E-ADDA: Unsupervised Adversarial Domain Adaptation Enhanced by a New Mahalanobis Distance Loss for Smart Computing.
Ye GaoBrian R. BaucomKaren RoseKristina GordonHongning WangJohn A. StankovicPublished in: SMARTCOMP (2023)
Keyphrases
- domain adaptation
- mahalanobis distance
- semi supervised
- metric learning
- noise model
- cross domain
- covariance matrix
- labeled data
- euclidean distance
- semi supervised learning
- unsupervised learning
- multiple sources
- transfer learning
- sentiment classification
- test data
- unlabeled data
- supervised learning
- data distribution
- noisy images
- learning algorithm
- maximum likelihood
- feature selection