globally reorganize separate learned orderings upon learning a single new item ("list-linking"; learn A>B>C and D>E>F, then C>D, and infer B>E), an instance of fast "reassembly" of exi

A central goal of modern neuroscience is to relate structure, dynamics, and computation in recurrent neural networks (RNNs), in particular those that can be created via biological learning rules. To date, most work on RNN training has focused on either efficient but non-biological rules (e.g., backprop, FORCE) or biologically plausible rules with poor learning efficiency (e.g., Hebbian learning). However, some of the fastest, strongest learning rules thought to act in vivo are not Hebbian, for instance in the fly mushroom body, where plasticity depends on high-dimensional presynaptic activity

…while maintaining information in working memory. Recent studies show how top-down inputs representing locomotive state modulates these timescales differently across visual areas in mice. In particular, the sign of correlations between speed and neural activity varies by region, highlighting the interplay between bottom-up sensory inputs, top-down pathways and connectivity motifs in shaping cortical dynamics. In addition, cortical circuits comprise a large diversity

Working memory, the process through which information is transiently maintained and manipulated over a brief period, is essential for most cognitive functions. However, the mechanisms underlying the generation and evolution of working- memory neuronal representations at the population level over long timescales remain unclear. Here, to identify these mechanisms, we trained head-fixed mice to perform an olfactory delayed-association task in which the mice made decisions depending on the sequential identity of two odours separated by a 5 s delay. Optogenetic inhibition of secondary motor neurons

…stabilize activity patterns, enable working memory, and incorporate chaot

Pawel Herman discusses models of working memory and the mechanisms that can maintain information over short timescales.

The prefrontal cortex (PFC) plays an important role in working memory [1], however, the underlying neural processes remains elusive. Studies on the hippocampus revealed that place cell sequences encode goal-directed behavior in rats [2]. In this study, we report about the existence of sequential activation patterns in PFC and their correlation with behavior in a working memory task. Experiments were conducted with one-photon calcium imaging in Thy1-GCaMP6f mice during an olfaction-guided spatial memory task in a three-arm arena consisting of sampling, choice, and reward conditions. In order to

Biophysical neuron models provide mechanistic insight about empirically observed phenomena. However, biophysical neuron models are expensive to simulate, thereby limiting the scale of biophysical models, and optimizing their parameters is notoriously difficult, thereby preventing the fitting of biophysical models to physiologically meaningful tasks or datasets (panel a). We built a new toolbox for simulation and inference in neuroscience, Jaxley, which allows to parallelize simulations on (multiple) GPUs and to compute the gradient of simulations with respect to parameters (panel b, c). We use

The exclusive attribution of working memory (WM) retention to persistent frontal activity [1,2] has been questioned by implications of spike-silent WM [3] in early visual cortex (EVC) [4,5,6]. Spike-silent WM might partly be explained by short-term plasticity (STP) [7], where perceptual encoding leaves a neural trace that can be retrieved using perturbation methods [8,9]. However, STP traces cannot easily withstand continual visual stimulation. In contrast, $modulatory$ top-down feedback [10] (i.e., altering spiking propensity without inducing spiking directly [11]) from frontal regions to EVC

…information, necessitating the use of memory. A recent study indicates that, in such cases, decisions align better with an urgency mechanism, which combines sensory evidence from memory or perception with a growing urgency signa

…This "perfect" memory of a co

A prominent theory proposes that maintaining sequential information in memory is reflected in ordered firing of neurons at different phases of theta oscillations (1). One study has recently challenged this view using single-unit and LFP recordings from epilepsy patients in the human medial temporal lobe (MTL, 2). In addition to predicting a specific temporal relationship between spiking and theta oscillations related to temporal order, the theory also assumes that 1. gamma-frequency activity encodes stimulus-relevant information that is maintained within memory and temporally aligned to spikin

Various working memory (WM) models based on a diverse set of biologically plausible synaptic and neural plasticity mechanisms have been proposed. We show that these proposed short-term plasticity mechanisms may not necessarily be competing explanations, but instead yield interesting functional interactions on a wide set of WM tasks and enhance the biological plausibility of spiking neural network models of WM. While reductionist models (WM explained by one particular mechanism) are theoretically appealing and have increased our understanding of specific mechanisms, they are insufficient. WM mo

The prefrontal cortex (PFC) is known to be involved in working memory, executive control, attention, and decision making. It integrates internal information and states with external information from the environment which results in timely and appropriate behavioral responses. It remains unknown how distinct excitatory cell types within PFC contribute to population activity and behavior. We investigate two major excitatory cell types within the PFC - extratelencephalic (ET) and intratelencephalic (IT) neurons. Anatomical analysis in transgenic mouse lines shows that ET neurons project to subcor

…requires allocation of attention and memory, response selection and inhibition, goal setting, self-control, self-monitoring, and skillful and flexible use of strategies (Eslinger, 1996).The current study atte