Domain-Adversarial Training of Neural Networks

Abstract

We introduce a new representation learning approach for domain adaptation, in which data at training and test time come from similar but different distributions. Our approach is directly inspired by the theory on domain adaptation suggesting that, for effective domain transfer to be achieved, predictions must be made based on features that cannot discriminate between the training (source) and test (target) domains.The approach implements this idea in the context of neural network architectures that are trained on labeled data from the source domain and unlabeled data from the target domain (no labeled target-domain data is necessary). As the training progresses, the approach promotes the emergence of features that are (i) discriminative for the main learning task on the source domain and (ii) indiscriminate with respect to the shift between the domains. We show that this adaptation behaviour can be achieved in almost any feed-forward model by augmenting it with few standard layers and a new gradient reversal layer. The resulting augmented architecture can be trained using standard backpropagation and stochastic gradient descent, and can thus be implemented with little effort using any of the deep learning packages.We demonstrate the success of our approach for two distinct classification problems (document sentiment analysis and image classification), where state-of-the-art domain adaptation performance on standard benchmarks is achieved. We also validate the approach for descriptor learning task in the context of person re-identification application.

A Convolution-LSTM-Based Deep Neural Network for Cross-Domain MOOC Forum Post Classification

“…Wei et al proposed a two-layer convolutional neural network (LM-CNN-LB) for cross-domain product review sentiment classification [40]. Various studies have also addressed learning representation via a deep architecture to reduce transfer loss [41][42][43][44][45]. To measure the transferability of the deep neural network in NLP, Mou et al studied the transferability of semantic-relative and semantic-irrelative tasks, layers and parameter initialization in multi-task learning and their combination in three datasets [46].…”

Inducing Semantic Micro-Clusters from Deep Multi-View Representations of Novels

“…Our objective function is inspired by prior work in multi-task learning and deep domain adaptation for classification (Ganin and Lempitsky, 2015;Ganin et al, 2016). They train neural networks to simultaneously learn classifiers which are accurate on their target task and are agnostic about feature fluctuation pertaining to domain shift.…”

Improving Chairlift Security with Deep Learning

“…This well-known machine learning problem is called Domain Adaptation. Many interesting approaches have been proposed in deep learning to tackle this problem such as the one from Ganin et al [3]. Their idea (as shown in Figure 1) is to train two networks that share the same first (convolutional) layers called the "feature extractor".…”

“…The second network aims at predicting the domain of an input example. Note that to train this second network, the examples of the target domain do not need to be labeled (and are not in [3]). The only information needed to train the second network is whether an example belongs to the source or to the target domain.…”

“…In all the reported experiments, the learning rate is set to 10 −5 (with an ADAM optimizer), the hinge loss margin to 0.33 for the domain adaptation component (with the reversal layer) and to 0.01 for the domain discriminator. The gradient reversal layer used for the domain adaptation induces a hyper-parameter γ set to 10 as defined by Ganin et al [3].…”