Compression of machine-learned models via entropy penalized weight reparameterization
Number of patents in Portfolio can not be more than 2000
United States of America
Stats
-
Apr 1, 2025
Grant Date -
Jul 4, 2024
app pub date -
Jan 10, 2024
filing date -
Jan 10, 2024
priority date (Note) -
In Force
status (Latency Note)
Importance
Abstract
Example aspects of the present disclosure are directed to systems and methods that learn a compressed representation of a machine-learned model (e.g., neural network) via representation of the model parameters within a reparameterization space during training of the model. In particular, the present disclosure describes an end-to-end model weight compression approach that employs a latent-variable data compression method. The model parameters (e.g., weights and biases) are represented in a “latent” or “reparameterization” space, amounting to a reparameterization. In some implementations, this space can be equipped with a learned probability model, which is used first to impose an entropy penalty on the parameter representation during training, and second to compress the representation using arithmetic coding after training. The proposed approach can thus maximize accuracy and model compressibility jointly, in an end-to-end fashion, with the rate-error trade-off specified by a hyperparameter.
First Claim
Other Claims data not available
Family
International Classification(s)
- 2024 Application Filing Year
- G06N Class
- 5602 Applications Filed
- 354 Patents Issued To-Date
- 6.32 % Issued To-Date
Cited Art Landscape
Patent Citation Ranking
- 0 Citation Count
- G06N Class
- 0 % this patent is cited more than
- < 1 Age
Maintenance Fees
| Fee | Large entity fee | small entity fee | micro entity fee | due date |
|---|---|---|---|---|
| 3.5 Year Payment | $1600.00 | $800.00 | $400.00 | Oct 1, 2028 |
| 7.5 Year Payment | $3600.00 | $1800.00 | $900.00 | Oct 1, 2032 |
| 11.5 Year Payment | $7400.00 | $3700.00 | $1850.00 | Oct 1, 2036 |