skip to main content

Title: Dynamics Are the Only Constant in Working Memory
Abstract In this short perspective, we reflect upon our tendency to use oversimplified and idiosyncratic tasks in a quest to discover general mechanisms of working memory. We discuss how the work of Mark Stokes and collaborators has looked beyond localized, temporally persistent neural activity and shifted focus toward the importance of distributed, dynamic neural codes for working memory. A critical lesson from this work is that using simplified tasks does not automatically simplify the neural computations supporting behavior (even if we wish it were so). Moreover, Stokes' insights about multidimensional dynamics highlight the flexibility of the neural codes underlying cognition and have pushed the field to look beyond static measures of working memory.  more » « less
Award ID(s):
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
Journal of Cognitive Neuroscience
Page Range / eLocation ID:
24 to 26
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Many daily activities and psychophysical experiments involve keeping multiple items in working memory. When items take continuous values (e.g., orientation, contrast, length, loudness) they must be stored in a continuous structure of appropriate dimensions. We investigate how this structure is represented in neural circuits by training recurrent networks to report two previously shown stimulus orientations. We find the activity manifold for the two orientations resembles a Clifford torus. Although a Clifford and standard torus (the surface of a donut) are topologically equivalent, they have important functional differences. A Clifford torus treats the two orientations equally and keeps them in orthogonal subspaces, as demanded by the task, whereas a standard torus does not. We find and characterize the connectivity patterns that support the Clifford torus. Moreover, in addition to attractors that store information via persistent activity, our networks also use a dynamic code where units change their tuning to prevent new sensory input from overwriting the previously stored one. We argue that such dynamic codes are generally required whenever multiple inputs enter a memory system via shared connections. Finally, we apply our framework to a human psychophysics experiment in which subjects reported two remembered orientations. By varying the training conditions of the RNNs, we test and support the hypothesis that human behavior is a product of both neural noise and reliance on the more stable and behaviorally relevant memory of the ordinal relationship between the two orientations. This suggests that suitable inductive biases in RNNs are important for uncovering how the human brain implements working memory. Together, these results offer an understanding of the neural computations underlying a class of visual decoding tasks, bridging the scales from human behavior to synaptic connectivity. 
    more » « less
  2. null (Ed.)
    The development of deep convolutional neural networks (CNNs) has recently led to great successes in computer vision and CNNs have become de facto computational models of vision. However, a growing body of work suggests that they exhibit critical limitations beyond image categorization. Here, we study one such fundamental limitation, for judging whether two simultaneously presented items are the same or different (SD) compared to a baseline assessment of their spatial relationship (SR). In both human subjects and artificial neural networks, we test the prediction that SD tasks recruit additional cortical mechanisms which underlie critical aspects of visual cognition that are not explained by current computational models. We thus recorded EEG signals from human participants engaged in the same tasks as the computational models. Importantly, in humans the two tasks were matched in terms of difficulty by an adaptive psychometric procedure: yet, on top of a modulation of evoked potentials, our results revealed higher activity in the low beta (16-24Hz) band in the SD compared to the SR conditions. We surmise that these oscillations reflect the crucial involvement of additional mechanisms, such as working memory and attention, which are missing in current feed-forward CNNs. 
    more » « less
  3. The record-breaking performance of deep neural networks (DNNs) comes with heavy parameter budgets, which leads to external dynamic random access memory (DRAM) for storage. The prohibitive energy of DRAM accesses makes it nontrivial for DNN deployment on resource-constrained devices, calling for minimizing the movements of weights and data in order to improve the energy efficiency. Driven by this critical bottleneck, we present SmartDeal, a hardware-friendly algorithm framework to trade higher-cost memory storage/access for lower-cost computation, in order to aggressively boost the storage and energy efficiency, for both DNN inference and training. The core technique of SmartDeal is a novel DNN weight matrix decomposition framework with respective structural constraints on each matrix factor, carefully crafted to unleash the hardware-aware efficiency potential. Specifically, we decompose each weight tensor as the product of a small basis matrix and a large structurally sparse coefficient matrix whose nonzero elements are readily quantized to the power-of-2. The resulting sparse and readily quantized DNNs enjoy greatly reduced energy consumption in data movement as well as weight storage, while incurring minimal overhead to recover the original weights thanks to the required sparse bit-operations and cost-favorable computations. Beyond inference, we take another leap to embrace energy-efficient training, by introducing several customized techniques to address the unique roadblocks arising in training while preserving the SmartDeal structures. We also design a dedicated hardware accelerator to fully utilize the new weight structure to improve the real energy efficiency and latency performance. We conduct experiments on both vision and language tasks, with nine models, four datasets, and three settings (inference-only, adaptation, and fine-tuning). Our extensive results show that 1) being applied to inference, SmartDeal achieves up to 2.44x improvement in energy efficiency as evaluated using real hardware implementations and 2) being applied to training, SmartDeal can lead to 10.56x and 4.48x reduction in the storage and the training energy cost, respectively, with usually negligible accuracy loss, compared to state-of-the-art training baselines. Our source codes are available at: 
    more » « less
  4. We initiate the study of biologically-inspired spiking neural networks from the perspective of streaming algorithms. Like computers, human brains face memory limitations, which pose a significant obstacle when processing large scale and dynamically changing data. In computer science, these challenges are captured by the well-known streaming model, which can be traced back to Munro and Paterson `78 and has had significant impact in theory and beyond. In the classical streaming setting, one must compute a function f of a stream of updates š’® = {uā‚,ā€¦,u_m}, given restricted single-pass access to the stream. The primary complexity measure is the space used by the algorithm. In contrast to the large body of work on streaming algorithms, relatively little is known about the computational aspects of data processing in spiking neural networks. In this work, we seek to connect these two models, leveraging techniques developed for streaming algorithms to better understand neural computation. Our primary goal is to design networks for various computational tasks using as few auxiliary (non-input or output) neurons as possible. The number of auxiliary neurons can be thought of as the "space" required by the network. Previous algorithmic work in spiking neural networks has many similarities with streaming algorithms. However, the connection between these two space-limited models has not been formally addressed. We take the first steps towards understanding this connection. On the upper bound side, we design neural algorithms based on known streaming algorithms for fundamental tasks, including distinct elements, approximate median, and heavy hitters. The number of neurons in our solutions almost match the space bounds of the corresponding streaming algorithms. As a general algorithmic primitive, we show how to implement the important streaming technique of linear sketching efficiently in spiking neural networks. On the lower bound side, we give a generic reduction, showing that any space-efficient spiking neural network can be simulated by a space-efficient streaming algorithm. This reduction lets us translate streaming-space lower bounds into nearly matching neural-space lower bounds, establishing a close connection between the two models. 
    more » « less
  5. Abstract

    Investigations into how individual neurons encode behavioral variables of interest have revealed specific representations in single neurons, such as place and object cells, as well as a wide range of cells with conjunctive encodings or mixed selectivity. However, as most experiments examine neural activity within individual tasks, it is currently unclear if and how neural representations change across different task contexts. Within this discussion, the medial temporal lobe is particularly salient, as it is known to be important for multiple behaviors including spatial navigation and memory, however the relationship between these functions is currently unclear. Here, to investigate how representations in single neurons vary across different task contexts in the medial temporal lobe, we collected and analyzed singleā€neuron activity from human participants as they completed a pairedā€task session consisting of a passiveā€viewing visual working memory and a spatial navigation and memory task. Five patients contributed 22 pairedā€task sessions, which were spike sorted together to allow for the same putative single neurons to be compared between the different tasks. Within each task, we replicated conceptā€related activations in the working memory task, as well as targetā€location and serialā€position responsive cells in the navigation task. When comparing neuronal activity between tasks, we first established that a significant number of neurons maintained the same kind of representation, responding to stimuli presentations across tasks. Further, we found cells that changed the nature of their representation across tasks, including a significant number of cells that were stimulus responsive in the working memory task that responded to serial position in the spatial task. Overall, our results support a flexible encoding of multiple, distinct aspects of different tasks by single neurons in the human medial temporal lobe, whereby some individual neurons change the nature of their feature coding between task contexts.

    more » « less