Poindexter/docs/plans/2026-02-16-math-expansion-design.md
Claude fa998619dc
Some checks failed
CI / build-test-wasm (pull_request) Has been cancelled
CI / build-test-gonum (pull_request) Has been cancelled
Tests / test (pull_request) Has been cancelled
feat: Add math modules — stats, scale, epsilon, score, signal
Centralizes common math operations used across core/go-ai, core/go,
and core/mining into Poindexter as the math pillar (alongside Borg=data,
Enchantrix=encryption).

New modules:
- stats: Sum, Mean, Variance, StdDev, MinMax, IsUnderrepresented
- scale: Lerp, InverseLerp, Remap, RoundToN, Clamp, MinMaxScale
- epsilon: ApproxEqual, ApproxZero
- score: WeightedScore, Ratio, Delta, DeltaPercent
- signal: RampUp, SineWave, Oscillate, Noise (seeded RNG)

235 LOC implementation, 509 LOC tests, zero external deps, WASM-safe.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-16 16:16:49 +00:00

46 lines
1.6 KiB
Markdown

# Poindexter Math Expansion
**Date:** 2026-02-16
**Status:** Approved
## Context
Poindexter serves as the math pillar (alongside Borg=data, Enchantrix=encryption) in the Lethean ecosystem. It currently provides KD-Tree spatial queries, 5 distance metrics, sorting utilities, and normalization helpers.
Analysis of math operations scattered across core/go, core/go-ai, and core/mining revealed common patterns that Poindexter should centralize: descriptive statistics, scaling/interpolation, approximate equality, weighted scoring, and signal generation.
## New Modules
### stats.go — Descriptive statistics
Sum, Mean, Variance, StdDev, MinMax, IsUnderrepresented.
Consumers: ml/coverage.go, lab/handler/chart.go
### scale.go — Normalization and interpolation
Lerp, InverseLerp, Remap, RoundToN, Clamp, MinMaxScale.
Consumers: lab/handler/chart.go, i18n/numbers.go
### epsilon.go — Approximate equality
ApproxEqual, ApproxZero.
Consumers: ml/exact.go
### score.go — Weighted composite scoring
Factor type, WeightedScore, Ratio, Delta, DeltaPercent.
Consumers: ml/heuristic.go, ml/compare.go
### signal.go — Time-series primitives
RampUp, SineWave, Oscillate, Noise (seeded RNG).
Consumers: mining/simulated_miner.go
## Constraints
- Zero external dependencies (WASM-compilable)
- Pure Go, stdlib only (math, math/rand)
- Same package (`poindexter`), flat structure
- Table-driven tests for every function
- No changes to existing files
## Not In Scope
- MLX tensor ops (hardware-accelerated, stays in go-ai)
- DNS tools migration to go-netops (separate PR)
- gonum backend integration (future work)