Bryan Catanzaro (https://twitter.com/ctnzr), the VP Applied Deep Learning Research at NVIDIA, joins Mark (https://twitter.com/Neurotic) and Melanie (https://twitter.com/nyghtowl) this week to discuss how his team uses applied deep learning to make NVIDIA products and processes better. We talk about parallel processing and compute with GPUs as well as his team’s research in graphics, text and audio to change how these forms of communication are created and rendered by using deep learning.

This week we are also joined by a special co-host, Sherol Chen (https://twitter.com/ffpaladin) who is a developer advocate on GCP and machine learning researcher on Magenta at Google. Listen at the end of the podcast where Mark and Sherol chat about all things GDC.

Bryan Catanzaro

Bryan Catanzaro is VP of Applied Deep Learning Research at NVIDIA, where he leads a team solving problems in domains ranging from video games to chip design using deep learning. Bryan earned his PhD from Berkeley, where he focused on parallel computing, machine learning, and programming models. He earned his MS and BS from Brigham Young University, where he worked on higher radix floating-point representations for FPGAs.

Bryan worked at Baidu to create next generation systems for training and deploying deep learning models for speech recognition. Before that, he was a researcher at NVIDIA, where he worked on programming models for parallel processors, as well as libraries for deep learning, which culminated in the creation of the widely used CUDNN library.

Cool things of the week

* NVIDIA Tesla V100s coming to Google Cloud site (https://cloudplatformonline.com/CP-2018-NA-NVIDIA.html)

* Automatic Severless Deployment with Cloud Source Repositories blog (https://cloudplatform.googleblog.com/2018/03/automatic-serverless-deployments-with-Cloud-Source-Repositories-and-Container-Builder.html)

* Magenta site (https://magenta.tensorflow.org/)

* NSynth Super site (https://magenta.tensorflow.org/nsynth-super)

* MusicVAE site (https://magenta.tensorflow.org/music-vae)

* Making music using new sounds generated with machine learnnig blog (https://www.blog.google/topics/machine-learning/making-music-using-new-sounds-generated-machine-learning/)

* Building Blocks of Interpretability blog (https://distill.pub/2018/building-blocks/)

Interview

* NVIDIA site (http://www.nvidia.com/page/home.html)

* NVIDIA GPU Technology Conference (GTC) site (https://www.nvidia.com/en-us/gtc/)

* CUDA site (https://developer.nvidia.com/cuda-zone)

* cuDNN site (https://developer.nvidia.com/cudnn)

* NVIDIA Volta site (https://www.nvidia.com/en-us/data-center/volta-gpu-architecture/)

* NVIDIA Tesla P4 docs (http://images.nvidia.com/content/pdf/tesla/Tesla-P4-Product-Brief.pdf)

* NVIDIA Tesla V100s site (http://www.nvidia.com/page/home.html)

* Silicon Valley AI Lab Baidu Research site (http://research.baidu.com/silicon-valley-ai-lab/)

* ICML: International Conference on Machine Learning site (https://icml.cc/)

* CVPR: Computer Vision and Pattern Recognition Conference site (http://cvpr2018.thecvf.com/)

Referenced Papers & Research:

* Deep learning with COTS HPC System paper (http://ai.stanford.edu/~acoates/papers/CoatesHuvalWangWuNgCatanzaro_icml2013.pdf)

* Building High-level Features Using Large Scale Unsupervised Learning paper (https://arxiv.org/pdf/1112.6209.pdf)

* OpenAI Learning to Generate Reviews and Discovering Sentiment paper (https://arxiv.org/pdf/1704.01444.pdf)

* Progressive Growing of GANs for Improved Quality, Stability, and Variation paper (https://arxiv.org/pdf/1710.10196.pdf) and CelebA dataset (http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html)

* High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs paper (https://arxiv.org/pdf/1711.11585.pdf)

* Deep Image Prior site (https://dmitryulyanov.github.io/deep_image_prior)

* How a Japanese cucumber farmer is using deep learning and TensorFlow blog (https://cloud.google.