AWS · MLA-C02 · BETA
Bias, class imbalance, and data for FM customization
MLA-C02 · Lesson 6 of 23 · 38:59
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Lesson 6 of the free MLA-C02 beta series (AWS Certified Machine Learning Engineer – Associate, domain 1, data preparation). In this lesson: measuring bias with SageMaker Clarify, mitigating bias and preparing multimodal data, handling class imbalance, and preparing data for foundation model customization and checking its integrity. With three quick checks and a recap, and AnyCompany Mobility, a car sharing service, as the running example.
Preview the topics
- The damage detection model works well on daytime photos but misses dents at night, and night photos are only three percent of the data. What should the ML engineer do first?
- A fraud model reports 99 percent accuracy, but it catches almost no fraud, which is one payment in five hundred. Which approach should the ML engineer take?
- The team has ten thousand support conversations, each with a question and an expert reply, and wants a model that answers in the same format and style. Which customization fits, with which data?