AI engineering glossary · Deep learning core
What is lottery ticket hypothesis?
Lottery ticket hypothesis: The proposal that a large randomly initialised network contains a small sub-network which, trained alone from its original starting weights, can match the full network's accuracy. It motivated research on pruning and on why large networks train well.
Lottery ticket hypothesis: an example
Pruning 90 percent of the weights and resetting the rest to their initial values can still reach similar accuracy in the original paper's experiments.
Related terms
More Deep learning core terms
- LSTM
- Maxout
- Mixed precision training
- Model parallelism
- Multi-task learning
- Multilayer perceptron
- Negative log-likelihood
- Neural network pruning
Where you learn it
Lottery ticket hypothesis is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.