Transfer Learning for Human & AI

Recent state-of-the-art active learning methods have mostly leveragedGenerative Adversarial Networks (GAN) for sample acquisition . However, GAN isusually known to suffer from instability and sensitivity to hyper-parameters . In contrast to these methods, we propose in this paper a novel active learningframework that we call Maximum Classifier Discrepancy for Active Learning(MCDAL) which takes the prediction discrepancies between multiple classifiers . We utilize two auxiliary classification layers that learntighter decision boundaries by maximizing the discrepancies among them . In this regard, wepropose a novel method to leverage the classifier discrepancies for theacquisition function for active learning . We also provide an interpretation ofour idea in relation to existing GAN based active learning

Author(s) : Jae Won Cho, Dong-Jin Kim, Yunjae Jung, In So Kweon

Links : PDF - Abstract

Code :


Keywords : active - learning - discrepancies - gan - classifier -

• Bioinformatics Specialization on Coursera
• fast.ai Machine Learning
• Deeplearning.ai Specialization by Andrew Ng
• Introduction to TensorFlow for AI, DL and ML by Andrew Ng
• Applied Data Science with Python
• Spark in R using sparklyr
• School of Data Science
• Python for Everybody Specialization