Adaptive Neural Networks Using Residual Fitting
This work addresses the computational inefficiency of methods like neural-architecture search for network sizing, offering a more efficient alternative for practitioners in machine learning.
The paper tackles the problem of determining appropriate neural network sizes by introducing a growth method that adds capacity based on residual error detection, achieving better performance than small static networks and similar results to larger pre-sized networks in tasks like classification and reinforcement learning.
Current methods for estimating the required neural-network size for a given problem class have focused on methods that can be computationally intensive, such as neural-architecture search and pruning. In contrast, methods that add capacity to neural networks as needed may provide similar results to architecture search and pruning, but do not require as much computation to find an appropriate network size. Here, we present a network-growth method that searches for explainable error in the network's residuals and grows the network if sufficient error is detected. We demonstrate this method using examples from classification, imitation learning, and reinforcement learning. Within these tasks, the growing network can often achieve better performance than small networks that do not grow, and similar performance to networks that begin much larger.