Stateless Neural Meta-Learning using Second-Order Gradients
This work addresses the need for efficient and effective meta-learning algorithms for researchers and practitioners in AI, offering an incremental improvement by combining insights from existing methods.
The paper tackled the problem of meta-learning requiring complex algorithms by showing that the meta-learner LSTM subsumes MAML and introducing TURTLE, a simpler yet more expressive algorithm that outperforms both in few-shot tasks like sine wave regression and image classification on miniImageNet and CUB, with accuracy gains of 1-6% and comparable computational cost to second-order MAML.
Deep learning typically requires large data sets and much compute power for each new problem that is learned. Meta-learning can be used to learn a good prior that facilitates quick learning, thereby relaxing these requirements so that new tasks can be learned quicker; two popular approaches are MAML and the meta-learner LSTM. In this work, we compare the two and formally show that the meta-learner LSTM subsumes MAML. Combining this insight with recent empirical findings, we construct a new algorithm (dubbed TURTLE) which is simpler than the meta-learner LSTM yet more expressive than MAML. TURTLE outperforms both techniques at few-shot sine wave regression and image classification on miniImageNet and CUB without any additional hyperparameter tuning, at a computational cost that is comparable with second-order MAML. The key to TURTLE's success lies in the use of second-order gradients, which also significantly increases the performance of the meta-learner LSTM by 1-6% accuracy.