SAGEMAKER WORKSHOP – BUILD, TRAIN AND DEPLOY ML MODELS AT SCALE WITH AMAZON SAGEMAKER
OCTOBER 8, 2018 | 9:30 A.M. – 12:30 P.M. ATKINSON HALL, QUALCOMM INSTITUTE, UC SAN DIEGO PLEASE REGISTER HERE
In this tutorial participants learn to solve Machine / Deep Learning problems using the tools available in the Amazon Web Services (AWS) cloud. The development and application of machine learning models is a vital part of scientific and technical computing. Increasing model training data size generally improves model prediction and performance, but deploying models at scale is a challenge. Participants will learn to use Amazon SageMaker, a new AWS service that simplifies the machine learning process and enables training on cloud stored datasets at any scale.
Applications will include:
The tutorial will walk attendees through the process of building a model, training it, and applying it for prediction. Working in web-based Jupyter Notebooks powered by AWS, we’ll explore common algorithms (e.g. k-means and PCA) and deep learning with MXNet and TensorFlow. Participants will become familiar with SDKs for Python and Spark and other APIs that make machine learning with AWS easy to use. With Amazon SageMaker, users take their code and analysis to the data, and participants will experiment on real-world datasets, such as Earth on AWS and the Cancer Genome Atlas. At the end of the session, attendees will have the resources and experience to start using Amazon SageMaker and other AWS services to accelerate their scientific research and time to discovery.
Chemistry, DeepChem: building an online compound solubility prediction workflow
Genomics, 1000 genomes dataset
Intended audience
Machine Learning Practitioners old and new: developers, scientists, data science practitioners, research staff, and any other interested persons. Participants should have some familiarity with:
AWS
Python
Jupyter notebooks
Basic machine learning methods
AGENDA9:30am: Speaker and Facilitator Introductions
9:35am: Introduction to Amazon Sagemaker
9:55am: Environment Setup
10:15am: Lab 1 – Digit Classification with the Amazon Linear Learner Algorithm; guided walk-through and recap
10:45am: Lab 2 – Distributed Training with TensorFlow (self-guided)
11:15am: Lab 3 – How to Bring Your Own Model (self-guided)
11:45am: Break
12:00pm: Lab 4 – Using Public Datasets (self-guided)
12:25pm: Closing and pizza
PREREQUISITES FOR WORKSHOP
AWS Account (already created)
Access to SageMaker, S3, ECR from your IAM role.
Access to SageMaker service role AmazonSageMaker-ExecutionRole or ability to create IAM roles.
*Please make sure to have these taken care of prior to the workshop. If you do not have an AWS account, please open an account following the directions on Blink here: https://blink.ucsd.edu/technology/cloud/aws/ (a $0 PO is fine.). If you have questions, please email matsonh@amazon.com.