The Microsoft Windows Azure-based system for neural network learning as an example of cloud processing application

Dariusz R Augustyn, Kamil Badura

Abstract


The paper presents the system for neural network learning based on the idea of Cloud computing. System implementation uses Microsoft Windows Azure technology. The well-known learning algorithm i.e. back propagation method was adopted for parallel and distributed execution. The architecture of cooperative worker role processes was proposed. The paper describes applying of methods of data storage like Windows Azure Table, Queue, Blob. The advantages of parallelization result from either applying multiple processes (instances) of WorkerRoles or applying Parallel Extension for .NET module in WorkeRole’s implementation.

Keywords


cloud computing; Microsoft Windows Azure; parallel processing; neural network learning

Full Text:

PDF (Polski)

References


Windows Azure (2012), http://msdn.microsoft.com/en-us/library/ddl79367.aspx.

Tour - Overview - Windows Azure (2012), http://www.windowsazure.com/en-us/ home/tour/overview.

Cloud Storage - Windows Azure (2012), http://www.windowsazure.com/en-us/ develop/net/fundamentals/cloud-storage.

Haridas J., Nilakantan N., Caíder B.: Windows Azure Table, http://www.scribd.com/ doc/63485303/Windows-Azure-Table-May-2009.

Tadeusiewicz R.: Sieci neuronowe. Akademicka Oficyna Wydawnicza, Warszawa 1993

Wilson R., Martinez T.: The General Inefficiency of Batch Training for Gradient Descent Learning. Neural Networks, 2003.

Prechelt L.: A Set of Neural Network Benchmark Problems and Benchmarking Rules. Technical Report, Karlsruhe 1994.

Parallel Programming in the .NET Framework (2012), http://msdn.microsoft.com/en-us/library/dd460693.aspx.

Data Parallelism (Task Parallel Library) (2012), http://msdn.microsoft.com/en-us/library/dd537608 .aspx.




DOI: http://dx.doi.org/10.21936/si2012_v33.n2A.132