Possible papers and topics

The two first papers will be discussed in the first lecture(s)

Dropout: a simple way to prevent neural networks from overfitting

N Srivastava, G Hinton, A Krizhevsky¡­ - The journal of machine ¡­, 2014 - jmlr.org

Dropout as a bayesian approximation: Representing model uncertainty in deep learning

Y Gal, Z Ghahramani - international conference on machine learning, 2016 - jmlr.org

Boosting With the L₂ Loss: Regression and Classification

Concrete dropout

Y Gal, J Hron, A Kendall - Advances in Neural Information Processing ¡­, 2017 - papers.nips.cc

Deep residual learning for image recognition

K He, X Zhang, S Ren, J Sun - Proceedings of the IEEE ¡­, 2016 - openaccess.thecvf.com

Probabilistic machine learning and artificial intelligence

Z Ghahramani - Nature, 2015 - nature.com

On the difficulty of training recurrent neural networks

R Pascanu, T Mikolov, Y Bengio - International conference on machine ¡­, 2013 - jmlr.org

Auto-encoding variational bayes

DP Kingma, M Welling - arXiv preprint arXiv:1312.6114, 2013 - arxiv.org

Explainable artificial intelligence: Understanding, visualizing and interpreting deep learning models

Possible topics:

• Gradient boosting
• Reinforcement learning
• Variational inference
• Reccurent neural networks
• Auto-encoding variational Bayes
• Gradient Boosting
• Bagging/Random forrest
• Boltzman machines
• Variational inference
• Explainable AI (Lime, Shapley, partial dependency plots)
• Ensemble learning/model averaging
• Regularization (Lasso, Ridge, Elastic net)
• H20, Keras, Adam,
• Probabilistic graphical models
• AutoML (https://arxiv.org/abs/1907.00909)
•