Deep Learning Day @ SUTD

NVIDIA Openning Address

Dr. Simon See

NVIDIA: an AI computing company

---

Introduction tp Deep Learning with GPUs

Jeffrey Adie, NVIDIA

AI -> ML -> NN -> DL
Deep Learning benefits:

• Robust
• Generalizable
• Scalable

NVIDIA DGX-1: First Deep Learning Machine
CUDA for Deep Leaning Development
cuDNN: Deep Learning Primitives
NVIDIA DIGITS: Interactive Deep Learning GPU Training System

---

Learning to Hash and diagnose with DNN

Ngai-Man Cheung

• Hash based approximate for image data indexing.
  DL: learn the hash function to project from high dimension image feature vector to binary code

• Skin Cancer detection
  limited data -> trained CNN, use FC output as feature and feed into other classifiers

• Chronic Would Assessment

---

Deeper Learning for Smarter Cities

Lin Shaowei

1. Step1: Traditional Programming -> Intentional Programming
   Functional Program
2. Step2: Data and APIs
   Content-centric networking(CCN)
   Named data networking/Named function networking
   Security: Block Chain vs. ICANN & DNS
3. Message Passing(coordinate services through topics)
   Publish-subscribe protocols eg. MQTT(IBM)