Intel Nervana Update + Productizing AI Research
summary
Intel announces Neon 2.0 (optimized with Intel MKL) and Nervana Graph, a framework- and hardware-independent computational graph that plugs into TensorFlow and others, applies optimization passes, and targets CPUs, GPUs, FPGAs and custom silicon. Claim: GoogLeNet inference runs ~98× faster with these optimizations. Roadmap shows Nervana silicon prototypes this year and products in 2018–20, plus Nervana Cloud and on‑prem options. Practices presented: synthetic data from games/GANs, edge feedback loops for live annotation, and model surgery for rotated satellite object detection.
AI-written summary · about
chapters (4)
- 0:00Welcome and introductions
- 6:18Interview with Naveen Rao
- 15:44Interview with Hanlin Tang
- 41:30Closing announcements
related talks
32mTowards new deep learning abstractions on top of exist frameworks
39mIntel Nervana DevCloud
30mAI: unleashing the next wave
50mLakshmi Narasimhan of Intel at Machine Learning Developers Summit 2019
11mGadi Singer (Intel) interviewed at the O'Reilly Artificial Intelligence Conference NY 2019
13mAccelerating AI algorithms
24mTools & Techniques for a cost optimized Analytics & AI pipeline
05mAccelerate AI development from cloud to client