Jun Yang

Chevron down

Deep Learning with PAI: a Case Study of AliMe

We introduce a machine learning and deep learning platform in Alibaba called Platform of AI (PAI) and its application in AliMe assistant bot. PAI deep learning (DL) platform employs Caffe and TensorFlow as the base, and supports the training and deployment of a comprehensive suite of DL algorithms with distributed CPUs and GPUs. We have optimized the PAI to ensure the efficiency of the DL algorithms in the distributed mode. With our optimization techniques, some algorithms achieve the fastest performance in distributed mode so far as we know and have over 10X speed-up against the default distributed implementation. Furthermore, we showcase an application of PAI DL in building the intention mining and open-domain conversation modules for the AliMe assistant bot. Results show our modules achieve very good performance and scale well for industrial applications.

Jun Yang is an algorithm architect in the Alibaba Cloud iDST Large-Scale-Learning group, where he leads the R&D work related to large-scale deep learning. His team's major focus is speeding up the training/inference process of deep learning models from both engineering and modeling perspective, and exploring new kinds of models to boost business impact. Previously, Jun was an architect in the Qihoo 360 Advertising Technology Department and tech lead of the Yahoo! Beijing performance advertising team.

Buttontwitter Buttonlinkedin
This website uses cookies to ensure you get the best experience. Learn more