search

ermongroup / mintnet

MintNet: Building Invertible Neural Networks with Masked Convolutions
38 stars 8 forks source link

Same model for classification as well as generating samples #3

Open Ektagavas opened 3 years ago

Ektagavas commented 3 years ago

Hey, thank you for sharing the code. It's great! I had one query. I have trained a Mintnet model for classification on CIFAR10. Can I use this same model for generating new samples i.e as a generative model (with some code modifications, of course)? If yes, can you suggest what part of the code needs to be changed? As far as I understood, currently, density estimation and classification experiments are run independently, but I want to make changes to achieve the above use-case. Thanks in advance!

yang-song commented 3 years ago

Hello, I don't think it's easy to convert a classifier to a generative model. For image generation, you will need to first have the logits layer of a MintNet classifier and then compute the inverse. The first step is non-trivial.