Evolutionary Training and Abstraction Yields Algorithmic Generalization of Neural Computers
This addresses the problem of algorithmic generalization in AI, enabling robust and scalable solutions for tasks requiring iterative or sequential reasoning, though it is incremental as it builds on memory-augmented networks.
The paper tackles the challenge of neural networks learning abstract, algorithmic strategies that generalize across problems, and presents the Neural Harvard Computer (NHC) which achieves perfect generalization and scaling to arbitrary task complexities beyond training data.
A key feature of intelligent behaviour is the ability to learn abstract strategies that scale and transfer to unfamiliar problems. An abstract strategy solves every sample from a problem class, no matter its representation or complexity -- like algorithms in computer science. Neural networks are powerful models for processing sensory data, discovering hidden patterns, and learning complex functions, but they struggle to learn such iterative, sequential or hierarchical algorithmic strategies. Extending neural networks with external memories has increased their capacities in learning such strategies, but they are still prone to data variations, struggle to learn scalable and transferable solutions, and require massive training data. We present the Neural Harvard Computer (NHC), a memory-augmented network based architecture, that employs abstraction by decoupling algorithmic operations from data manipulations, realized by splitting the information flow and separated modules. This abstraction mechanism and evolutionary training enable the learning of robust and scalable algorithmic solutions. On a diverse set of 11 algorithms with varying complexities, we show that the NHC reliably learns algorithmic solutions with strong generalization and abstraction: perfect generalization and scaling to arbitrary task configurations and complexities far beyond seen during training, and being independent of the data representation and the task domain.