The Computing
Services Centre (CSC) of City University of Hong Kong (CityU) organized a Seminar
entitled “NVIDIA Technology Sharing Session – AI Application and Research” on
14th Aug 2019. I met the
speaker Mr. Samuel Lo and took a photo for memory.
Mr. Samuel Lo
(General Manager, Nvidia AI Technology Center) was the first speaker and his
topic named “AI / HPC as a Service @ Higher Education”. Then he briefed Nvidia history which formed
by Jensen Huang in 1993 and the first GPU issued in 1996, then developed High Performance
Computing (HPC) in 2006. Recently, they focused on AI ecosystem development
since 2016.
Nvidia
developed HPC and AI had intersection in transforming science.
After that Mr.
Lo briefed the AI for Science using transformative tool to accelerate
innovation including accuracy improvement and accelerating time to solution.
The blueprint of Nvidia AI/HPC as a Service was mentioned that including Deep Learning (DL), Machine Learning (ML), Metropolis (Smart City), Clara (Healthcare), etc.
Mr. Lo highlighted
the dramatically improvement using GPU-accelerated Data Centre that only 1
T4-Accelerated Server (2K Watts) used equal to 200 CPU Servers (60K Watts)
Finally, he discussed some challenges utilizing AI & HPC software included expertise, installation, optimization, productivity and maintenance.
Lastly, Mr. Samuel
Lo described the cooperation between Data Scientist (Developer) and System Administrator
(DevOps) to drive productivity and faster discoveries. At the end, he briefed
the resource for AI/HPC included Deep Learning Institute, Inception for AI
Startups, GPU Technology Conference (GTC) and Nvidia AI Technology Centre, etc.
The second
speaker was Charles Lo (Nvidia) and his talk introduced The AI Brain for Smart
City and Smart Manufacturing. In this example, the system was able to identify suspicious
car
For foundry,
using AI detected defective wafer and got 7 times improvement.
Then Charles
briefed ISAAC SDK for robotics and using simulation and deploy to robot that
could learn much faster.
Intelligent
Video Analytics (IVA) Deep Learning Workflow Management was introduced and discussed
in which included Transfer Learning Toolkit and DeepStream Software Development
Kit (SDK). The DL workflow was showed as follows.
And then
DeepStream SDK transfer learning workflow was mentioned. (Now it has just launched DeepStream SDK 4.0)
The last
speaker was Mr. Juniper who discussed the Clara Echo for healthcare session.
Firstly, he introduced Clara AI which used to learn different unlabeled
data from MR, CT, X-ray and Ultrasound for diagnosis.
Then he
introduced the Clara Train SDK in which contained 12 pre-trained models in
different syndromes.
After that he
mentioned different challenges in medical deep learning included Data
Annotation, Algorithms and Training Pipelines.
Finally, he
briefed how to get start the Clara Train SDK.
Lastly, he
introduced AI employed in Financial sector included Investment, Credit Scoring,
Fraud Detection and Customer Experience.
Reference:
沒有留言:
發佈留言