Advanced Analytics on a Data Platform without Data
Project Member(s): Liu, W.
Funding or Partner Organisation: National ICT Australia
Commonwealth Scientific and Industrial Research Organisation (Commonwealth Scientific & Industrial Research Organisation)
Start year: 2015
Summary: Data owners increasingly expose their data to others to consume and analyse, but many analytics need to be done over a variety of federated data sources without moving all the data into a centralised data platform. Distributed data mining and machine learning research, in the past, have addressed some of the issues. However, there are a number of new challenges including adapting to heterogeneous environments cross different data owners, cross-cluster/data centres, Hadoop/Spark based systems, data source variety and real time streams. This collaborative research project will devise advanced analytics over a data platform that can deal with federated data sources and streams without needing to hold all data in the platform. This includes new algorithm designs for both primitive operations (e.g. tensor / matrix factorisation) and high-level data analytics and mining (e.g. graph mining, ensemble learning) for such scenarios.
Verma, S, Liu, W, Wang, C & Zhu, L 2018, 'Hybrid networks: Improving deep learning networks via integrating two views of images', Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), International Conference on Neural Information Processing, Springer Link, Siem Reap, Cambodia, pp. 46-58.
View/Download from: UTS OPUS or Publisher's site
FOR Codes: Pattern Recognition and Data Mining, Information Processing Services (incl. Data Entry and Capture)