Published On Aug 2, 2020
Rewriting a Deep Generative Network, presented at ECCV 2020 (oral).
In this paper, we show how a deep generative network can be reprogrammed by directly rewriting model weights. https://rewriting.csail.mit.edu https://github.com/davidbau/rewriting https://arxiv.org/pdf/2007.15646.pdf
Rewriting a model is challenging, because doing it effectively requires the user to develop a causally correct understanding of the structure, behavior, and purpose of the internal parameters of the network. Up to now it has been unknown whether direct model rewriting is a reasonable proposition.
Our paper shows that model rewriting is feasible.
The video demonstrates a user interface for editing model weights to change the rule that you choose. The paper, with more examples, code, demos, and details are linked at our website.
Website: https://rewriting.csail.mit.edu
Code: https://github.com/davidbau/rewriting
Paper: https://arxiv.org/pdf/2007.15646.pdf