Incremental Learning Through Deep Adaptation

Given an existing trained neural network, it is often desirable to be able to add new capabilities without hindering performance of already learned tasks. Existing approaches either learn sub-optimal solutions, require joint training, or incur a substantial increment in the number of parameters for each added task, typically as many as the original network. We propose a method which fully preserves performance on the original task, with only a small increase (around 20%) in the number of required parameters while performing on par with more costly finetuning procedures, which typically double the number of parameters. The learned architecture can be controlled to switch between various learned representations, enabling a single network to solve a task from multiple different domains. We conduct extensive experiments showing the effectiveness of our method and explore different aspects of its behavior.