Invited Talk: "The Apperception Engine" - Richard Evans

This talk attempts to answer a central question in unsupervised learning: what does it mean to “make sense” of a sensory sequence? In our formalization, making sense involves constructing a symbolic causal theory that both explains the sensory sequence and also satisfies a set of unity conditions. The unity conditions insist that the constituents of the causal theory – objects, properties, and laws – must be integrated into a coherent whole. On our account, making sense of sensory input is a type of program synthesis, but it is unsupervised program synthesis. I will show how our system makes sense of a variety of sensory sequences, including rhythmic sequences, sequence induction IQ tasks, and occlusion tasks. It is noteworthy that our system is able to achieve human-level performance on these IQ tasks, even though it was not designed to solve those particular tasks.

In the second half I will describe our neuro-symbolic framework for distilling interpretable theories out of streams of raw, unprocessed sensory experience. First, we extend the definition of the apperception task to include ambiguous (but still symbolic) input: sequences of sets of disjunctions. Next, we use a neural network to map raw sensory input to disjunctive input. Our binary neural network is encoded as a logic program, so the weights of the network and the rules of the theory can be solved jointly as a single SAT problem. This way, we are able to jointly learn how to perceive (mapping raw sensory information to concepts) and apperceive (combining concepts into declarative rules).

Program

13:15 - 13:30 Doors open

13.30 - 13.40 Introduction (Luc De Raedt)

13.40 - 14.30 Keynote “The Apperception Engine” (Richard Evans - DeepMind)

14.30 - 15.00 Presentation “RoadMap TAILOR” (Marc Schoenauer, Michela Milano)

15:00 - 15:15 Break

15.15 - 15.45 Breakout room

15.45 - 16.00 Wrapping up

16.00 - 16.30 Poster Spotlights

16.30 - 16.45 Virtual Coffee

16.45 - 18.00 Poster Session

18.00 - 18.15 Next Steps