Optimal Design for Hetero-Associative Memory: Hippocampal CA1 Phase Response Curve and Spike-Timing-Dependent Plasticity

Abstract

Recently reported experimental findings suggest that the hippocampal CA1 network stores spatio-temporal spike patterns and retrieves temporally reversed and spread-out patterns. In this paper, we explore the idea that the properties of the neural interactions and the synaptic plasticity rule in the CA1 network enable it to function as a hetero-associative memory recalling such reversed and spread-out spike patterns. In line with Lengyel’s speculation (Lengyel et al., 2005), we firstly derive optimally designed spike-timing-dependent plasticity (STDP) rules that are matched to neural interactions formalized in terms of phase response curves (PRCs) for performing the hetero-associative memory function. By maximizing object functions formulated in terms of mutual information for evaluating memory retrieval performance, we search for STDP window functions that are optimal for retrieval of normal and doubly spread-out patterns under the constraint that the PRCs are those of CA1 pyramidal neurons. The system, which can retrieve normal and doubly spread-out patterns, can also retrieve reversed patterns with the same quality. Finally, we demonstrate that purposely designed STDP window functions qualitatively conform to typical ones found in CA1 pyramidal neurons.

Heteroassociative storage of hippocampal pattern sequences in the CA3 subregion

“…For instance, Yamaguchi (2003) studied a model that tries to explains theta phase precession using heteroassociative connections from CA3 to CA1 and in the CA3 recurrent connections. Other studies also used simple feedforward network models that associate pairs of patterns ( Lytton, 1998 ; Miyata, Ota & Aonishi, 2013 ). But these studies did not evaluate the problem of sequence completion, noise tolerance and operation with limited connectivity.…”

A Model for the Peak-Interval Task Based on Neural Oscillation-Delimited States

“…The proposed model works with neural networks that generate reproducible sequences of states at steady rates. Heteroassociative networks (Sompolinsky and Kanter, 1986;Camargo et al, 2018) can be used to implement our model as a biological neural network, as well as Synfire chains (Abeles, 1991;Miyata et al, 2013). The later introduced the idea of a rapid sequential activation of groups of neurons, representing the states in our model.…”

“…Accordingly, heteroassociative networks are promising as a biological implementation of the model, since they can produce long sequences of state activation within a single network while performing pattern completion (Camargo et al, 2018). This kind of networks are present in both the CA3 (Lisman et al, 2005) and CA1 (Miyata et al, 2013) subregions of the hippocampus, which is involved in some time-related tasks (Meck et al, 1984). Accumulated evidence suggests that timing mechanisms are decentralized (Paton and Buonomano, 2018) and the hippocampus is one from brain areas where sequences of states could be evoked (Buzsáki and Tingley, 2018).…”

Integration of exteroceptive and interoceptive information within the hippocampus: a computational study

“…By evidence of associative synaptic plasticity at most of the hippocampal synapses (Andersen et al, 2007 ), auto- and heteroassociative networks have already been validated as prototypical models of the mnemonic function of the hippocampus. While autoassociative learning has long been ascribed to CA3 because of its recurrent connections (Marr, 1971 ), heteroassociative learning has been reported to occur at the Schaffer collaterals projecting from CA3 to CA1 (Miyata et al, 2013 ), at CA3 backprojections on hilar mossy cells and DG granule cells (Lisman and Otmakhova, 2001 ). Recently, the functional homogeneity of CA3 has been called into question (de Almeida et al, 2007 ; Witter, 2007 ; Hunsaker et al, 2008 ; Thompson et al, 2008 ; Bush et al, 2010 ).…”