NeurIPS 2018: Intel Advances AI Research and Development

The Conference on Neural Information Processing Systems (NeurIPS) is a multi-track machine learning and computational neuroscience conference. It includes invited talks, demonstrations, symposia, and oral and poster presentations of papers selected by the program chairs. Originally established in 1986 as a conference for exploring biological and artificial neural networks, NeurIPS has become a highly-anticipated annual gathering for the AI community of data scientists, researchers, and academics.

This year’s sold-out conference runs Monday, December 3 through Saturday, December 8 at the Montreal Convention Centre (Palais des Congrès de Montréal). For the first time in the conference’s 32-year history, an Expo Day precedes the NeurIPS event on Sunday, December 2.

Intel is a proud diamond-level sponsor and participant at NeurIPS. We are committed to breaking the barriers between model and reality for AI developers with dedicated hardware, software, and tools, and we also support the AI community with our own research efforts. Among our contributions to NeurIPS 2018 are 11 accepted conference papers, including two spotlight papers. Read on for a complete guide to all our activities at NeurIPS 2018.

NeurIPS 2018 Expo Day – Sunday, December 2

The NeurIPS Expo is a one-day event taking place Sunday, December 2, prior to the NeurIPS conference, that provides a showcase for sponsor technologies. Remember, while the NeurIPS Expo is free of charge, registration is required. Please join the team from Intel AI at the following sessions:

Intel® AI DevCamp and Intel® AI DevJam on NeurIPS 2018 Expo Day – Sunday, December 2

Intel is hosting two exclusive events on NeurIPS 2018 Expo Day at the St. James Theatre (265 St. Jacques Street, two blocks from the Convention Centre). They are free to attend but require registration in advance as space is limited.

  • Intel AI DevCamp 2:00 pm – 6:00 pm Take your research and knowledge to the next level at this 4-hour workshop-style training session where you can join other developers to see how the AI community is advancing and solving complex problems. Intel AI evangelists will deliver hands-on and lecture-based training on topics like deep learning, computer vision, and reinforcement learning. Take a look at demos that show novel ways your contemporaries are applying AI. Then try your own hand in a mini AI competition to test the skills you learned during the workshop.
  • Intel AI DevJam 6:00 pm – 9:00 pm The DevJam is your chance to meet developers showcasing projects on a variety of AI frameworks and tools including TensorFlow*, Caffe*, the Intel® Distribution of OpenVINO™ toolkit, Intel Movidius™ technology, Intel FPGAs, and Intel® RealSense™ Technology. Over a dozen poster sessions will be on display from our worldwide Intel® AI Student Ambassador community. Bring your research questions, and network with leaders from the AI space. You can also learn about Intel career opportunities when you visit with recruiters at the event and enjoy great giveaways and delicious food.

Accepted Papers

Intel supports the AI developer community with a breadth of research efforts. Among our contributions to NeurIPS 2018 are 11 accepted conference papers (including two spotlight papers), as well as 10 workshop papers. Please join us at our poster and workshop sessions and spotlight talks to meet our Intel AI researchers.

Monday, December 3rd

  • Using Multi-Task Learning to Improve Out-of-Distribution Generalization in Pose Estimation
    Ahanita Bhiwandiwalla and Kavya Ravichandran (Intel summer intern)
    Presented at WiML 1:00 PM – 2:30 PM
  • Conversational Intent Understanding for Passengers in Autonomous Vehicles
    Eda Okur, Shachi H Kumar, Saurav Sahay, Asli Arslan Esme, Lama Nachman
    Presented at WiML 1:00 PM – 2:30 PM
  • The Importance of Socio-Cultural Differences for Annotating and Detecting the Affective States of Students
    Eda Okur, Sinem Aslan, Nese Alyuz, Asli Arslan Esme, Ryan S. Baker
    Presented at WiML 1:00 PM – 2:30 PM

Tuesday, December 4th

Wednesday, December 5th

Thursday, December 6th

Friday, December 7th

Saturday, December 8th

Intel AI Sponsored Workshops: Monday, December 3 – Saturday, December 8

Intel AI is proud to extend our support for the AI developer community by sponsoring five workshops throughout the conference, focused on diversity in data science and applying AI for greater social good.

  • Bronze sponsor of LatinX in AI Saturday, December 8 8:00 AM- 7:00 PM

Intel Demos: Monday, December 3 – Friday, December 7

Intel is showcasing its technology developments and research efforts in Booth #403 at the Montreal Convention Center. Our demos span topics that range from computer vision technology, model optimization, natural language processing, and reinforcement learning to privacy in machine learning.

Intel AI Researchers, Data Scientists, and Developers look forward to meeting you at the Intel AI booth to meet up and answer questions. Ask about the Intel AI Passport Program and get some great swag.

Booth attendees can also get professional headshots taken – complements of Intel. You may want to get a haircut ahead of the conference so that you look sharp on the day. These photos are ideal for your LinkedIn profile.

Computer Vision Demos

  • Intel® Neural Compute Stick 2 (Intel® NCS2)Powered by the Intel® Movidius™ Myriad™ X Vision Processing Unit (VPU), the Intel NCS2 is a plug-and-play AI development kit that fits in your pocket.
  • Speech Recognition Using Intel® Distribution of OpenVINO™ Toolkit This demo showcases a speech recognition model that translates a .wav input file to text and executes robotic code driving a bionic arm. You can access the demo’s code here.
  • Human Pose Estimation Using Intel Distribution of OpenVINO Toolkit By leveraging the Intel Distribution of OpenVINO toolkit and Intel® processors, the lightweight model in this demo performs a resource-consuming task in real time.
  • NASA FDL (Frontier Development Lab): Learn how Intel AI solutions accelerate lunar exploration by using algorithms to map images and detect the presence of craters.
  • Large-Memory Training on CPUs:  High-Content Screening for Drug Discovery High-content cellular microscopy is important in the field of drug discovery, but using it at scale poses data science and performance optimization challenges. In this demo, we show model convergence of the Multiscale Convolutional Neural Network topology powered by Intel® Xeon® Processors.
  • Robot Vision with the Intel® Neural Compute Stick 2 See how you can give your robotics project the ability to run neural network inferences and enhance the perception of the real world with the Intel® NCS2. This demo robot uses an object detection network to follow specific objects in real time.
  • Skeletal Tracking with Intel® RealSense™ Technology and 3DiVi The Intel® RealSense™ Depth Camera D400-Series uses stereo vision to calculate depth. Combined with 3DiVi facial and gesture recognition software, this solution enables skeletal tracking capabilities even in low light or bright sun, allowing for better results in a variety of conditions.

Natural Language Processing (NLP) demos

  • NLP Architect by Intel® AI Lab NLP Architect is an open source Python* library intended to be a platform for ongoing research and collaboration in NLP and Deep Learning. Code and further documentation can be found on our GitHub.
  • Unsupervised Aspect-based Sentiment Analysis This deep learning-based sentiment analysis system analyzes public opinions on selected entities (aspects) in Twitter and presents sentiment statistics that can be used by data scientists to extract business insight.

Privacy in Machine Learning Demos

  • Secure Federated Learning Intel has partnered with the University of Pennsylvania to show the first use of federated learning to train a deep learning model that identifies brain tumors from MRIs without needing to share sensitive patient data.
  • Homomorphic Encryption: Maintaining Privacy in Inference Tasks Homomorphic encryption, or the ability to perform computations on encrypted data, addresses increasing concerns about data privacy in machine learning. However, building models that operate on ciphertext is currently labor-intensive. Discover how Intel’s nGraph compiler can be extended to enable data scientists and developers to use their favorite deep learning frameworks to train and deploy models that operate on private, encrypted data.

Reinforcement Learning Demos

  • Deep Reinforcement Learning Research Collaborative Evolutionary Reinforcement Learning (CERL) is a multilevel optimization framework that leverages a collection of agents (learners) to solve a Reinforcement Learning (RL) problem. The learners explore independently while exploiting their collective experiences jointly through sharing of data and policies during learning. The ‘collective learner’ that emerges from this tightly integrated collaborative learning process combines the best of its composite approaches. This work will be presented at a NeurIPS competition track on Friday, December 7th.
  • RL Coach by Intel® AI Lab Coach is an open source reinforcement learning research framework containing the implementation of many state-of-the-art algorithms. It enables easy experimentation with existing algorithms and simplifies the development of new algorithms.

Model Optimization Demos

  • Next-Gen Intel® Xeon® Scalable Processors + Intel® Deep Learning Boost (Intel DL Boost) Discover how Intel DL Boost, a set of processor technologies designed to accelerate AI deep learning use cases, leverages new Vector Neural Network Instructions on next-gen Intel Xeon Scalable Processors.
  • Quantization Research + Distiller Learn about cutting-edge model optimization and quantization techniques to support high-speed inference using Intel DL Boost technology and Distiller, an open-source Python package for neural network compression research.
  • nGraph: Compiler for Cross-Hardware & Framework Flexibility The nGraph Compiler is an open source compiler, runtime and computational library for deep learning acceleration. This demo will show how nGraph can accelerate training & inference of a simple vision workload on TensorFlow*. Participants are encouraged to bring their own laptop running on either Linux & Mac OS and will be provided with step-by-step guidance to install and use the nGraph compiler with TensorFlow.

Intel® AI Academy

  • Meet Intel® Software Innovators and Student Ambassadors as they showcase their AI projects, leveraging Intel tools and optimized frameworks.
  • The Academy provides free courses, community & tools to boost AI development. Stop by the booth to meet some of our innovators and student ambassadors.

We’re all in for NeurIPS 2018

We hope you’ll join us at our Day 0 events, the Intel AI DevCamp and DevJam, and throughout the week at our poster presentations, workshops, and booth demos. Our ever-growing team of researchers is excited to talk to conference attendees about our current projects – and how we’re accelerating the path to future insights. Please follow us for ongoing news about AI advances – and have a great week at NeurIPS!

Intel technologies’ features and benefits depend on system configuration and may require enabled hardware, software or service activation. Performance varies depending on system configuration. No computer system can be absolutely secure. Check with your system manufacturer or retailer or learn more at intel.com.

Intel, the Intel logo, Movidius, OpenVINO, and Xeon are trademarks of Intel Corporation or its subsidiaries in the U.S. and/or other countries.

*Other names and brands may be claimed as the property of others.

© Intel Corporation

Related Posts