Highway Value Iteration Networks
Authors: Yuhui Wang, Weida Li, Francesco Faccio, Qingyuan Wu, Jürgen Schmidhuber
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct a series of experiments to evaluate how highway VINs can improve the long-term planning capabilities of VINs for complex tasks. |
| Researcher Affiliation | Academia | 1AI Initiative, King Abdullah University of Science and Technology 2National University of Singapore 3The Swiss AI Lab IDSIA/USI/SUPSI 4The University of Liverpool. |
| Pseudocode | No | The paper describes algorithms in text and mathematical equations, and provides architectural diagrams, but it does not contain any explicitly labeled "Pseudocode" or "Algorithm" blocks. |
| Open Source Code | Yes | The source code of highway VIN is available at https://github.com/wangyuhuix/Highway VIN. |
| Open Datasets | Yes | Our experimental settings follow those outlined in the paper on GPPN (Lee et al., 2018). For maze navigation tasks, the training, validation, and test datasets comprise 25K, 5K, and 5K mazes, respectively. |
| Dataset Splits | Yes | For maze navigation tasks, the training, validation, and test datasets comprise 25K, 5K, and 5K mazes, respectively. |
| Hardware Specification | Yes | Additionally, the following table details the GPU memory consumption and training duration for each method when employing 300 layers on NVIDIA A100 GPUs. |
| Software Dependencies | No | The paper mentions optimizers (RMSprop) and training parameters, but does not list specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | All models are trained for 30 epochs using the RMSprop optimizer with a learning rate of 0.001 and a batch size of 32. We also specify a kernel size of 5 for convolutional operations in the planning module, as mentioned in Eq. (3). For the neural network that maps the observation to the latent MDP, we set the hidden dimension to 150. |