Random 70/30 Train/Test Split Can Fail Under Distribution Shift
A random 70%/30% split into training and test sets can be a bad idea when the available training distribution differs from the distribution the system ultimately needs to perform well on.
0
1
Tags
Machine Learning
Deep Learning
Machine Learning Strategy
Supervised Learning
Dive into Deep Learning @ D2L
Data Science
Machine Learning Yearning @ DeepLearning.AI
Related
Avoid Randomly Shuffling Mixed-Source Data into Dev/Test Sets
Include Some Target-Distribution Examples in Training Alongside Auxiliary Data
Down-Weighting Auxiliary Data from a Different Distribution
Training Dev Set
Error Table Across Two Data Distributions and Three Error Types
Data Mismatch Between Training and Dev Set Distributions
Limited Practical Scope of Domain Adaptation for Different Data Distributions
Domain Adaptation for Different Data Distributions
Website Images and Mobile Phone Pictures as a Distribution Mismatch Example
Random 70/30 Train/Test Split Can Fail Under Distribution Shift
Which data should define the dev and test sets for the cat-picture app?
Dev and test sets should represent the future data distribution of interest.
Complete the principle: Dev and test sets should reflect _____ data.
Match each cat-app data group with its appropriate role or property.
Order the decisions for building datasets when auxiliary and target data differ.
Explain why different training and evaluation distributions can be appropriate.
Diagnose the evaluation-set mistake in a mobile cat-classification app.
Why did success on website images fail to ensure success on mobile uploads?
Which dataset design best uses both target and auxiliary cat images?
Using extra internet images for training requires internet images in dev and test sets.
Learn After
Why can a random 70/30 train/test split be a bad idea in modern applications?
True or False: A random 70/30 split guarantees the test set matches the target distribution.
Before the modern era of big data, a common rule was to use a random _____ split for train/test sets.
Match each data scenario to whether a random 70/30 split is appropriate.
Order the reasoning steps for evaluating a random 70/30 split's suitability.
Explain why distribution mismatch undermines a random 70/30 split.
Diagnose a mobile app team's train/test split decision.
When does the random 70/30 split practice fail, according to the source?
Which example illustrates the risk of a random 70/30 split?
True or False: The 70/30 split was historically common before the era of big data.