Additive Decoders for Latent Variables Identification and Cartesian-Product Extrapolation
Dec 15, 2023
Speakers
Sébastien Lachapelle
Speaker · 0 followers
Divyat Mahajan
Speaker · 1 follower
Ioannis Mitliagkas
Speaker · 0 followers
Simon Lacoste-Julien
Speaker · 1 follower
About
We tackle the problems of latent variables identification and "out-of-support'' image generation in representation learning. We show that both are possible for a class of decoders that we call additive, which are reminiscent of decoders used for object-centric representation learning (OCRL) and well suited for images that can be decomposed as a sum of object-specific images. We provide conditions under which exactly solving the reconstruction problem using an additive decoder is guaranteed to identify the blocks of latent variables up to permutation and block-wise invertible transformations. This guarantee relies only on very weak assumptions about the distribution of the latent factors, which might present statistical dependencies and have an almost arbitrarily shaped support. Our result provides a new setting where nonlinear independent component analysis (ICA) is possible and adds to our theoretical understanding of OCRL methods. We also show theoretically that additive decoders can generate novel images by recombining observed factors of variations in novel ways, an ability we refer to as Cartesian-product extrapolation. We show empirically that additivity is crucial for both identifiability and extrapolation on simulated data.We tackle the problems of latent variables identification and "out-of-support'' image generation in representation learning. We show that both are possible for a class of decoders that we call additive, which are reminiscent of decoders used for object-centric representation learning (OCRL) and well suited for images that can be decomposed as a sum of object-specific images. We provide conditions under which exactly solving the reconstruction problem using an additive decoder is guaranteed to id…
Organizer
NeurIPS 2023
Account · 645 followers
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
Differentially Private Image Classification by Learning Priors from Random Processes
Watch later
Xinyu Tang, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Outlier-Robust Gromov Wasserstein for Graph Data
Watch later
Lemin Kong, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Conformal Prediction for Time Series with Modern Hopfield Networks
Watch later
Andreas Auer, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Spectral Co-Distillation for Personalized Federated Learning
Watch later
Zihan Chen, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
AircraftVerse: A Large-Scale Multimodal Dataset of Aerial Vehicle Designs
Watch later
Adam Cobb, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
PlanBench: An Extensible Benchmark for Evaluating Large Language Models on Planning and Reasoning about Change
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%