NeuroSolutions CUDA - Sample OCR Project

This sample project uses a Optical Character Recognition (OCR) dataset along with a single hidden-layer Mutli-Layer Perceptron (MLP) neural network using the Levenberg-Marquardt (LM) learning algorithm. The Levenberg-Marquardt is the default learning algorithm in NeuroSolutions and is the most powerful form of back-propagation learning available.

Test Environment

Dell Studio XPS 9000
Operating System	Windows 7 64-bit
Processor	Intel® Core™ i7-920 processor(8MB L3 Cache, 2.66GHz)
Memory	9GB DDR3 SDRAM at 1066MHz - 6 DIMMs
Video Card	NVIDIA GeForce GTS 240 1024MB
Hard Drive	750GB 7200 RPM SATA Hard Drive

Benchmarks

The sample project is trained on a total of 17,000 exemplars/samples for 100 epochs in NeuroSolutions. The graph below shows that the simulation run on a traditional (CPU) processor took 3-hours, 27-minutes and 8-seconds to complete while the fastest CUDA-enabled version of NeuroSolutions took only 3-minutes and 16-seconds to complete - that is a 65 times speed increase!

Through our research with several varieties of NVIDIA graphics cards we've found that the performance increase is fairly linear to the number of cores the graphics card contains. That means even with a mid-grade NVIDIA GeForce GTX 260 you could see speed increases up to 20 times faster than a traditional (CPU) processor for less than $200 in upgrade cost!

Download this project and try for yourself - LetterRecognition.zip

NeuroDimension, Inc. announces the release of NeuroSolutions 6.07

"BTW, love the product. To keep current, I've tried others and nothing even comes close to the power, performance, flexibility and ease of use. You also have the best tech support on the planet."
- John Wester, CTO, Net Shepherd, Inc.