Core-Norm
Bounded, asymmetric and invertible preprocessing for numeric machine-learning features, with public benchmark evidence and browser-verifiable datasets.
SOFTWARE
Open implementations of research methods and scientific utilities. Each software entry is designed to expose the code, assumptions, evidence, and reproducibility material behind it—not only a polished interface.
SOFTWARE CATALOG
This catalog is data-driven so additional software can be added without redesigning the page or primary navigation. Each entry receives its own permanent software page.
Bounded, asymmetric and invertible preprocessing for numeric machine-learning features, with public benchmark evidence and browser-verifiable datasets.
Non-interventional mechanistic interpretability for neuron roles, layer fingerprints, and semantic pathways, with 45 browser-verifiable public datasets.
Local-first truth-vs-prediction evaluation for regression, classification, segmentation, and image registration—with visual diagnostics and report-ready outputs.
Paste or upload code and receive an inspectable asymptotic estimate, loop/recurrence evidence, space model, confidence, and limitations.
Find expensive structure choices—list membership, front removals, full sorts, dense graph storage—and get safer alternatives with complexity impact.
Inspect code structure and generate worst-path test ideas: absent targets, dense graphs, recursion depth, pathological ordering, and scale sweeps.
Compare inferred time/space growth, inspect evidence, estimate constant-factor cost, and find an approximate theoretical crossover point.
Apply inspectable, rule-based transformations such as list-queue → deque, full-sort → min/max, and repeated membership → set lookup.
Matrix multiplication intelligence with operation analysis, zero-skipping diagnostics, 64 public datasets/reference datasets, and browser-local user uploads.
Visual transpose, normalization, slicing, Hadamard, and Kronecker transformations over 64 public datasets/reference datasets or user-supplied matrices.
Sparsity, structural-pattern, and dense-versus-CSR storage intelligence across 64 public datasets/reference datasets and user-supplied numerical matrices.
Cell-by-cell matrix multiplication debugger that exposes the exact scalar products forming each selected output value on public or user-owned data.