Event

Whole-brain recordings in Drosophila and digital twins test intrinsically motivated agents

1 reports · 1 sources

Declan Kutscher and co-authors published a paper combining whole-brain recordings in Drosophila with a neuromechanical digital twin to test whether agents trained with intrinsic motivation algorithms can reproduce the fly's spontaneous exploratory behavior in a virtual tethered-ball experiments. Whole-brain recordings in Drosophila show that neural representations of its spontaneous exploration decisions can appear more than 10 seconds in advance. The agents select actions from a discrete library of low-level motor primitives and are trained without external reward using 20 intrinsic motivation algorithms, covering three classes of methods: curiosity, entropy, and competence; evaluation is conducted in a virtual tethered-ball experiment modeled on the Drosophila experiments, comparing spontaneous movement statistics and internal representations, and decoding future turning direction from kinematics and neural activity on the same timescale as the fly. The authors report that among these, DIAYN was closest to Drosophila in behavior and neural anticipatory encoding, and reproduced the additive and subtractive rules of interhemispheric kinematic encoding.