Notes
Short-form research notes by Markus Pollak. These are working documents: observations, design memos, and research-question writeups that sit between full papers and the code. They describe what I am working on, why, and what I expect to find.
Notes do not replace papers. A note is a thread that may or may not become a paper; a paper is a result I am willing to defend in the literature. Treat the two accordingly.
Latest
Quality you can read: R², discounted for size
A search that trades accuracy against size needs one number to rank candidates by. Ufinq uses description length for the search, and reports it on a scale every reader already knows: R², with the unexplained share inflated by the size of the formula.
Read note § 2.2 · 2026-10-01 · RESEARCHRoles instead of strategies: a design that was measured and removed
AHISTER used to give successor branches alternating search strategies, each with its own objective and gene pool. Three A/B comparisons showed no advantage over one objective with different branch roles, so the strategies were removed. A negative result, recorded.
Read note § 2.3 · 2026-10-01 · RESEARCHHow the benchmark is run, and what it does not show
The Ufinq benchmark runs every week against PySR, gplearn and Operon with one fixed configuration per algorithm. This note is about the rules that make the numbers worth reading, and about the limits that come with them.
Read note § 2.4 · 2026-05-02 · RESEARCHEvery Ufinq individual is already a local optimum
Standard genetic programming evolves raw candidates and lets selection do all the work. Ufinq does the opposite: every candidate is refined to a local optimum before it joins the gene pool. That single inversion explains an unusually large fraction of the system’s behavior.
Read note § 2.5 · 2026-04-30 · RESEARCHDiversity is two things, not one
Most genetic-programming systems collapse all diversity into a single concept. Ufinq treats structural diversity and behavioral diversity as independent properties, because they fail in different ways, for different reasons, at different times.
Read note § 2.6 · 2026-04-22 · RESEARCHApproximate evaluation, as a research problem
Full-fidelity evaluation of evolved candidates is the dominant compute cost in symbolic regression at scale. CRESCENT treats evaluation itself as a search problem: which candidates deserve a full score, and which can be safely scored cheaply?
Read note