ATLAS: Universal Function Approximator for Memory Retention
This addresses the problem of catastrophic forgetting for continual learning in AI, offering a foundational but incremental improvement over existing methods.
The paper tackles catastrophic forgetting in artificial neural networks by introducing ATLAS, a novel architecture based on a new universal approximation theorem using single-variable and exponential functions, which achieves some memory retention and continual learning with predictable off-target effects.
Artificial neural networks (ANNs), despite their universal function approximation capability and practical success, are subject to catastrophic forgetting. Catastrophic forgetting refers to the abrupt unlearning of a previous task when a new task is learned. It is an emergent phenomenon that hinders continual learning. Existing universal function approximation theorems for ANNs guarantee function approximation ability, but do not predict catastrophic forgetting. This paper presents a novel universal approximation theorem for multi-variable functions using only single-variable functions and exponential functions. Furthermore, we present ATLAS: a novel ANN architecture based on the new theorem. It is shown that ATLAS is a universal function approximator capable of some memory retention, and continual learning. The memory of ATLAS is imperfect, with some off-target effects during continual learning, but it is well-behaved and predictable. An efficient implementation of ATLAS is provided. Experiments are conducted to evaluate both the function approximation and memory retention capabilities of ATLAS.