Quantifying and Learning Linear Symmetry-Based Disentanglement
Authors: Loek Tonnaer, Luis Armando Perez Rey, Vlado Menkovski, Mike Holenderski, Jim Portegies
ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the utility of our metric by showing that (1) common VAE-based disentanglement methods don t learn LSBD representations, (2) LSBD-VAE, as well as other recent methods, can learn LSBD representations needing only limited supervision on transformations, and (3) various desirable properties expressed by existing disentanglement metrics are also achieved by LSBD representations. |
| Researcher Affiliation | Collaboration | 1Eindhoven University of Technology (TU/e), Eindhoven, The Netherlands 2Eindhoven Artificial Intelligence Systems Institute (EAISI), Eindhoven, the Netherlands 3Prosus, Amsterdam, The Netherlands. |
| Pseudocode | No | The paper does not contain explicit pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code available at https://github.com/ luis-armando-perez-rey/lsbd-vae |
| Open Datasets | Yes | Data We evaluate the disentanglement of several models on three different image datasets (Square, Arrow, and Airplane) with a known group decomposition G = SO(2) SO(2) describing the underlying transformations... we also evaluate the same models on two datasets Model Net40 (Wu et al., 2014) and COIL-100 (Nene et al., 1996)... This dataset was obtained using Tensor Flow Datasets (2021). |
| Dataset Splits | Yes | For the Square, Arrow, and Airplane datasets we test LSBDVAE with transformation-labeled batches of size M = 2. More specifically, for each experiment we randomly select L disjoint pairs of data points, and label the transformation between the data points in each pair. We vary the number of labeled pairs L from 0 (corresponding to a VAE) to N/2 (in which case each data point is involved in exactly one labeled pair). |
| Hardware Specification | Yes | The hardware used across all experiments was a DGX station with 4 NVIDIA GPUs V100 and 32GB . Only one GPU was used per experiment. |
| Software Dependencies | No | The paper mentions software like "Blender v2.7" for data generation and "disentanglement lib" for other methods, but does not provide specific version numbers for core software dependencies used for running their own experiments (e.g., Python, PyTorch/TensorFlow). |
| Experiment Setup | Yes | Table 3 shows the hyperparameters used to train each model for all datasets. Table 4 shows the hyperparameters used to train the LSBD-VAE models for each dataset. In the latter case, the number of epochs for the LSBD-VAE model were increased. |