Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Improving Relational Regularized Autoencoders with Spherical Sliced Fused Gromov Wasserstein
Authors: Khai Nguyen, Son Nguyen, Nhat Ho, Tung Pham, Hung Bui
ICLR 2021 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we conduct extensive experiments to show that the new proposed autoencoders have favorable performance in learning latent manifold structure, image generation, and reconstruction. In this section, we conduct extensive experiments on MNIST (Le Cun et al., 1998) and Celeb A datasets (Liu et al., 2015) to evaluate the performance of s-DRAE, ps-DRAE and m(p)s-DRAE with various autoencoders |
| Researcher Affiliation | Collaboration | Khai Nguyen Vin AI Research, Vietnam EMAIL Son Nguyen Vin AI Research, Vietnam EMAIL Nhat Ho University of Texas, Austin Vin AI Research, Vietnam EMAIL Tung Pham Vin AI Research, Vietnam EMAIL Hung Bui Vin AI Research, Vietnam EMAIL |
| Pseudocode | Yes | To generate samples from v MF, we follow the procedure in (Ulrich, 1984), which is described in Algorithm 1 in Appendix B. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described (e.g., a specific repository link, explicit code release statement, or code in supplementary materials). |
| Open Datasets | Yes | In this section, we conduct extensive experiments on MNIST (Le Cun et al., 1998) and Celeb A datasets (Liu et al., 2015) to evaluate the performance of s-DRAE, ps-DRAE and m(p)s-DRAE with various autoencoders |
| Dataset Splits | Yes | For s-DRAE, ps-DRAE and m(p)s-DRAE (10 v MF components with uniform weights and same concentration parameters), we search for κ {1, 5, 10, 50, 100} which gives the best FID score on the validation set of the corresponding dataset. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper mentions using "Adam optimizer" but does not specify version numbers for any software, libraries, or frameworks required to replicate the experiment. |
| Experiment Setup | Yes | To guarantee the fairness of the comparison, we use the same autoencoder architecture, Adam optimizer with learning rate = 0.001, β1 = 0.5 and β2 = 0.999; batch size = 100; latent size = 8 on MNIST and 64 on Celeb A; coefficient λ=1; fused parameter β = 0.1. We set the number of components K = 10 for autoencoder with a mixture of Gaussian distribution as the prior. More detailed descriptions of these settings are in Appendix F. |