{"id":22979,"date":"2019-07-24T08:30:10","date_gmt":"2019-07-24T15:30:10","guid":{"rendered":"https:\/\/insidebigdata.com\/?p=22979"},"modified":"2019-07-25T18:32:50","modified_gmt":"2019-07-26T01:32:50","slug":"darwinai-generative-synthesis-platform-and-intel-optimizations-for-tensorflow-accelerate-neural-networks","status":"publish","type":"post","link":"https:\/\/insidebigdata.com\/2019\/07\/24\/darwinai-generative-synthesis-platform-and-intel-optimizations-for-tensorflow-accelerate-neural-networks\/","title":{"rendered":"DarwinAI Generative Synthesis Platform and Intel Optimizations for TensorFlow Accelerate Neural Networks"},"content":{"rendered":"\n
\"\"<\/figure><\/div>\n\n\n\n

DarwinAI<\/a>, a Waterloo, Canada startup creating next-generation technologies for Artificial Intelligence development, announced that the company\u2019s Generative Synthesis<\/em> platform \u2013 when used with Intel technology and optimizations \u2013 generated neural networks with a 16.3X improvement in image classification inference performance. Intel shared the optimization results in a recently published solution brief<\/a>. <\/p>\n\n\n\n

\u201cThe complexity of deep neural networks makes them a challenge to build, run and use, especially in edge-based scenarios such as autonomous vehicles and mobile devices where power and computational resources are limited,\u201d said Sheldon Fernandez, CEO of DarwinAI.  \u201cOur Generative Synthesis platform is a key technology in enabling AI at the edge \u2013 a fact bolstered and validated by Intel\u2019s solution brief.\u201d<\/p><\/blockquote>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

DarwinAI and Intel Optimizations Deliver As Much As 16.3X Performance Increase<\/strong><\/p>\n\n\n\n

Intel engineers ran image classification\nperformance tests with ResNet50 and NASNet, two popular neural networks, using\nIntel Optimizations for TensorFlow with Intel Math Kernel Library (Intel MKL\nand Intel MKL-DNN). The results: for ResetNet50, Darwin\u2019s Generative Synthesis\nplatform combined with Intel\u2019s optimizations delivered a 16.3X improvement in\ninference speed over baseline measurements on an Intel Xeon Platinum 8153\nprocessor.  Meanwhile, for NASNet, the\nDarwinAI-Intel combination yielded a 9.6X improvement in inference speed on the\nsame hardware. <\/p>\n\n\n\n

\u201cAI Building AI\u201d Technology Produces High-Performance Deep Learning Solutions<\/strong><\/p>\n\n\n\n

DarwinAI\u2019s Generative Synthesis platform<\/a> uses AI itself to examine and learn from a neural network in order to construct new highly compact versions without sacrificing functional accuracy. This patented \u201cAI building AI\u201d technology dramatically reduces the size, complexity and guesswork in designing efficient, high-performance deep learning solutions.  This unique method of understanding also facilitates \u201cexplainable\u201d deep learning — the ability to understand why a network makes the decisions it does. <\/p>\n\n\n\n

The benefits of DarwinAI\u2019s Generative\nSynthesis technology include:<\/p>\n\n\n\n