Domain Adaptation with Asymmetrically-Relaxed Distribution Alignment

Domain adaptation addresses the common problem when the target distribution generating our test data drifts from the source (training) distribution. While absent assumptions, domain adaptation is impossible, strict conditions, e.g. covariate or label shift, enable principled algorithms. Recently-proposed domain-adversarial approaches consist of aligning source and target encodings, often motivating this approach as minimizing two (of three) terms in a theoretical bound on target error. Unfortunately, this minimization can cause arbitrary increases in the third term, e.g. they can break down under shifting label distributions. We propose asymmetrically-relaxed distribution alignment, a new approach that overcomes some limitations of standard domain-adversarial algorithms. Moreover, we characterize precise assumptions under which our algorithm is theoretically principled and demonstrate empirical benefits on both synthetic and real datasets.

[1]  J. Heckman Sample Selection Bias as a Specification Error (with an Application to the Estimation of Labor Supply Functions) , 1977 .

[2]  H. Shimodaira,et al.  Improving predictive inference under covariate shift by weighting the log-likelihood function , 2000 .

[3]  Marco Saerens,et al.  Adjusting the Outputs of a Classifier to New a Priori Probabilities: A Simple Procedure , 2002, Neural Computation.

[4]  Koby Crammer,et al.  Analysis of Representations for Domain Adaptation , 2006, NIPS.

[5]  Bernhard Schölkopf,et al.  Correcting Sample Selection Bias by Unlabeled Data , 2006, NIPS.

[6]  Koby Crammer,et al.  A theory of learning from different domains , 2010, Machine Learning.

[7]  Karsten M. Borgwardt,et al.  Covariate Shift by Kernel Mean Matching , 2009, NIPS 2009.

[8]  Yishay Mansour,et al.  Domain Adaptation: Learning Bounds and Algorithms , 2009, COLT.

[9]  Tyler Lu,et al.  Impossibility Theorems for Domain Adaptation , 2010, AISTATS.

[10]  Mehryar Mohri,et al.  Domain Adaptation in Regression , 2011, ALT.

[11]  Yaoliang Yu,et al.  Analysis of Kernel Mean Matching under Covariate Shift , 2012, ICML.

[12]  Bernhard Schölkopf,et al.  Domain Adaptation under Target and Conditional Shift , 2013, ICML.

[13]  Ruth Urner,et al.  Domain adaptation–can quantity compensate for quality? , 2013, Annals of Mathematics and Artificial Intelligence.

[14]  Trevor Darrell,et al.  Deep Domain Confusion: Maximizing for Domain Invariance , 2014, CVPR 2014.

[15]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[16]  George Trigeorgis,et al.  Domain Separation Networks , 2016, NIPS.

[17]  François Laviolette,et al.  Domain-Adversarial Training of Neural Networks , 2015, J. Mach. Learn. Res..

[18]  Sebastian Nowozin,et al.  f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization , 2016, NIPS.

[19]  Aaron C. Courville,et al.  Improved Training of Wasserstein GANs , 2017, NIPS.

[20]  Stefano Ermon,et al.  A DIRT-T Approach to Unsupervised Domain Adaptation , 2018, ICLR.

[21]  Jianmin Wang,et al.  Partial Adversarial Domain Adaptation , 2018, ECCV.

[22]  Jianmin Wang,et al.  Partial Transfer Learning with Selective Adversarial Networks , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[23]  Alexander J. Smola,et al.  Detecting and Correcting for Label Shift with Black Box Predictors , 2018, ICML.

[24]  Taesung Park,et al.  CyCADA: Cycle-Consistent Adversarial Domain Adaptation , 2017, ICML.

[25]  Jian Shen,et al.  Wasserstein Distance Guided Representation Learning for Domain Adaptation , 2017, AAAI.