Multi-Domain Adversarial Learning
arXiv:1903.09239
Abstract
Multi-domain learning (MDL) aims at obtaining a model with minimal average risk across multiple domains. Our empirical motivation is automated microscopy data, where cultured cells are imaged after being exposed to known and unknown chemical perturbations, and each dataset displays significant experimental bias. This paper presents a multi-domain adversarial learning approach, MuLANN, to leverage multiple datasets with overlapping but distinct class sets, in a semi-supervised setting. Our contributions include: i) a bound on the average- and worst-domain risk in MDL, obtained using the H-divergence; ii) a new loss to accommodate semi-supervised multi-domain learning and domain adaptation; iii) the experimental validation of the approach, improving on the state of the art on two standard image benchmarks, and a novel bioimage dataset, Cell.
Accepted at ICLR'19
Cited by in corpus (7)
- Learning Robust Global Representations by Penalizing Local Predictive Power
- Generalizing to unseen domains via distribution matching
- Estimating Counterfactual Treatment Outcomes over Time Through Adversarially Balanced Representations
- Domain Aggregation Networks for Multi-Source Domain Adaptation
- One Model to Serve All: Star Topology Adaptive Recommender for Multi-Domain CTR Prediction
- Adversarial Camera Alignment Network for Unsupervised Cross-camera Person Re-identification
- Multi-source Few-shot Domain Adaptation