Princeton Neuroscience Institute
"Predictive Representations in Memory and Planning”
As we navigate the world, what kinds of representations does the brain store in memory? Which representations does the brain retrieve to make decisions while planning? And how does the brain update these representations? I will present computational, behavioral, and fMRI evidence that memory and planning interact: (1) the brain builds and stores predictive representations in memory that cache multi-step dependencies at multiple scales, i.e., the successor representation (SR), and uses them for fast and flexible planning; (2) predictive representations and planning policies are updated via offline memory processes, such as replay. This SR-replay combination may be a mechanism for segmenting and integrating the flow of experience into memory at different levels of abstraction, which could serve as predictive representations for perception and flexible planning.