Geometric Deep Learning: a Temperature Based Analysis of Graph Neural Networks
This work provides a theoretical analysis for researchers in geometric deep learning, but it is incremental as it builds on existing temperature definitions without demonstrating new practical gains.
The authors analyzed Graph Neural Networks (GCN and GAT models) as thermodynamic systems by treating weights as particles and applying a temperature concept from prior work, examining temperature variations across layers without reporting concrete numerical results.
We examine a Geometric Deep Learning model as a thermodynamic system treating the weights as non-quantum and non-relativistic particles. We employ the notion of temperature previously defined in [7] and study it in the various layers for GCN and GAT models. Potential future applications of our findings are discussed.