Tuning semantic representations using episodic memory Computational Neuroscience

Description

Semantic memories, which are general facts about the world, and episodic memories, which are about personally experienced events, interact with each other. In the  SPI (Serial-Parallel-Independent) model, Tulving (1995) hypothesized that encoding is serial, i.e., sensory information is passed through the semantic system before being encoded into episodic memory. The semantic representation is therefore crucial for episodic memory, but the SPI model does not allow for episodic memory to influence the learning of semantic representations. Based on experimental findings that show that semantic learning can occur without episodic memory, but that such learning is slow, we hypothesize that episodic memory may also facilitate the learning of semantic representations.

We have recently developed a computational model of the interaction between episodic memory and the semantic representation (Fang 2018). The semantic representation is modeled using slow feature analysis (Wiskott 2002), which reduces the spatial and temporal dimensionality of sensory input. Episodic memory is modeled by a simple sequence storage algorithm. Using this model we have been able to show that having episodic memory leads to a more optimized semantic representation.

The hippocampus, a prominent brain structure in mammals, has long been thought to be responsible for memory only, but studies with hippocampally lesioned patients suggest that it may have a function in perception as well (Lee 2012). Using simplified inputs, we have been able to show that perceptual deficits in hippocampal patients may occur because of the less optimized semantic representation, not because of a perceptual function of the hippocampus. The goal of this project is to extend the model to more complex types of inputs and to model human experiments more closely. This project requires knowledge of python (with numpy).

Literature:

Fang et al. (2018). The Interaction between Semantic Representation and Episodic Memory. Neural Computation, 30(2), 293–332

Lee et al. (2012). The hippocampus and visual perception. Front. Hum. Neurosci. 6:91

Tulving (1995). Organization of memory: Quo vadis? In Gazzaniga (Ed.), The cognitive neurosciences (pp. 839-847). Cambridge, MA: MIT Press.

Wiskott & Sejnowski (2002). Slow feature analysis: Unsupervised learning of invariances. Neural Computation 14(4): 715-770.

Supervisors: Richard Görler and Prof. Dr. Sen Cheng

The Institut für Neuroinformatik (INI) is a central research unit of the Ruhr-Universität Bochum. We aim to understand the fundamental principles through which organisms generate behavior and cognition while linked to their environments through sensory systems and while acting in those environments through effector systems. Inspired by our insights into such natural cognitive systems, we seek new solutions to problems of information processing in artificial cognitive systems. We draw from a variety of disciplines that include experimental approaches from psychology and neurophysiology as well as theoretical approaches from physics, mathematics, electrical engineering and applied computer science.

Universitätsstr. 150, Building NB, Room 3/32
D-44801 Bochum, Germany

Tel: (+49) 234 32-28967
Fax: (+49) 234 32-14210