Hebbian plasticity

Connectomic traces of Hebbian plasticity in the entorhinalhippocampal system

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

It is beyond doubt that cognitive phenomena such as language and reasoning are the direct product of the activity of neurons and synapses. Making progress on this open question, often called the bridging problem, was identified by Richard Axel as the most important challenge in neuroscience today. The research community must identify, from among the ever growing volume of neuroscientific knowledge a conceptual core of basic elements and principles which may suffice for bringing about the brain’s function. Here we describe such a minimalistic model of the brain, and we demonstrate its power by

…shape behavior through processes like Hebbian plasticity [2]. Understanding how these mechanisms combine to shape learning will require measurements of anatomical and functional c

Synaptic changes driven by Hebbian plasticity are unstable and require compensatory mechanisms to homeostatically control neuronal activity. Synaptic scaling regulates the synaptic efficacies to bring individual neurons to a set firing regime [1]. However, the timescales of synaptic scaling are too slow to counterbalance the rapid growth due to Hebbian mechanisms [2]. Heterosynaptic plasticity that operates on timescales similar to Hebbian processes can prevent runaway potentiation [3] but its ability to homeostatically control activity is uncertain. Here, we introduce aggregate scaling, a si

The brain computes internal representations by applying highly recurrent dynamics to feed-forward input. Such dynamics may be viewed as analogous to the inference step in latent variables models, where one usually follows the gradient of the latent posterior distribution until a stable state is reached. At the stable state, when this gradient has approximately zero norm, a parameter optimization step can be applied to slightly increase the likelihood of the current sample under the model. Here we suggest this procedure, which closely matches the expectation-maximization (EM) algorithm, can be

…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 hippocampus plays a central role in memory processing. In order to avoid deleterious interference between stored and ongoing experience, theoretical considerations require a separation between phases of memory encoding and recall within the same neural circuit, putatively controlled by hippocampal theta oscillations [1]. However, the neural mechanisms subserving this separation remain unknown. Classical models either remain moot about these mechanisms [2], or assume purpose-built neuromodulatory interactions that are in conflict with biologically realistic timescales of synaptic modulation

Timely reaction to salient stimuli is essential for survival. The subcortical thalamic-basolateral amygdala (BLA) pathway ensures rapid processing. Here, we show that BLA neuronal, and thalamic axonal activity mirror the defensive behaviour evoked by an innate-visual-threat as well as an auditory-learned-threat. Perturbing this pathway compromises defensive responses to both threats. Our side-by-side examination highlights innate and learned threat processing in the subcortical amygdala pathway, thus providing new fundamental insights. Next, we examined the role of homo- and hetero-synaptic pl

Feedforward sensory processing in the brain is generally construed as proceeding through a hierarchy of layers, each constructing increasingly abstract and invariant representations of sensory inputs. This interpretation is at odds with the observation that activity in sensory processing layers is heavily modulated by contextual signals, such as cross modal information or internal mental states. While it is tempting to assume that such modulations bias the feedforward processing pathway towards detection of relevant input features given a context, this induces a dependence on the contextual st

The hippocampus is the episodic memory and spatial navigation center of the brain. The cognitive map metaphor identifies episodic memory with ``navigation'' through a space of arbitrary learned relationships. However, what structures are admissible as ``cognitive maps'' remains unclear, particularly whether these maps are constrained to obey the same rules as physical space. We hypothesize that the hippocampus learns metric spaces as its cognitive maps. We test this hypothesis by training mice to learn and solve relational queries in a finite metric space which cannot be embedded in 2D Euclide

Living organisms receive an immense amount of information per unit time, but much of it is of little relevance for behavior. A simple approach to process this high-dimensional input is to focus on a lower-dimensional subspace and ignore the directions that are less informative. Principal component analysis (PCA) achieves this by discarding the directions with low variance. Such an approach is, however, inefficient in cases where the irrelevant directions are very noisy, thus having greater variance than the relevant ones. If we have access to representative samples of variability in irrelevant

Path integration is an internal computation of position by integrating velocities over time. In the absence of external landmarks, path integration accumulates error, causing the internal representation of position to drift relative to the external world [1,2]. In addition to the well-established role of landmarks in correcting this error, recent experiments identified a new role of landmarks: recalibrating the path-integration gain, a factor mapping the animal’s movement through space to an updating of position in its internal representation [3]. Here, we set out to understand the mechanisms

…While Hebbian plasticity is believed to play a pivotal role in biological memory, it has so far been analyzed mostly in the context of pattern completion and unsupervised learning. Here, we propose that Hebbian plasticity is fundamental for computations in biological neural systems. We introduce a novel spiking neural network (SNN) architecture that is enriched by Hebbian synaptic plasticity. We experimentally

…can be stored as memories with Hebbian plasticity rules, these models considered only a particular Hebbian rule that is not observed in experiments. Here we consider a recently introduced model for arbitrary Hebbian plasticity rules and show how the choice of these rules and of the neural activity patterns influences memorized sequence retrieval. In particular, we derive a general theory that predicts the speed of sequence traversal. Th

The free-energy principle and active inference have received a significant attention in the fields of neuroscience and machine learning. However, it remains to be established whether active inference is an apt explanation for any given neural network that actively exchanges with its environment. To address this issue, we show that a class of canonical neural networks of rate coding models implicitly performs variational Bayesian inference under a well-known form of partially observed Markov decision process model (Isomura, Shimazaki, Friston, Commun Biol, 2022). Based on the proposed theory, w

Learning in the brain is thought to be largely unsupervised since the bulk of an animal’s natural experience arrives neither with associated reinforcement signals nor explicit supervisory signals. A compelling account of unsupervised learning in animals is provided by the notion that the brain learns by trying to predict the future, thereby identifying statistical redundancies in sensory inputs over time. Based on this principle, self-supervised learning (SSL) techniques have enabled deep artificial neural networks (ANNs) to achieve competitive results on challenging perceptual tasks. SSL atte

…curves are thought to be shaped by Hebbian plasticity during development through the strengthening of synapses between neurons with similar functional preferences. This raises the hypothesis that the strength of synapses should correlate with the difference in functional preference between pre- and postsynaptic activity. A recent study, however, found that the functional preference of cortical neurons was correlated with the number of spines a

…synapses through activity-dependent plasticity, and synapses influence neurons by shaping network dynamics. While previous studies have analyzed recurrent circuits with static synapses, or synapses displaying short-term facilitation and depression, the impact of ongoing Hebbian plasticity on network behavior is not well elucidated. Such an understanding is required to map the full landscape of neural-circuit dynamics and to probe computational roles of dynamic synapses. To address this knowledge ga

Orientation selectivity is a key feature of primary visual cortex (V1). Orientation specific receptive fields (OSRFs) appear to arise from specific feedforward connectivity from lateral geniculate nucleus (LGN) to V1 cells. OSRF development has been hypothesized to arise from a competition between ON- and OFF-center LGN cells, if the difference between same-type and opposite-type correlation functions is a non-monotonic, "Mexican hat" function of cell-cell separation. However experiments found this difference decays monotonically. Many models have assumed "Mexican hat" recurrent connectivity (

…model of long-term memory, relying on Hebbian plasticity for storage and attractor dynamics for recall. In contrast, memory-augmented neural networks in machine learning commonly use a key-value mechanism to store and read out memories in a single step. Such augmented networks achieve impressive feats of memory compared to traditional variants, yet it remains unclear whether they can be implemented by biological systems. In our work, we bridge this gap by proposing a set of of biologically plausible three-fact

…correction mechanism to grid cells via Hebbian plasticity, and experiments have found th

In networks of neurons, correlations and synaptic strengths are heavy-tailed, with a small number of neurons interacting much more strongly than the vast majority of pairs. Yet, it remains unclear whether, and how, such heavy-tailed connectivity emerges from simple underlying mechanisms. Building upon recent advances in neuroimaging, here we show that the correlations between thousands of neurons in the mouse visual cortex have heavy-tailed distributions spanning three decades in strength. Moreover, we find that these distributions are robust to variations in the visual stimuli, even persistin

…The concept of Hebbian plasticity suggests that these assemblies are generated through synaptic plasticity, strengthening the recurrent connections within select groups of neurons that receive correlated stimulation. To remain stable in absence of such stimulation, these assemblies need to be self-reinforcing under the plasticity rule. Previous models of such assembly generation and maintenance required mechanisms of heterosynaptic competition or homeostatic plasticity often with biologic