Dimitrije Marković
Bernstein Conference 2020
Satellite Workshop: "Dynamic probabilistic inference in the brain"
Introduce a computational model that represents temporal structure of a dynamic environment.
Infer learned temporal structure from human behaviour.
Recent empirical evidence of neuronal circuitry supporting anticipatory behaviour:
Accurate temporal representation → anticipating events.
Marković, et al. PLoS computational biology (2019).
two hidden states
st∈{A,B}
Hidden semi-Markov model
Transition probability
p(st+1∣st,ft)={I2,J2−I2, for ft<n+1 for ft=n+1
Duration probability
p(ft+1∣ft)→p(d)
Phase transitions
p(ft∣ft−1)
M Varmazyar, et al., Journal of Industrial Engineering International (2019).
Phase transitions
p(ft∣ft−1)
Duration distribution
p(d)=(d−1d+n−2)(1−δ)d−1δn
M Varmazyar, et al., Journal of Industrial Engineering International (2019).
K Friston, et al., Neural computation (2017).
history of past outcomes and choices Ht−1=(ot−1:1,at−1:1)
belief updating (Bayes rule)
p(st,ft∣Ht) = p(ot∣at,Ht−1)p(ot∣st,at)p(st,ft∣Ht−1)
Generative process
Action selection
learning phase
model fitting
model testing
Condition with regular reversals
Condition with irregular reversals
duration [d]
duration [d]
Thanks to: