A spiking-neuron model of memory encoding and replay in hippocampus
© Trujillo and Eliasmith; licensee BioMed Central Ltd. 2014
Published: 21 July 2014
Hippocampal cells replay sequences of neural activity that have been experienced in the past, implicating the hippocampus in episodic memory. We present a spiking-neuron attractor network model that can encode arbitrary experiences in a combination of learned connection weights and neural activity. The model can then recall part or all of the encoded experience based on contextual cues, simulating hippocampal replay.
The simulation is run using the Neural Engineering Framework (NEF)  and the Nengo neural modelling tool . It consists of 77740 simulated spiking LIF neurons, divided into populations representing hippocampal areas CA3 and CA1, and part of EC. The NEF provides a method for dynamically representing vector-values and computing transformations on them with neural populations.
Our model takes as input a sequence of high-dimensional vectors representing sensory information entering the hippocampus from EC. These vectors can represent position information (place cells) or any other form of processed sensory data. Strong recurrently connected populations in area CA3 cycle through a series of temporal indices, and a Hebbian learning rule allows for these indices to be uniquely generated for a given environment. Using the NEF to compute a binding operator , the network associatively binds the sensory input vectors with an index. These bindings are stored in the neural activity of the recurrent CA1-EC loop. The network can switch between encoding and recall modes, and when in recall mode can look up the corresponding index bound to the experience being recalled and, using the same indexing system in CA3, can complete the pattern of neural activity corresponding to its previously remembered experience.
To the best of our knowledge, this is the first spiking neural model to take into account both timing and sensory tuning of hippocampal cells, while exhibiting the ability to recall previously experienced sequences. In addition, it operates on arbitrary sensory vectors as input, thus not constraining the model to spatial or non-spatial information and allowing it to be extended to perform spatial navigation tasks.
Air Force Office of Scientific Research grant FA8655-13-1-3084, Canada Research Chairs, NSERC Discovery grant, Canadian Foundation for Innovation, Ontario Innovation Trust, NSERC Postgraduate Scholarship.
- Eliasmith C, Anderson CH: Neural Engineering. 2004, The MIT PressGoogle Scholar
- Nengo. [http://www.nengo.ca]
- Plate TA: Holographic reduced representations. Neural networks, IEEE transactions. 1995, 6 (3): 623-641. 10.1109/72.377968.View ArticleGoogle Scholar
- Ji D, Wilson MA: Coordinated memory replay in the visual cortex and hippocampus during sleep. Nature neuroscience. 2007, 10 (1): 100-107. 10.1038/nn1825.View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.