Brick-by-Brick: Combinatorial Construction with Deep Reinforcement Learning
Authors: Hyunsoo Chung, Jungtaek Kim, Boris Knyazev, Jinhwi Lee, Graham W. Taylor, Jaesik Park, Minsu Cho
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our image-conditioned 3D object assembly of B3 in three scenarios: (i) MNIST construction, (ii) randomly-assembled object construction, and (iii) Model Net construction. For an evaluation metric, we measure the episode return or Io U between the constructed object and the desired target at the end of each episode: |
| Researcher Affiliation | Collaboration | 1POSTECH 2University of Guelph 3Vector Institute 4POSCO |
| Pseudocode | No | The paper describes the model and algorithms using mathematical formulations and textual descriptions but does not include a distinct pseudocode block or algorithm listing. |
| Open Source Code | No | The paper does not provide an explicit statement about open-sourcing the code or a link to a code repository. |
| Open Datasets | Yes | For an evaluation metric, we measure the episode return or Io U between the constructed object and the desired target at the end of each episode: (...) MNIST dataset (...) Model Net dataset [40] |
| Dataset Splits | Yes | In particular, 500 images from one of available classes are chosen, further divided into 400 samples for a training dataset and 100 samples for a test dataset. (...) we sample images from 800 target objects for a training dataset while 200 target objects are used for a test dataset. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions using “Open AI Gym” and the “PPO algorithm [35]” but does not provide specific version numbers for these or other software libraries/frameworks. |
| Experiment Setup | No | The paper mentions using the PPO algorithm and optimizing a clipped surrogate objective, but it does not provide concrete values for hyperparameters such as learning rate, batch size, or number of epochs in the main text. |