The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.
Error code: FileFormatMismatchBetweenSplitsError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Lottery Ticket Pocket
A reproducible Lottery Ticket Hypothesis retest on handwritten digits. The dense network is trained, globally pruned by weight magnitude to 20% of its connections, rewound to its exact initial weights, and retrained. An equally sparse random mask is trained from the same initialization as the causal control.
The saved report measures clean accuracy and Gaussian-noise accuracy for all three models. The Space runs them side by side on identical held-out examples.
Verified local result
The dense 6,570-parameter network reached 97.04% clean accuracy. Its 1,399-active-parameter magnitude ticket matched 97.04%, but an equally sparse random-mask control reached 97.78% and also led under Gaussian noise (93.33% versus 91.85%). On this benchmark, sparsity survived rewinding, but magnitude selection did not establish a special winning-ticket advantage.
uv run python projects/lottery-ticket-pocket/train.py
uv run pytest tests/test_lottery_ticket_pocket.py
Results are verified local artifacts; Hub deployment remains a separate step.
- Downloads last month
- 2