eResearch NZ
Browse
Buliding_machine_learning_systems_on_microsoft_Azure_cloud_machines.pdf (965.32 kB)

Buliding machine learning systems on Microsoft Azure cloud machines

Download (965.32 kB)
presentation
posted on 2020-03-10, 03:54 authored by Chenghao CaiChenghao Cai, Jing SunJing Sun, Gill DobbieGill Dobbie

We complete two machine learning projects using the Microsoft Azure Virtual Machines.


The first project is automatic voice recognition, which is the use of machine learning techniques to convert human speech to text. We build Gaussian mixture models, hidden Markov models and deep neural networks on the Azure VM, then use 100 hours of voice data to train the models. We find that better machine learning models and more training data can lead to increased accuracy of voice recognition, while background noise can reduce the recognition accuracy.


The second project is automated program repair. In this project, machine learning models such as support vector machines and random forests are used to learn the semantics programs. The training of such machine learning models, model checking processes and constraint solving processes are completed using the Azure VM. As both the model checking and machine learning techniques require considerable computational resources, we suggest using these techniques with Azure cloud computing services.


ABOUT THE AUTHOR

Chenghao Cai is a PhD student at the School of Computer Science, University of Auckland, whose study has been financially supported by the China Scholarship Council (CSC). His PhD work provided substantial contributions to the field of automated software engineering, especially in the area of machine learning approach to formal design model repair. Chenghao is in stage of finishing the PhD study. His thesis consists of eleven chapters, where the content of the chapters is supported by internationally peer reviewed publications.


Chenghao has published ten research papers to date, among which was the 52-pages manuscript published in the Automated Software Engineering (ASE) journal. ASE is a top quality and prestigious international journal in the field of Software Engineering, which has

an A-tier ranking by the Computing Research and Education Association of Australasia (CORE). Furthermore, Chenghao received the Microsoft Asia Cloud Research Software Fellowship (CRSF) Award in June 2019.

History

Usage metrics

    eResearch NZ 2020

    Categories

    Licence

    Exports