- author = "Esteban Real and Alok Aggarwal and Yanping Huang and Quoc V. Le",
- title = "Regularized Evolution for Image Classifier Architecture Search",
- booktitle = "Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19)",
- year = "2019",
- editor = "Pascal Van Hentenryck and Zhi-Hua Zhou",
- pages = "4780--4789",
- address = "Hilton Hawaiian Village, Honolulu, Hawaii, USA",
- month = jan # " 27-" # feb # " 1",
- publisher = "AAAI",
- note = "AAAI Technical Track: Machine Learning",
- keywords = "genetic algorithms, genetic programming, GPU, ANN, LSTM, aging evolution, regularized evolution, reinforcement learning, NASNet, image classification, CIFAR-10, model augmentation trick",
- isbn13 = "978-1-57735-809-1",
-
URL = "
https://ojs.aaai.org/index.php/AAAI/article/view/4405",
-
URL = "
https://ojs.aaai.org/index.php/AAAI/article/download/4405/4283.pdf",
-
DOI = "
10.1609/aaai.v33i01.33014780",
-
broken_code_url = "https://colab.research.google.com/github/google-research/google-research/blob/master/evolution/regularizedevolutionalgorithm/regularizedevolution.ipynb",
-
broken_code_url = "https://tfhub.dev/google/imagenet/amoebanetan18f448/classification/1",
- size = "10 page",
- abstract = "The effort devoted to hand-crafting neural network image classifiers has motivated the use of architecture search to discover them automatically. Although evolutionary algorithms have been repeatedly applied to neural network topologies, the image classifiers thus discovered have remained inferior to human-crafted ones. Here, we evolve an image classifier, AmoebaNet-A, that surpasses hand-designs for the first time. To do this, we modify the tournament selection evolutionary algorithm by introducing an age property to favor the younger genotypes. Matching size, AmoebaNet-A has comparable accuracy to current state-of-the-art ImageNet models discovered with more complex architecture-search methods. Scaled to larger size, AmoebaNet-A sets a new state-of-the-art 83.9percent top-1 / 96.6percent top-5 ImageNet accuracy. In a controlled comparison against a well known reinforcement learning algorithm, we give evidence that evolution can obtain results faster with the same hardware, especially at the earlier stages of the search. This is relevant when fewer compute resources are available. Evolution is, thus, a simple method to effectively discover high-quality architectures.",
-
notes = "It evolves trees representing algorithms in the style
of GP, but the encoding restricts the search space to
the architecture of neural networks.
p4780 'we evolve image classifiers that surpass (human) hand-designs'
mutation randomly reconnecting the origin of edges to different vertices and randomly relabeling the edges, covering the full NASnet search space.
fig 5. Model accuracy (0.957--0.967) v. Model Cost (GigaFLOPS) (0.75--1.35).
p4786 Model speed part of fitness (asynchronous EA, ie not generational).
supplements can be found online at: https://arxiv.org/abs/1802.01548
Google Brain",
