Course Correcting Koopman Representations
Authors: Mahan Fathi, Clement Gehring, Jonathan Pilault, David Kanaa, Pierre-Luc Bacon, Ross Goroshin
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We empirically evaluated our proposed approach on a number of highly nonlinear environments with varying dimensionality. We begin by modelling the forward dynamics of well known, nonlinear, low dimensional dynamical systems. We further extend the results to more practical, higher dimensional, robotic environments implemented in Mu Jo Co. We use the D4RL dataset by Fu et al. (2020) to train our Koopman autoencoder. |
| Researcher Affiliation | Collaboration | Mahan Fathi Google Deep Mind, Mila Université de Montréal Clement Gehring Mila Université de Montréal Jonathan Pilault Mila Polytechnique Montréal David Kanaa Mila Pierre-Luc Bacon CIFAR AI Chair, Mila Université de Montréal Ross Goroshin Google Deep Mind |
| Pseudocode | No | The paper does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement about releasing source code or a link to a code repository for the described methodology. |
| Open Datasets | Yes | We use the D4RL dataset by Fu et al. (2020) to train our Koopman autoencoder. |
| Dataset Splits | Yes | We use 80% of the trajectories for training and evaluate on the remainder 20%, consisting of trajectories of length 300. |
| Hardware Specification | No | The paper mentions running experiments but does not provide specific details about the hardware used, such as GPU models, CPU types, or cloud computing instances. |
| Software Dependencies | No | The paper mentions 'jax.experimental.ode.odeint()' and 'JAX' (Bradbury et al., 2018), and 'Adam W optimizer by Loshchilov & Hutter (2017)', but does not provide specific version numbers for these software components. |
| Experiment Setup | Yes | We employ Adam W optimizer by Loshchilov & Hutter (2017) with a slight weight decay of value 1e 4 and learning rate of 1e 4. We utilize a custom learning rate of 1e 5 for the dynamics components, which are either the Koopman matrices or 2-layer MLPs. We use embedding size of 128 for the dynamical systems. We use embeddings sizes of 512 and 256 for the state and action encodings, respectively. We utilize training sequences with a length of 10 and 100, for the dynamical systems and D4RL state prediction tasks, respectively. Moreover, we observe that using reencoding training schemes of 20 and 50 steps are beneficial for most of D4RL state prediction tasks, compared to training without reencoding. We use Re LU activations for all experiments to ensure sparse activations. The encoders used for dynamical systems and D4RL are 4-layer and 6-layer standard MLPs, respectively. We use a small sparsity inducing L1 loss, 1e 3, applied to the Koopman embeddings to encourage region-dedicated dynamics. |