When, where, and how to add new neurons to ANNs
This work addresses a challenging and understudied issue in structural learning for neural networks, offering a novel approach to dynamic network growth.
The paper tackles the problem of neurogenesis in artificial neural networks by introducing a framework and strategies for dynamically adding neurons during learning, resulting in networks that converge to an efficient size with improved performance across various supervised tasks.
Neurogenesis in ANNs is an understudied and difficult problem, even compared to other forms of structural learning like pruning. By decomposing it into triggers and initializations, we introduce a framework for studying the various facets of neurogenesis: when, where, and how to add neurons during the learning process. We present the Neural Orthogonality (NORTH*) suite of neurogenesis strategies, combining layer-wise triggers and initializations based on the orthogonality of activations or weights to dynamically grow performant networks that converge to an efficient size. We evaluate our contributions against other recent neurogenesis works across a variety of supervised learning tasks.