paper discussion: Neural Architecture Search with Reinforcement Learning

[Chad DeChant]

There has been a flood of papers submitted to ICLR and this week we will be reading one of them. 

2pm Wednesday 11/9 in the Computer Science Department conference room in the CS building in Mudd.

http://openreview.net/forum?id=r1Ue8Hcxg

Neural Architecture Search with Reinforcement Learning

Barret Zoph, Quoc Le

Abstract: Neural networks are powerful and flexible models that work well for many diffi- cult learning tasks in image, speech and natural language understanding. Despite their success, neural networks are still hard to design. In this paper, we use a re- current network to generate the model descriptions of neural networks and train this RNN with reinforcement learning to maximize the expected accuracy of the generated architectures on a validation set. On the CIFAR-10 dataset, our method, starting from scratch, can design a novel network architecture that rivals the best human-invented architecture in terms of test set accuracy. Our CIFAR-10 model achieves a test error rate of 3.84, which is only 0.1 percent worse and 1.2x faster than the current state-of-the-art model. On the Penn Treebank dataset, our model can compose a novel recurrent cell that outperforms the widely-used LSTM cell, and other state-of-the-art baselines. Our cell achieves a test set perplexity of 62.4 on the Penn Treebank, which is 3.6 perplexity better than the previous state-of- the-art.