The authors study monotone spectral filters, estimators that regularize regression by applying a monotone weighting rule to directions associated with different covariance eigenvalues.
Principal component regression dominates monotone spectral filters
The authors prove finite-sample risk guarantees placing optimally tuned PCR within a constant factor of every monotone spectral filter, with polynomial advantages on some problems.
Big Tech
Juno Kim · Hengyu Fu · Peter Bartlett · Jason D. Lee · Jingfeng Wu
University of California, Berkeley · Google DeepMind
Research Digest··2 min read
Kim et al.
Why this paper
From Google DeepMind and University of California, Berkeley
In one line
Principal component regression dominates all monotone spectral filters for linear regression.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§