Skip to content

constrained_lasso: say where the CLA alone falls short - #65

Merged
tschm merged 2 commits into
mainfrom
constrained-lasso-wording
Oct 24, 2026
Merged

tschm merged 2 commits into
mainfrom
constrained-lasso-wording

Conversation

@tschm

@tschm tschm commented Oct 24, 2026 •

Copy link
Copy Markdown
Member

Wording fixes ahead of the JRS submission, following #64.

  • Section 1: the prediction that a portfolio CLA returns the knots "with no change to its code" fell short in two ways, not one. Besides the two defects fixed in cvxcla 2.6.0, degenerate starts need cvxcla.Lasso to run the CLA twice on a relaxed problem and combine the paths, outside the CLA.
  • Section 4: the penalty form cannot take a zero sum directly; it reaches one by relaxing the constraint, as Section 6 and Table 5 show.
  • Section 5, Scale: name the machine behind the cost ratios, an Apple M4 Pro; the text said only "one machine".

Compiles to 15 pages, no overfull lines.

🤖 Generated with Claude Code

tschm and others added 2 commits October 24, 2026 16:48
Two wording fixes ahead of submission:

- Section 1: the prediction that a portfolio CLA returns the knots with
  no change to its code fell short in two ways, not one. Besides the two
  defects fixed in cvxcla 2.6.0, degenerate starts need cvxcla.Lasso to
  run the CLA twice on a relaxed problem and combine the paths, outside
  the CLA.
- Section 4: the penalty form cannot take a zero sum directly; it reaches
  one by relaxing the constraint, as Section 5 and Table 5 show.

Compiles to 15 pages, no overfull lines.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The cost ratios were measured on an Apple M4 Pro (as recorded in d48f37e);
the text said only "one machine". Ratios travel better than times, but
naming the machine lets a reproducer judge how far theirs differs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@tschm
tschm merged commit a783bbc into main Oct 24, 2026
5 checks passed
@tschm
tschm deleted the constrained-lasso-wording branch October 24, 2026 15:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant