Rapid Adaptation with Conditionally Shifted Neurons
This addresses the problem of enabling machines to adapt quickly to new tasks with limited data, which is incremental as it builds on existing metalearning frameworks.
The paper tackles rapid adaptation in neural networks by introducing conditionally shifted neurons, which modify activations with task-specific shifts from a memory module, achieving state-of-the-art results on metalearning benchmarks in vision and language domains.
We describe a mechanism by which artificial neural networks can learn rapid adaptation - the ability to adapt on the fly, with little data, to new tasks - that we call conditionally shifted neurons. We apply this mechanism in the framework of metalearning, where the aim is to replicate some of the flexibility of human learning in machines. Conditionally shifted neurons modify their activation values with task-specific shifts retrieved from a memory module, which is populated rapidly based on limited task experience. On metalearning benchmarks from the vision and language domains, models augmented with conditionally shifted neurons achieve state-of-the-art results.