While NVIDIA was mostly known among gamers and Cryptocurrency miners, in recent years it has earned a name for itself in the Machine Learning community. NVIDIA has been powering the Deep Learning revolution for about a decade now. NVIDIA GPU Technology Conference or GTC is one of the most awaited events of the year. Machine Learning enthusiasts wait for NVIDIA GTC to learn about the recent developments in the field of Hardware accelerators, software optimizations, new libraries among other things. And in March 2022, GTC is back with another edition of GTC.
This edition of GTC also came with a lot of sessions on various topics from Augmented Reality, Gaming, Machine Learning, and Omniverse across a host of industries. In this blog post, we will talk about the 5 sessions at NVIDIA GTC 2022 that you should absolutely attend if you are a beginner in the field of Machine Learning and Data Science.
5 Steps to starting a career in AI – SE2572
This is a particularly interesting session for beginners to attend. The session is designed for people who are beginners in the field of Machine Learning and Data Science and aspire to make a career in the same. The speakers will give insights into their journey and discuss the top five most practical steps you can take when beginning a career in artificial intelligence.
A Developer’s Guide to Choosing the Right GPUs for Deep Learning on AWS (Presented by Amazon Web Services) – S42497
Let’s face it, choosing the right GPU is something that every Machine Learning Engineer and Data Scientist has spent countless hours on. Amazon Web Services(AWS) is the leading cloud provider in the world and provides various CPU and GPU-based hardware architectures for supporting Machine Learning models during Training and Inference.
In this session, the speaker will take on a deep dive into how to choose the right GPU-based EC2 instance on AWS for your deep learning projects from the most performant instance for training, the best instance for prototyping, and the most cost-effective instance for inference deployments. At the end of the session, you will be capable to make an informed choice for GPU instances for every machine learning workload.
AI For Good: Intersecting Humanity and Intelligent Systems – S42250
While Deep Learning models keep on achieving state-of-the-art results in various fields and applications. However, they are Black Box models and this raises many ethical questions about biases and transparency of Machine Learning models. In recent years there has been a very high interest in the field of Fair AI, Explainable AI(XAI), and ethical AI.
This session will be about industry trends in making our world more equitable with the help of AI.
AI Building Blocks for Industry 4.0 (Presented by Supermicro) – S42564
Industry 4.0 and the critical place that Artificial Intelligence has it has long been discussed. Industry 4.0 has wide application in the field of health care, finance, services, manufacturing, energy, and other industries. One of the prime concerns of Industry 4.0 is to preserve privacy. This is where Federated Learning comes in. Federated Learning enables collaboration to share data, while preserving data privacy, to build better AI models with reduced bias.
In this session, Supermicro will describe an implementation of federated deep learning using Supermicro AI platforms for health care and life sciences, using a resilient architecture built with MONAI, NVIDIA FLARE, NVIDIA CLARA, Kubernetes, and CEPH.
Convolutions vs Transformers: From EfficientNets V2 to CoAtNet – S42621
At the starting of the previous decade, Convolutional Neural Networks changed the course of Deep Learning and Computer Vision forever when AlexNet crushed the ImageNet challenge.
However, after the recent success of Transformers in the field of Natural Language Understanding models such as Bert, GPT-2, and GPT-3, there has been some interest in the field of Transformers in the field of Computer Vision as well. Although earlier this year, a team at Facebook AI Research(FAIR) has already proved the latest generation of ConvNets- the ConvNext have better accuracy as compared to vision transformers.
This session will discuss in detail the properties of convolutions and transformers, and explore how to effectively combine the benefits of both with a hybrid approach. The session will also show how a properly scaled hybrid model can achieve state-of-the-art results on both fully-supervised and zero-shot transfer accuracy on ImageNet.
Other Articles you might be Interested In
Trends in Artificial Intelligence in 2021
Qiskit opens applications for Quantum Error Correction Summer School 2022