Search Papers | Poster Sessions | All Posters
Poster C50 in Poster Session C - Friday, August 9, 2024, 11:15 am – 1:15 pm, Johnson Ice Rink
Model Interval Timing with Laplace Neural Manifolds of Past and Future
Rui Cao1 (), Aakash Sarkar1, Marc Howard1; 1Boston University
Recent data suggest that the brain maintains Laplace transformed neural timelines of the past and the planned future. We apply a cognitive model that uses this representation to explain canonical behavior patterns in interval timing tasks. The model comprises three components: 1. a population of exponentially decaying neurons that encode the Laplace transform of past events at various rates across neurons; 2. a weight matrix that stores Hebbian associations of past events with the present; 3. a population of exponentially ramping neurons at various rates that encode the Laplace transform of the expected future given the present. This model allows an agent to continuously update memory and predicted future in relation to the present moment as events unfold. Unlike typical recurrent neural networks (RNNs) for timing tasks, each component in our model maps to concrete cognitive functions, which enables the agent to adjust and manipulate a logarithmically compressed timeline to meet various task demands.
Keywords: interval timing time perception neural manifold cognitive model