Associative learning

How are associations between stimuli or events formed through synaptic plasticity?

Temporal associations in the hippocampus

Episodic memory requires linking discontiguous events in time and depends on the hippocampus. This temporal association learning is often modeled using trace fear conditioning. We probed the ensemble activity in hippocampal CA1 during trace fear learning to differentiate between candidate activity mechanisms and directly resolve the underlying representation.
  • ​​​M. Ahmed*, J. B. Priestley*, A. Castro, F. Stefanini, E. Balough, E. Lavoie, L. Mazzucato, S. Fusi, A. Losonczy,  "Changes in Effective Hippocampal Network Coupling Mediate Learning and Memory of Associations Between Temporally Discontiguous Stimuli." Biological Psychiatry 83 (9), S115. doi:10.1016/j.biopsych.2018.02.304.​​
  • M. S. Ahmed*, J. B. Priestley*, A. Castro, F. Stefanini, E. M. Balough, E. Lavoie, L. Mazzucato, S. Fusi, A. Losonczy, "Hippocampal network reorganization underlies the formation of a temporal association memory," biorxiv, https://doi.org/10.1101/613638 
    [biorxiv]

Stimulus identification with reinforcement learning​

How does a neural network learn to represent stimuli via experience-dependent synaptic plasticity and discriminate between them? In existing models, identity and onset/offset times of relevant stimuli are known to the learning agent. Departing from this view, we introduced an autonomous learning system that identified action-relevant stimuli in a continuous input stream without prior knowledge of them, while learning to ignore stimuli that are not behaviorally relevant. Based on a feedforward spiking network, this model uses a biologically plausible reinforcement learning rule for maximizing the average reward obtained for correct decisions and learn to recognize spatiotemporal activity patterns recorded from the sensory cortex of alert rats.
  • ​L. Le Donne, L. Mazzucato, R. Urbanczik, W. Senn, G. La Camera, "Spike--based reinforcement learning for temporal stimulus segmentation and decision making", Computing with Spikes Workshop, NIPS 2016.