Understanding Visual Concepts with Continuation Learning
This work addresses the challenge of symbolic representation learning in AI, which could benefit fields like computer vision and reinforcement learning, though it appears incremental as it builds on existing concepts like gating and frame prediction.
The authors tackled the problem of learning factorized symbolic representations from visual data by introducing a neural network architecture and learning algorithm that uses consecutive frames to isolate discrete gated units representing factors of variation. They demonstrated the approach on datasets of faces undergoing 3D transformations and Atari 2600 games, but did not provide concrete numerical results in the abstract.
We introduce a neural network architecture and a learning algorithm to produce factorized symbolic representations. We propose to learn these concepts by observing consecutive frames, letting all the components of the hidden representation except a small discrete set (gating units) be predicted from the previous frame, and let the factors of variation in the next frame be represented entirely by these discrete gated units (corresponding to symbolic representations). We demonstrate the efficacy of our approach on datasets of faces undergoing 3D transformations and Atari 2600 games.