Automated Machine Learning and Neural Architecture Search applied to engineering problem solving
Project Leader: Saman Halgamuge
Staff: Damith Senanayake Wei Wang
Sponsors: Australian Research Council
Primary Contact: Saman Halgamuge (firstname.lastname@example.org)
Keywords: artificial intelligence; control and signal processing; deep learning; machine learning; mathematical modelling
Disciplines: Mechanical Engineering
The goal of machine learning is to automatically learn from data and make decisions or predictions. Traditionally, architectures of machine learning methods are manually designed and associated hyper parameters are set based on the experience of the designer. In the past decade, Deep Neural Networks (DNNs) based on Supervised Learning have revolutionised various fields. To reduce the development cost of DNNs, a recent idea proposed is to automate the DNN design, which leads to an emerging field called automatic machine learning (Auto-ML).
Existing Auto-ML methods have attempted to optimise every step of the data analysis pipeline including data preparation, feature engineering, model generation, training, and evaluation. Among them, Neural Architecture Search (NAS) methods that explicitly find DNN architectures for a given supervised learning task. This is achieved by encoding the candidate architecture as a solution in some search space and treating the architecture design as an optimisation problem. Self-Adapting Neural Network Architectures instead of ‘searching for the best’ has been our strategy to this problem.