Computational models can replicate the capacity of human recognition memory

Computational models can distinguish the contribution from different mechanisms to familiarity recognition

Familiarity is the strange feeling of knowing that something has already been seen in our past. Over the past decades, several attempts have been made to model familiarity using artificial neural networks. Recently, two learning algorithms successfully reproduced the functioning of the perirhinal cortex, a key structure involved during familiarity: Hebbian and anti-Hebbian learning. However, performance of these learning rules is very different from one to another thus raising the question of their complementarity. In this work, we designed two distinct computational models that combined Deep Learning and a Hebbian learning rule to reproduce familiarity on natural images, the Hebbian model and the anti-Hebbian model, respectively. We compared the performance of both models during different simulations to highlight the inner functioning of both learning rules. We showed that the anti-Hebbian model fits human behavioral data whereas the Hebbian model fails to fit the data under large training set sizes. Besides, we observed that only our Hebbian model is highly sensitive to homogeneity between images. Taken together, we interpreted these results considering the distinction between absolute and relative familiarity. With our framework, we proposed a novel way to distinguish the contribution of these familiarity mechanisms to the overall feeling of familiarity. By viewing them as complementary, our two models allow us to make new testable predictions that could be of interest to shed light on the familiarity phenomenon.

Face familiarity detection with complex synapses

Synaptic plasticity is a complex phenomenon involving multiple biochemical processes that operate on different timescales. Complexity can greatly increase memory capacity when the variables characterizing the synaptic dynamics have limited precision, as shown in simple memory retrieval problems involving random patterns. Here we turn to a real-world problem, face familiarity detection, and we show that synaptic complexity can be harnessed to store in memory a large number of faces that can be recognized at a later time. The number of recognizable faces grows almost linearly with the number of synapses and quadratically with the number of neurons. Complex synapses outperform simple ones characterized by a single variable, even when the total number of dynamical variables is matched. Complex and simple synapses have distinct signatures that are testable in experiments. Our results indicate that a system with complex synapses can be used in real-world tasks such as face familiarity detection.

A neural theory for counting memories

Keeping track of the number of times different stimuli have been experienced is a critical computation for behavior. Here, we propose a theoretical two-layer neural circuit that stores counts of stimulus occurrence frequencies. This circuit implements a data structure, called a count sketch, that is commonly used in computer science to maintain item frequencies in streaming data. Our first model implements a count sketch using Hebbian synapses and outputs stimulus-specific frequencies. Our second model uses anti-Hebbian plasticity and only tracks frequencies within four count categories ("1-2-3-many"), which trades-off the number of categories that need to be distinguished with the potential ethological value of those categories. We show how both models can robustly track stimulus occurrence frequencies, thus expanding the traditional novelty-familiarity memory axis from binary to discrete with more than two possible values. Finally, we show that an implementation of the "1-2-3-many" count sketch exists in the insect mushroom body.

Meta-learning synaptic plasticity and memory addressing for continual familiarity detection

Over the course of a lifetime, we process a continual stream of information. Extracted from this stream, memories must be efficiently encoded and stored in an addressable manner for retrieval. To explore potential mechanisms, we consider a familiarity detection task in which a subject reports whether an image has been previously encountered. We design a feedforward network endowed with synaptic plasticity and an addressing matrix, meta-learned to optimize familiarity detection over long intervals. We find that anti-Hebbian plasticity leads to better performance than Hebbian plasticity and replicates experimental results such as repetition suppression. A combinatorial addressing function emerges, selecting a unique neuron as an index into the synaptic memory matrix for storage or retrieval. Unlike previous models, this network operates continuously and generalizes to intervals it has not been trained on. Our work suggests a biologically plausible mechanism for continual learning and demonstrates an effective application of machine learning for neuroscience discovery.

A computational model of familiarity detection for natural pictures, abstract images, and random patterns: Combination of deep learning and anti-Hebbian training

We present a neural network model for familiarity recognition of different types of images in the perirhinal cortex (the FaRe model). The model is designed as a two-stage system. At the first stage, the parameters of an image are extracted by a pretrained deep learning convolutional neural network. At the second stage, a two-layer feed forward neural network with anti-Hebbian learning is used to make the decision about the familiarity of the image. FaRe model simulations demonstrate high capacity of familiarity recognition memory for natural pictures and low capacity for both abstract images and random patterns. These findings are in agreement with psychological experiments.

Single-exposure visual memory judgments are reflected in inferotemporal cortex

Our visual memory percepts of whether we have encountered specific objects or scenes before are hypothesized to manifest as decrements in neural responses in inferotemporal cortex (IT) with stimulus repetition. To evaluate this proposal, we recorded IT neural responses as two monkeys performed a single-exposure visual memory task designed to measure the rates of forgetting with time. We found that a weighted linear read-out of IT was a better predictor of the monkeys’ forgetting rates and reaction time patterns than a strict instantiation of the repetition suppression hypothesis, expressed as a total spike count scheme. Behavioral predictions could be attributed to visual memory signals that were reflected as repetition suppression and were intermingled with visual selectivity, but only when combined across the most sensitive neurons.

As we go about our daily lives, we store visual memories of the objects and scenes that we encounter. This type of memory, known as visual recognition memory, can be remarkably powerful. Imagine viewing thousands of images for only a few seconds each, for example. Several days later, you will still be able to distinguish most of those images from previously unseen ones. How does the brain do this?

Visual information travels from the eyes to an area of the brain called visual cortex. Neurons in a region of visual cortex called inferotemporal cortex fire in a particular pattern to reflect what is being seen. These neurons also reflect memories of whether those things have been seen before, by firing more when things are new and less when they are viewed again. This decrease in firing, known as repetition suppression, may be the signal in the brain responsible for the sense of remembering.

Meyer and Rust have now tested this idea by training macaque monkeys to report whether images on a screen were new or familiar. The monkeys were very good at remembering the images they had seen more recently, although they tended to forget some of the images with time. Then, the rate at which the monkeys forgot the images was compared with the rate at which repetition suppression disappeared in inferotemporal cortex. The results showed that the total number of firing events in this region was not a great predictor of how long the monkeys remembered images. However, a decrease in the number of firing events for a particular subset of the neurons did predict the remembering and forgetting. Repetition suppression in certain inferotemporal cortex neurons can thus account for visual recognition memory.

Brain disorders and aging can both give rise to memory deficits. Identifying the mechanisms underlying memory may lead to new treatments for memory-related disorders. Visual recognition memory may be a good place to start because of our existing knowledge of how the brain processes visual information. Understanding visual recognition memory could help us understand the mechanisms of memory more broadly.

In search of a recognition memory engram

•

The role of the perirhinal cortex in familiarity discrimination is reviewed.

•

Behavioural, pharmacological and electrophysiological evidence is considered.

•

The cortex is found to be essential for memory acquisition, retrieval and storage.

•

The evidence indicates that perirhinal synaptic weakening is critically involved.

A large body of data from human and animal studies using psychological, recording, imaging, and lesion techniques indicates that recognition memory involves at least two separable processes: familiarity discrimination and recollection. Familiarity discrimination for individual visual stimuli seems to be effected by a system centred on the perirhinal cortex of the temporal lobe. The fundamental change that encodes prior occurrence within the perirhinal cortex is a reduction in the responses of neurones when a stimulus is repeated. Neuronal network modelling indicates that a system based on such a change in responsiveness is potentially highly efficient in information theoretic terms. A review is given of findings indicating that perirhinal cortex acts as a storage site for recognition memory of objects and that such storage depends upon processes producing synaptic weakening.

Mechanisms of synaptic plasticity and recognition memory in the perirhinal cortex

Learning is widely believed to involve synaptic plasticity, employing mechanisms such as those used in long-term potentiation (LTP) and long-term depression (LTD). In this chapter, we will review work on mechanisms of synaptic plasticity in perirhinal cortex in vitro and relate these findings to studies underlying recognition memory in vivo. We describe how antagonism of different glutamate and acetylcholine receptors, inhibition of nitric oxide synthase, inhibition of CREB phosphorylation, and interfering with glutamate AMPA receptor internalization can produce deficits in synaptic plasticity in vitro. Inhibition of each of these different mechanisms in vivo also results in recognition memory deficits. Therefore, we provide strong evidence that synaptic plastic mechanisms are necessary for the information processing and storage that underlies object recognition memory.

Recognition memory and synaptic plasticity in the perirhinal and prefrontal cortices

Work is reviewed that relates recognition memory to studies of synaptic plasticity mechanisms in perirhinal and prefrontal cortices. The aim is to consider evidence that perirhinal cortex and medial prefrontal cortex store rather than merely transmit information necessary for recognition memory and, if so, to consider what mechanisms are potentially available within these cortices for producing such storage through synaptic change. Interventions with known actions on plasticity mechanisms are reviewed in relation to their effects on recognition memory processes. These interventions importantly include those involving antagonism of glutamatergic and cholinergic receptors but also inhibition of plasticity consolidation and expression mechanisms. It is concluded that there is strong evidence that perirhinal cortex is involved in information storage necessary for object recognition memory and, moreover, that such storage involves synaptic weakening mechanisms including the removal of AMPA glutamate receptors from synapses. There is good evidence that medial prefrontal cortex is necessary for associative and temporal order recognition memory and that this cortex expresses plasticity mechanisms that potentially allow the storage of information. However, the case for medial prefrontal cortex acting as a store requires further support.

What pharmacological interventions indicate concerning the role of the perirhinal cortex in recognition memory

Findings of pharmacological studies that have investigated the involvement of specific regions of the brain in recognition memory are reviewed. The particular emphasis of the review concerns what such studies indicate concerning the role of the perirhinal cortex in recognition memory. Most of the studies involve rats and most have investigated recognition memory for objects. Pharmacological studies provide a large body of evidence supporting the essential role of the perirhinal cortex in the acquisition, consolidation and retrieval of object recognition memory. Such studies provide increasingly detailed evidence concerning both the neurotransmitter systems and the underlying intracellular mechanisms involved in recognition memory processes. They have provided evidence in support of synaptic weakening as a major synaptic plastic process within perirhinal cortex underlying object recognition memory. They have also supplied confirmatory evidence that that there is more than one synaptic plastic process involved. The demonstrated necessity to long-term recognition memory of intracellular signalling mechanisms related to synaptic modification within perirhinal cortex establishes a central role for the region in the information storage underlying such memory. Perirhinal cortex is thereby established as an information storage site rather than solely a processing station. Pharmacological studies have also supplied new evidence concerning the detailed roles of other regions, including the hippocampus and the medial prefrontal cortex in different types of recognition memory tasks that include a spatial or temporal component. In so doing, they have also further defined the contribution of perirhinal cortex to such tasks. To date it appears that the contribution of perirhinal cortex to associative and temporal order memory reflects that in simple object recognition memory, namely that perirhinal cortex provides information concerning objects and their prior occurrence (novelty/familiarity).

An Infomax Algorithm Can Perform Both Familiarity Discrimination and Feature Extraction in a Single Network

Psychological experiments have shown that the capacity of the brain for discriminating visual stimuli as novel or familiar is almost limitless. Neurobiological studies have established that the perirhinal cortex is critically involved in both familiarity discrimination and feature extraction. However, opinion is divided as to whether these two processes are performed by the same neurons. Previously proposed models have been unable to simultaneously extract features and discriminate familiarity for large numbers of stimuli. We show that a well-known model of visual feature extraction, Infomax, can simultaneously perform familiarity discrimination and feature extraction efficiently. This model has a significantly larger capacity than previously proposed models combining these two processes, particularly when correlation exists between inputs, as is the case in the perirhinal cortex. Furthermore, we show that once the model fully extracts features, its ability to perform familiarity discrimination increases markedly.

Recognition memory: material, processes, and substrates

The proposal that a system centering on the perirhinal cortex is responsible for familiarity discrimination, particularly for single items, whereas a system centering on the hippocampus is responsible for recollective and more complex associational aspects of recognition memory is reviewed in the light of recent findings. In particular, the proposal is reviewed in relation to recent animal work with rats and results from human clinical studies. Notably, progress has been made in determining potential neural memory substrate mechanisms within the perirhinal cortex in rats. Recent findings have emphasized the importance of specifying the type of material, the type of test, and the strategy used by subjects to solve recognition memory tests if substrates are to be accurately inferred. It is to be expected that the default condition is that both the hippocampal and perirhinal systems will contribute to recognition memory performance. Indeed, rat lesion experiments provide examples of where cooperation between both systems is essential. Nevertheless, there remain examples of the independent operation of the hippocampal and perirhinal systems. Overall, it is concluded that most, though not all, of the recent findings are in support of the proposal. However, there is also evidence that the systems involved in recognition memory need to include structures outside the medial temporal lobe: there are significant but as yet only partially defined roles for the prefrontal cortex and sensory association cortices in recognition memory processes.