%0 Generic %1 zhang2016understanding %A Zhang, Chiyuan %A Bengio, Samy %A Hardt, Moritz %A Recht, Benjamin %A Vinyals, Oriol %D 2016 %K 2016 arxiv deep-learning neural-networks paper %T Understanding deep learning requires rethinking generalization %U http://arxiv.org/abs/1611.03530 %X Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family, or to the regularization techniques used during training. Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization, and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. We interpret our experimental findings by comparison with traditional models.

@misc{zhang2016understanding, abstract = {Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance. Conventional wisdom attributes small generalization error either to properties of the model family, or to the regularization techniques used during training. Through extensive systematic experiments, we show how these traditional approaches fail to explain why large neural networks generalize well in practice. Specifically, our experiments establish that state-of-the-art convolutional networks for image classification trained with stochastic gradient methods easily fit a random labeling of the training data. This phenomenon is qualitatively unaffected by explicit regularization, and occurs even if we replace the true images by completely unstructured random noise. We corroborate these experimental findings with a theoretical construction showing that simple depth two neural networks already have perfect finite sample expressivity as soon as the number of parameters exceeds the number of data points as it usually does in practice. We interpret our experimental findings by comparison with traditional models.}, added-at = {2018-04-10T07:39:20.000+0200}, author = {Zhang, Chiyuan and Bengio, Samy and Hardt, Moritz and Recht, Benjamin and Vinyals, Oriol}, biburl = {https://www.bibsonomy.org/bibtex/236aa98f1b8768649920884bc712f821f/achakraborty}, description = {Understanding deep learning requires rethinking generalization}, interhash = {7f151ed8acacf7b68d2e74f2671111f4}, intrahash = {36aa98f1b8768649920884bc712f821f}, keywords = {2016 arxiv deep-learning neural-networks paper}, note = {cite arxiv:1611.03530Comment: Published in ICLR 2017}, timestamp = {2018-04-10T07:39:20.000+0200}, title = {Understanding deep learning requires rethinking generalization}, url = {http://arxiv.org/abs/1611.03530}, year = 2016 }