Baidu to Open Source New Platform for Deep Learning Community


BEIJING, CHINA--(Marketwired - Sep 1, 2016) - Baidu, Inc. (NASDAQ: BIDU), the leading Chinese language Internet search provider, today announced it will release an open source software platform for the deep learning community. 

The software will be released on GitHub on Sept. 30 with full documentation and specs. A pre-release alpha version is available now.

Originally developed for internal use by and for Baidu's own engineers, the platform provides an intuitive and powerful interface for data handling and specifying model structure. Its ease of use makes it a natural starting point for programmers and enthusiasts who want to apply deep learning to their projects and products. 

The platform, codenamed "PaddlePaddle," has been used to develop a range of Baidu's leading-edge products and technologies in areas such as advertising, search ranking, large-scale image classification, optical character recognition and machine translation.

The announcement was made at the annual Baidu World conference by Andrew Ng, Baidu chief scientist, who said, "PaddlePaddle opens up technology widely used by Baidu to programmers looking for easy-to-learn and easy-to-use tools to build their own deep learning applications."

Pieter Abbeel, Associate Professor, UC Berkeley EECS and Research Scientist, OpenAI, commented, "Progress in AI is critically dependent on time spent in software development to prototype and test new ideas. Sharing software development frameworks, like Baidu is doing with its platform, is key to accelerating progress for the entire community."

Xu Wei, Baidu distinguished scientist and leader of PaddlePaddle development, said, "With this platform, designing a deep learning model is like writing pseudocode. Engineers can focus on the high-level structure of their model without worrying about the low-level details. We expect it to be useful to programmers who want to quickly apply deep learning models to problems in areas that can really benefit from AI, such as health care and finance."

Dr. Ng added, "Other deep learning platforms have been a great boon to researchers wanting to invent new deep learning algorithms. but their high degree of flexibility limits their ease of use. In contrast, PaddlePaddle focuses on making it easy for enthusiasts and programmers -- not just machine learning researchers -- to learn and use powerful deep learning tools."

Features
Powerful: Supports widely-used neural network architectures, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs). 
Efficient: To leverage the power of heterogeneous computing, optimization occurs at a number of levels including computing, memory, architecture and communication. PaddlePaddle optimizes mathematical operations via state-of-the-art BLAS libraries, e.g. Intel MKL, ATLAS, OpenBLAS and cuBLAS, etc. 
Scalable: PaddlePaddle can easily scale to large numbers of GPUs/CPUs on multiple machines. 

Availability
PaddlePaddle will be released on GitHub on Sept. 30 with full documentation and specs. A pre-release alpha version is available now at https://github.com/baidu/paddle. PaddlePaddle is being released under an open-source Apache license.

About Baidu
Baidu, Inc. is the leading Chinese language Internet search provider. As a technology-based media company, Baidu aims to provide the best and most equitable way for people to find what they're looking for. In addition to serving individual Internet search users, Baidu provides an effective platform for businesses to reach potential customers. Baidu's ADSs trade on the NASDAQ Global Select Market under the symbol "BIDU". Currently, ten ADSs represent one Class A ordinary share.

Contact Information:

Contact:
Calisa Cole
Director, Communications
Baidu, Inc.
calisacole@baidu.com