Abstract
Many machine learning algorithms depend on hyperparameters. Hyperparameters are special parameters that cannot be learned such as the typical model parameters. They have to be set before the learning process. The choice of hyperparameters is crucial for the performance of an algorithm such that it is necessary that they are optimized. Often, hyperparameters are tuned manually by experts or using a grid search. Recent research has shown that automatic tuning can yield better results or at least saves you time.
In this seminar students will gather knowledge about alternative hyperparameter tuning strategies. We will focus on recent work in the field of automatic hyperparameter optimization. Attendance of the lecture "Machine Learning" may help but is not necessary.
The first meeting is on April 15th, 16:15 in D 017. Please have a look at the bibliography before the meeting.
Instructor: Martin Wistuba