Chelsea Finn, Pieter Abbeel, Sergey Levine
2017 · arXiv
We propose an algorithm for meta-learning that is model-agnostic, in the sense that it is compatible with any model trained with gradient descent and applicable to a variety of different learning problems, including classification, regression, and reinforcement learning. The goal of meta-learning is to train a model on a variety of learning tasks, such that it can solve new learning tasks using only a small number of training samples. In our approach, the parameters of the model are explicitly trained such that a small number of gradient steps with a small amount of training data from a new task will produce good generalization performance on that task. In effect, our method trains the model to be easy to fine-tune. We demonstrate that this approach leads to state-of-the-art performance on two few-shot image classification benchmarks, produces good results on few-shot regression, and accelerates fine-tuning for policy gradient reinforcement learning with neural network policies.
Metadata: arXiv. Abstracts retain their original authors’ rights.
Original HN comments, with labels from simple text rules. A label describes the whole comment and may miss context or irony. Read the discussion to make up your own mind. About the labels
3 selected excerpts · 3 total mentions
Hyperparameter optimization (including architectures) is not really meta-learning. Meta-learning, also known as "learning to learn", is more like MAML[0], RL2[1], L2RL[2], etc. 0. https://arxiv.org/abs/1703.03400 1. https://arxiv.org/abs/1611.02779 2. https://arxiv.org/abs/1611.05763
Many concerns in this paper, especially about deep learning being data hungry, limited capacity for transfer, and integrating prior knowledge are addressed by recent papers in meta-learning and learning-to-learn [1,2,3 and many many others] both in the supervised and reinforcement learning contexts. In the case of meta-reinforcement learning there has been recent work [5] which seems to indicate that this mechanism is very similar to how learning works in the brain. In fact, my group recently published a paper [4] about learning-to-learn in spiking networks (making it biologically more realistic) and showing that the network learns priors on families of tasks for…
supervised learning, and learns useful exploration strategies automatically when doing meta-reinforcement learning. While I don't claim this is the right path to AGI, it's a very promising and new direction in deep learning research which this paper seems to ignore. [1] http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.5.32... [2] https://arxiv.org/abs/1611.05763 [3] https://arxiv.org/abs/1703.03400 [4] https://arxiv.org/abs/1803.09574 [5] https://www.nature.com/articles/s41593-018-0147-8 (Preprint: https://www.biorxiv.org/content/early/2018/04/06/295964)
I think we may be talking to different customers. I talked with ~200 customers last year, and the most common question was "What do I use ML for?" and the second most common was "How do I get started?" Put another way, the average customer has ~zero ML usage today. I'd guess that 95%+ of all businesses have ML usage today. Further guessing would say that <1% of ML usage actually care about levels of accuracy beyond "it's better than the hacked together set of rules/filters I use today." These are very large businesses with lots of money to spend on a solution. There are many ways to measure "better", and AutoML does apply here. This includes "better == faster to train or…
develop" [1], "better == you need less data" [2], "better == lower error rates"[3]. While I agree that many of measures do not apply across datasets, most customers only have one dataset per problem. [1] Predictive accuracy and run-time - https://repositorio-aberto.up.pt/bitstream/10216/104210/2/19... [2] Less data - https://arxiv.org/abs/1703.03400 [3] https://link.springer.com/article/10.1007/s10994-017-5687-8 Disclosure: I work at Google on Kubeflow
More from the same topics.
Tom B. Brown, Benjamin Mann, Nick Ryder +28
Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task.
Anders Sandberg, Eric Drexler, Toby Ord
The Fermi paradox is the conflict between an expectation of a high {\em ex ante} probability of intelligent life elsewhere in the universe and the apparently lifeless universe we in fact observe.
Chiyuan Zhang, Samy Bengio, Moritz Hardt +2
Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance.
Scott Aaronson
In honor of Alan Turing's hundredth birthday, I unwisely set out some thoughts about one of Turing's obsessions throughout his life, the question of physics and free will.