InvNet: Encoding Geometric and Statistical Invariances in Deep Generative Models
Authors: Ameya Joshi, Minsu Cho, Viraj Shah, Balaji Pokuri, Soumik Sarkar, Baskar Ganapathysubramanian, Chinmay Hegde4377-4384
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our approach on several synthetic and real world problems: navigating manifolds of geometric shapes with desired sizes; generation of binary two-phase materials; and the (challenging) problem of generating multi-orientation polycrystalline microstructures. |
| Researcher Affiliation | Academia | Ameya Joshi, 1 Minsu Cho, 1 Viraj Shah,1 Balaji Pokuri,2 Soumik Sarkar,2 Baskar Ganapathysubramanian,2 Chinmay Hegde1 1Dept. of Electrical and Computer Engineering, 2Dept. of Mechanical Engineering Iowa State University, Ames {ameya, chomd90, viraj, balajip, soumiks, baskarg, chinmay}@iastate.edu |
| Pseudocode | Yes | Algorithm 1 Training Inv Nets Require: Set learning rates, termination conditions; Training data: x [c]d. 1: while LI large and θ has not converged do 2: for l 1 to NG do 3: θ θ ηG θ L(Gθ(z, r)) Generator update 4: end for 5: for m 1 to ND do 6: ψ ψ + ηD ψ L( x) Discriminator update 7: end for 8: for n 1 to NI do 9: θ θ ηD θLI Projection step 10: end for 11: end while |
| Open Source Code | Yes | All code and models can be found at https://github.com/ chomd90/invnet.git. A webapp showcsing the results of our models can also be found at https://tiny.cc/invnet. |
| Open Datasets | Yes | We use a publicly available dataset of 2D binary microstructures containing 34k images across the wide range of statistical moments (Pokuri et al. 2019) |
| Dataset Splits | No | The paper states 'The training data consists of 30k images of size 128 128.' and mentions evaluating on generated images, but it does not specify explicit train/validation/test dataset splits with percentages or sample counts. |
| Hardware Specification | Yes | Uses Intel 4-core CPU with 32 GB RAM. Uses 1 NVIDIA Tesla V100 GPU, 32 GB GDDR5 on Tensor Flow GPU version 1.4. |
| Software Dependencies | Yes | Tensor Flow GPU version 1.4. |
| Experiment Setup | No | The paper mentions 'Set learning rates, termination conditions' in Algorithm 1, but it does not explicitly provide specific hyperparameter values (e.g., learning rate, batch size, number of epochs) or detailed training configurations in the main text. |