Proper Laplacian Representation Learning

Authors: Diego Gomez, Michael Bowling, Marlos C. Machado

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate three different aspects of the proposed max-min objective: eigenvector accuracy, eigenvalue accuracy, and the necessity of each of the components of the proposed objective.
Researcher Affiliation Academia Diego Gomez, Michael Bowling , Marlos C. Machado Department of Computing Science, University of Alberta Alberta Machine Intelligence Institute (Amii) Canada CIFAR AI Chair Edmonton, AB T6G 2R3, Canada {gomeznor,mbowling,machado}@ualberta.ca
Pseudocode No The paper does not contain a dedicated pseudocode or algorithm block.
Open Source Code Yes 6Accompanying code is available here: https://github.com/tarod13/laplacian_dual_ dynamics.
Open Datasets No The paper uses custom-generated grid environments ('Grid Room-1', 'Grid Room-16', 'Grid Maze-19', '12 different grid environments') and describes the process of generating transition samples within them. However, it does not provide concrete access information (link, DOI, formal citation for a specific dataset release) for these environments as publicly available datasets.
Dataset Splits No The paper does not explicitly provide details about training, validation, or test dataset splits, percentages, or sample counts.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions general software components like 'neural network' and 'stochastic gradient descent-based optimizer' but does not specify any software libraries or tools with version numbers.
Experiment Setup Yes We use the (x, y) coordinates as inputs to a fully-connected neural network ϕθ : R2 Rd, parameterized by θ, with 3 layers of 256 hidden units to approximate the d dimensional Laplacian representation ϕ, where d = 11. The network is trained using stochastic gradient descent... for the same initial barrier coefficients as in Figure 1. ...the best barrier increasing rate, αbarrier, for our method across the same environments (αbarrier = 0.01).