Hyperparameter Optimization for Machine Learning

No Thumbnail Available

URL

Journal Title

Journal ISSN

Volume Title

Perustieteiden korkeakoulu | Master's thesis
Ask about the availability of the thesis by sending email to the Aalto University Learning Centre oppimiskeskus@aalto.fi

Date

2016-06-14

Department

Major/Subject

Applied Mathematics

Mcode

SCI3016

Degree programme

Master’s Programme in Applied and Engineering Mathematics (N5TeAM)

Language

en

Pages

73+0

Series

Abstract

In the recent years, there have been significant developments in the field of machine learning, with the modern methods like deep learning, significantly overpassing previous state-of-the-art results on a variety of tasks. These modern methods however, come at the cost of increased complexity and require careful tuning of multiple hyperparameters which specify the model. The common practice still is manual tuning of the hyperparameters, making the use of deep learning methods, more of an art than a science. In this thesis, we will explore some of the methods for automated hyperparameter optimization, focusing on the gradient-based approach. The goal is to provide a gentle introduction to this topic, by first providing a solid overview of essential concepts from both optimization and machine learning.

Description

Supervisor

Hollanti, Camilla

Thesis advisor

Raiko, Tapani

Keywords

hyperparameter, machine learning, deep learning, optimization, gradient-based

Other note

Citation