Space-Time Graph Neural Networks

Authors: Samar Hadou, Charilaos I Kanatsoulis, Alejandro Ribeiro

ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical experiments with decentralized control systems showcase the effectiveness and stability of the proposed ST-GNNs. Our theoretical findings are also supported by thorough experimental analysis based on decentralized control applications.
Researcher Affiliation Academia Department of Electrical and Systems Engineering University of Pennsylvania {selaraby, kanac, aribeiro}@seas.upenn.edu
Pseudocode No No pseudocode or algorithm blocks were found in the paper.
Open Source Code No We used the GNN library at https://github.com/alelab-upenn/graph-neural-networks
Open Datasets No The dataset is generated according to the mobility model in (95) and (96). The dataset consists of 500 time-varying graph signals {Xm}500 m=1 that are calculated under optimal centralized policies {U m}500 m=1.
Dataset Splits Yes We split the data into 460 examples for training, 20 for validation and 20 for testing.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory amounts) used for running experiments were mentioned in the paper.
Software Dependencies No We used the GNN library at https://github.com/alelab-upenn/graph-neural-networks
Experiment Setup Yes We train a 2-layer ST-GNN on the training data and optimize the mean squared loss using ADAM algorithm with learning rate 0.01 and decaying factors β1 = 0.9 and β2 = 0.999. Table 1: Simulation parameters in Experiments #1 and #2. parameter value... ST-GNN feature/layer, F0:2 4, 16, 2 (#1) and 6, 64, 2 (#2) Filter taps/layer, K1:2 4, 1 Activation function, σ tanh