Snowflake: Scaling GNNs to High-Dimensional Continuous Control via Parameter Freezing
This addresses a scaling bottleneck for GNNs in robotics and control domains, enabling their use in high-dimensional settings where they previously failed, though it is incremental as it builds on existing GNN architectures.
The paper tackled the problem of scaling graph neural networks (GNNs) to high-dimensional continuous control, where performance deteriorates with larger agents, by introducing Snowflake, a training method that freezes parameters to combat overfitting. The result is that Snowflake significantly boosts GNN performance on large agents, matching multi-layer perceptrons (MLPs) and offering superior transfer properties.
Recent research has shown that graph neural networks (GNNs) can learn policies for locomotion control that are as effective as a typical multi-layer perceptron (MLP), with superior transfer and multi-task performance (Wang et al., 2018; Huang et al., 2020). Results have so far been limited to training on small agents, with the performance of GNNs deteriorating rapidly as the number of sensors and actuators grows. A key motivation for the use of GNNs in the supervised learning setting is their applicability to large graphs, but this benefit has not yet been realised for locomotion control. We identify the weakness with a common GNN architecture that causes this poor scaling: overfitting in the MLPs within the network that encode, decode, and propagate messages. To combat this, we introduce Snowflake, a GNN training method for high-dimensional continuous control that freezes parameters in parts of the network that suffer from overfitting. Snowflake significantly boosts the performance of GNNs for locomotion control on large agents, now matching the performance of MLPs, and with superior transfer properties.