Cloud-based Machine Learning Offerings - Beginner's Perspective
summary
Compares hands-on runs of Jupyter notebooks on AWS SageMaker, Google AI Platform and Azure ML using the UCI Dota2 CSV, trains a scikit‑learn decision tree, and deploys models as endpoints. Reports roughly 53% accuracy on raw hero‑composition prediction, shows data prep with Azure’s visual drag‑and‑drop and three‑way splits, highlights AWS Mechanical Turk and an algorithm marketplace for production tuning, and praises Google’s AutoML for automated algorithm selection and simplified defaults.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:17Talk Outline
- 2:17What is Machine Learning?
- 5:13What is DotA 2?
- 7:10DotA Data - Enhanced
- 8:44Training the Model
- 9:23Using the Model
- 9:36Evaluating the Model
- 10:54Why Jupyter?
- 13:02Getting The Data Loaded - AV
- 16:23The Machine Learning Pipeline
- 19:11Getting The Data Loaded - GO
- 21:25AWS Mechanical Turk
- 23:06AWS: Many Options
- 29:47Google Cloud: Smaller Config
- 32:52Azure: Visual Pipeline Editor
- 36:20Big Takeaway
related talks
44mPragmatic Machine Learning in the Cloud
49mAutomated Machine Learning on Azure
58mDeep Dive into Azure Automated Machine Learning
01h 05mAzure Machine Learning for the Developer
44mPractical deep dive into machine learning techniques and MLOps
37mStreamlining Machine Learning Operations
50mBuilding Intelligent Mobile Apps using Machine Learning
52mGenerate ML predictions without writing any code