Abstract - Human episodic memory has been studied for several years. It provides us with the ability to remember experiences and recognize people, situations and so on. Researchers have come up with multiple explanations of the working of the human memory system but no final conclusion has been reached. Based on these explanations, various theories and models have been proposed. In this paper, we look at a few such computational models of human episodic memory and perform a comparative study on them. Keywords - Episodic memory, Recall, Recognition, Event, Cues. Episodic memory is the type of memory system that handles our ability to recall previously experienced events and to recognize things as having been encountered previously [1]. It also …show more content…
The model uses the neural network for encoding of experience as events and the spatio-temporal relations amongst these events. This model is constructed by joining two fusion ART networks. It consists of three layers of memory fields. The first layer, F1, holds the values of activation of all attributes, generally situational attributes. According to the pattern of activation in the F1 layer, a node in the F2 layer is selected. The activation pattern of an event is learnt by adjusting the weights of the connections between F1 and F2 layers. The F2 layer also behaves as a buffer for event activations. A sequence of events produces a series of activations in F2 layer. This pattern, represents an episode. It is similarly learnt as weight adjustments on the connections between F2 layer and selected category of F3 layer. An episode can be recognized based on the selected node in F3 layer and can be reproduced by a readout process. The corresponding events can be read out from F2 to F1 layer. Thus encoding, storing and retrieval of events is performed based on computational principles and …show more content…
It is shown in Fig. 2. The EC is divided into two layers, one sends inputs and the other receives outputs. This is modeled by having two layers called EC_in and EC_out. While encoding, the hippocampal region links co-active units in the EC_in to a cluster of units in the CA3 region, thus serving as the hippocampal representation of the episode. To facilitate recall, active units in CA3 region are linked to each other and to the CA1 region which contains an invertible representation of the input pattern. Thus when an item’s representation is activated, the activity is spread to the representation in