Transforming task representations to perform novel tasks

Abstract

An important aspect of intelligence is the ability to adapt to a novel task without any direct experience (zero-shot), based on its relationship to previous tasks. Humans can exhibit this cognitive flexibility. By contrast, models that achieve superhuman performance in specific tasks often fail to adapt to even slight task alterations. To address this, we propose a general computational framework for adapting to novel tasks based on their relationship to prior tasks. We begin by learning vector representations of tasks. To adapt to new tasks, we propose meta-mappings, higher-order tasks that transform basic task representations. We demonstrate the effectiveness of this framework across a wide variety of tasks and computational paradigms, ranging from regression to image classification and reinforcement learning. We compare to both human adaptability and language-based approaches to zero-shot learning. Across these domains, meta-mapping is successful, often achieving 80-90% performance, without any data, on a novel task, even when the new task directly contradicts prior experience. We further show that meta-mapping can not only generalize to new tasks via learned relationships, but can also generalize using novel relationships unseen during training. Finally, using meta-mapping as a starting point can dramatically accelerate later learning on a new task, and reduce learning time and cumulative error substantially. Our results provide insight into a possible computational basis of intelligent adaptability and offer a possible framework for modeling cognitive flexibility and building more flexible artificial intelligence systems.

arXiv comment45 pages

References in corpus (19)

• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• Empirical Evaluation of Rectified Activations in Convolutional Network
• Exact solutions to the nonlinear dynamics of learning in deep linear neural networks
• Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks
• Zero-Shot Learning Through Cross-Modal Transfer
• Deep Learning: A Critical Appraisal
• Neural Programmer-Interpreters
• Unsupervised Learning via Meta-Learning
• Grounded Language Learning in a Simulated 3D World
• Towards Explaining the Regularization Effect of Initial Large Learning Rate in Training Neural Networks
• Zero-Shot Task Generalization with Multi-Task Deep Reinforcement Learning
• Safe Exploration in Finite Markov Decision Processes with Gaussian Processes
• Diversity is All You Need: Learning Skills without a Reward Function
• HyperNetworks
• Generalization without systematicity: On the compositional skills of sequence-to-sequence recurrent networks
• LGM-Net: Learning to Generate Matching Networks for Few-Shot Learning
• What shapes feature representations? Exploring datasets, architectures, and training
• Learning to Make Analogies by Contrasting Abstract Relational Structure
• One-shot and few-shot learning of word embeddings

Cited by in corpus (5)

• Abstraction and Analogy-Making in Artificial Intelligence
• Learning the Travelling Salesperson Problem Requires Rethinking Generalization
• Continual Robot Learning using Self-Supervised Task Inference
• Towards mental time travel: a hierarchical memory for reinforcement learning agents
• Task Guided Compositional Representation Learning for ZDA