Intel AI Research at NeurIPS 2018

Now that the excitement of the 32nd annual Conference on Neural Information Processing Systems (NeurIPS) is upon us, it’s a great time to take stock of everything that Intel AI research has worked on this year, and how much we’ve grown: we have more papers to present, more open source software tools to share, and we’ve launched exciting new academic partnerships on three continents. Here are some of the highlights:

Research publications: This year we are pleased to announce that Intel researchers will be presenting 11 posters and 10 workshop papers at NeurIPS. Reflecting the breadth of technologies that Intel works on, these papers cover a wide range of topics, from fundamental research in structure learning, robustness to adversarial examples, limited-precision training, and combinatorial optimization, to applications in pose estimation, mapping for flood and pandemic response, and tackling the problem of information poverty in the developing world.

Open sourcing AI tools: One of the key reasons that deep learning has advanced so rapidly in the last few years is its culture of publishing results and open source software. In this spirit, we are strong advocates of releasing tools to the community to accelerate their research and development of AI systems. Here are some highlights of 2018:

  • Beta release of the nGraph compiler stack for deep learning systems
  • HE-transformer: an extension to the nGraph compiler that brings privacy-preserving machine learning (in the form of homomorphic encryption) to deep learning frameworks for research.
  • A new release of Reinforcement Learning Coach, plus integration with AWS Sagemaker. Coach is a research framework for training and evaluating reinforcement learning agents by harnessing the power of multi-core CPU processing to achieve state of the art results.
  • A new set of features for NLP Architect, our open source library for natural language processing research and development
  • Open-source availability of Neural Network Distiller, a Python* package for neural network compression research that provides a PyTorch* environment for prototyping and analyzing compression algorithms
  • Availability of the world’s first open dataset of driving conditions in India to facilitate autonomous vehicle training

Academic partnerships and collaborations: Intel has been an active participant in the academic research community ever since our founding 50 years ago. The field of deep learning, and AI more broadly, has recently experienced some of the fastest transitions of technologies from academic research to real-world products of any engineering field in recent memory. We have been collaborating with key AI researchers in academia—students and faculty—to shape the next 50 years of computing, which we believe will be ineluctably tied to the future of AI. Here are some of the highlights of programs that we initiated in 2018:

  • We announced the “AI Future Pioneers Program” at the Intel AI Developer Conference debut in Beijing on Nov. 14, 2018. The Program, which will work with over 20 universities and research institutes, aims to explore new development models through industry/university/academy collaborations and to promote multiple efforts such as joint research, faculty development, talent cultivation, and commercialization.
  • We announced a deep collaboration with the Technion’s AI Institute, a new center combining elements of Technion’s computer science, electrical engineering, industrial engineering and management departments, among others, all collaborating to drive a closer relationship between academia and industry in the race to AI.
  • We grew our Student Ambassador program to over 200 students from top institutions around the world using Intel AI in academic research, some of whom have published at NeurIPS and WiML this year. To see examples of Student Ambassador research and meet some of our Ambassadors first hand, join us at the Intel AI DevJam on Sunday, Dec 2nd or visit the Intel AI booth.
  • We launched the AI Residency program, taking in our first cohort of residents in September (and seeking applicants for 2019). Our AI Residents spend a year embedded with Intel AI researchers and collaborate with us on projects of their design with Intel mentorship across our entire range of AI technologies—spanning everything from materials science and computer architecture to algorithms and hyperscale systems.
  • We launched the Intel Strategic Research Alliance (ISRA) for Probabilistic Computing, and the Intel Neuromorphic Research Community (INRC), which enables researchers and developers to access our Loihi neuromorphic test chip.

It’s been a busy year at Intel AI research, and 2019 is shaping up to be even bigger. Here is a complete list of all of our activities at NeurIPS 2018, and we look forward to meeting you at our booth if you’d like to learn more and meet the team.