Synthetic examples that seem believable to people are often easier to make than examples that match a model's learned patterns.
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
What makes synthetic data difficult to use effectively?
Human-likeness guarantees machine-likeness in synthetic data
Synthetic Data Can Look Real to a Human First
Match each synthetic-data concept to the description it best fits in the realism dilemma.
Put the checks in a sensible order when deciding whether generated data is usable for training.
A team generates synthetic café-noise clips that human listeners say sound convincing. What should they check before using them for training?
A generated image can look ____ to a person while still triggering a detector that says it was synthesized.
Match Each Scenario to the Correct Realism Concept
Order the steps for checking whether synthetic customer-support chats are useful for training.
Why Synthetic Samples Can Fool People but Not Models
Testing Whether Synthetic Data Works for a Model
Why Human Approval Does Not Prove Synthetic Data Works for a Model