University of Cambridge > > Machine Learning @ CUED > Towards true end-to-end learning & optimization

Towards true end-to-end learning & optimization

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Pat Wilson.

Deep neural networks automatically learn representations from raw data, but their architectures and hyperparameters still typically need to be defined manually by human experts. In this talk, I will discuss extensions of Bayesian optimization for effectively searching in this combined space of architectures and hyperparameters, thereby paving the way to fully automated machine learning (AutoML) with neural networks. I will first show competition-winning practical AutoML systems and then focus on speeding up AutoML (sometimes up to 100-fold) by reasoning over data subsets and partial learning curves. I will also briefly show related applications to the end-to-end optimization of algorithms for solving hard combinatorial problems and discuss recent progress on weight optimization by variants of stochastic gradient descent.

This talk is part of the Machine Learning @ CUED series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2024, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity