Preliminary program. Click the picture for a larger view.

Update: The bus leaves at 9.30 Monday, and we will be back Friday around 15.00.

 

Topics covered:

Lars Kai Hansen: Classics of semi-supervised learning

• Exponential convergence in mixture models with known input distribution (Castelli-Cover result)
• Co-training  (Blum, Mitchell)
• Kernel methods (clustering hypothesis, label propagation)
• Bayesian methods for principled integration of generative and discriminative models (Minka, Bishop Laserre)

Serge Belongie: Learning perceptual embeddings with triplet based comparisons

• Learning Concept Embeddings
• t-Stochastic Triplet Embedding (t-STE)
• Crowd Kernel Learning (CKL)
• Siamese Networks

Marco Loog: Minimax and projection estimators for semi-supervised learning

• EM, self-learning, and assumptions in SSL
• Safety, Contrast, and Pessimism

Ole Winther: Variational auto-encoders and their application to semi-supervised learning

 

For reading material go to: webpage.