Model-Based Control with Sparse Neural Dynamics

Authors: Ziang Liu, Genggeng Zhou, Jeff He, Tobia Marcucci, Fei-Fei Li, Jiajun Wu, Yunzhu Li

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical and hardware experiments show that, despite the aggressive sparsification, our framework can deliver better closed-loop performance than existing state-of-the-art methods.
Researcher Affiliation Academia 1Cornell University 2Stanford University 3Massachusetts Institute of Technology 4University of Illinois Urbana-Champaign
Pseudocode No The paper describes procedures in text and equations, but does not include a dedicated pseudocode or algorithm block.
Open Source Code No The paper mentions 'Please see our website at robopil.github.io/Sparse-Dynamics/ for additional visualizations.' but does not explicitly state that the source code for their method is available at this link or elsewhere.
Open Datasets Yes For closed-loop control evaluation, we additionally present the performance of our framework on two standardized benchmark environments from Open AI Gym [7], Cart Pole-v1 and Reacher-v4.
Dataset Splits No The paper describes data collection for training (e.g., '1,600 transition pairs') and mentions evaluating 'long-horizon predictive capability' but does not explicitly provide details about specific train/validation/test dataset splits (e.g., percentages or sample counts) for their collected data or how they set up validation for the OpenAI Gym environments.
Hardware Specification No The paper mentions 'Numerical and hardware experiments' but does not provide any specific details about the hardware used, such as GPU models, CPU types, or memory specifications.
Software Dependencies Yes We generated 50,000 transition pairs using the Pymunk simulator [6]. The formulation in Equation 11 is the simplest mixed-integer encoding of a Re LU network, and a variety of strategies are available in the literature to accelerate the solution of our MIPs. ... and solve the problem using a commercial optimization solver, Gurobi [18]. Post-pruning, model speedup is performed using Neural Network Intelligence Library [51].
Experiment Setup Yes Instead of limiting the number of regular Re LUs from the very beginning of the training process, we start with a randomly initialized neural network and use gradient descent to optimize ω and π by minimizing the following objective function until convergence: E[L(θ)] + λR(π), where the regularization term R(π) aims to explicitly reduce the use of the regular Re LU function. ...We then take an iterative approach by starting with a relatively large ε1 and gradually decrease its value for K iterations with ε1 > ε2 > > εK = ε. ...This yields a spectrum of models with varying degrees of sparsification.