Purpose of error analysis in a mismatch study
Question: What is an error analysis trying to reveal when you suspect a mismatch between data sources?
Sample answer: It is trying to identify the important ways the training data differs from the dev data and to understand how those differences hurt performance.
Key points:
- Identify important differences.
- Compare the training data with the dev data.
Rubric: The correct answer must mention the differences between the training and dev sets and explain that the goal is to understand how those differences affect errors.
0
1
Tags
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Data Science
Machine Learning Strategy
Machine Learning Yearning @ DeepLearning.AI
Related
Train-Station Audio as a Speech Recognition Data Mismatch Example
What is the main purpose of error analysis when a model works well on suburban data but fails more often on downtown data?
Data Mismatch Is Diagnosed on the Test Set
The goal of error analysis in a data mismatch review is to identify the important _____ between the training data and the dev data.
Match each item to its role in a data mismatch review.
Arrange the steps for diagnosing a mismatch between training performance and dev-set performance.
What most directly causes a data mismatch between training and development performance?
Goal of Error Analysis in a Data-Mismatch Study
In a dataset-mismatch review, compare the training set against the _____ set.
Match each step in a distribution-gap review to its main purpose.
Order the steps for diagnosing a training–validation mismatch using error analysis.
Using error analysis to investigate a train-dev mismatch
Diagnosing a Training–Validation Gap
Purpose of error analysis in a mismatch study