Continuous Neural Algorithmic Planners
This work addresses a limitation in neural algorithmic planning for real-world tasks, but it is incremental as it builds on existing methods with discretization and selective expansion.
The authors tackled the problem of extending neural algorithmic reasoning to continuous action spaces by expanding the XLVIN method, which previously only supported discrete actions, and demonstrated that their CNAP approach outperforms model-free baselines in continuous control tasks like MuJoCo, particularly in low-data settings.
Neural algorithmic reasoning studies the problem of learning algorithms with neural networks, especially with graph architectures. A recent proposal, XLVIN, reaps the benefits of using a graph neural network that simulates the value iteration algorithm in deep reinforcement learning agents. It allows model-free planning without access to privileged information about the environment, which is usually unavailable. However, XLVIN only supports discrete action spaces, and is hence nontrivially applicable to most tasks of real-world interest. We expand XLVIN to continuous action spaces by discretization, and evaluate several selective expansion policies to deal with the large planning graphs. Our proposal, CNAP, demonstrates how neural algorithmic reasoning can make a measurable impact in higher-dimensional continuous control settings, such as MuJoCo, bringing gains in low-data settings and outperforming model-free baselines.