Daniel Lee, Travers Rhodes
We lifted 2 functions out of this paper's own repositories and ran 2 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.
| Repository | Role | Ran |
|---|---|---|
| travers-rhodes/jlonevae | canonical | 2 of 2 |
| Function | Status | Where it lives |
|---|---|---|
| compute_generator_jacobian_optimized | Ran | travers-rhodes/jlonevae/jlonevae_lib/architecture/vae_jacobian.py code served (permissive licence) · get_code("36bd19e41d6d0203") |
| jacobian_loss_function | Ran | travers-rhodes/jlonevae/jlonevae_lib/architecture/vae_jacobian.py code served (permissive licence) · get_code("a396932d7dc104c9") |
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
There have been many recent advances in representation learning; however, unsupervised representation learning can still struggle with model identification issues related to rotations of the latent space. Variational Auto-Encoders (VAEs) and their extensions such as β-VAEs have been shown to improve local alignment of latent variables with PCA directions, which can help to improve model disentanglement under some conditions. Borrowing inspiration from Independent Component Analysis (ICA) and sparse coding, we propose applying an L 1 loss to the VAE's generative Jacobian during training to encourage local latent variable alignment with independent factors of variation in images of multiple objects or images with multiple parts. We demonstrate our results on a variety of datasets, giving qualitative and quantitative results using information theoretic and modularity measures that show our added L 1 cost encourages local axis alignment of the latent representation with individual factors of variation.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2106.02923")
get_code_for_paper("2106.02923")
have("2106.02923")
Connect an agent — have() is free.