# WassimTenachi/PhySO

Physical Symbolic Optimization

Repository: https://github.com/WassimTenachi/PhySO
Canonical: https://ross.abutalabs.com/products/physo
Homepage: https://physo.readthedocs.io/
Language: Python
License: MIT
License Family: permissive
Topics: deep-learning, machine-learning, physics, python, reinforcement-learning, symbolic-regression, equation-discovery
Last push: 2026-02-21T13:49:54+00:00

## Health v2 (maintenance only)
Score: 52/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 68, release rhythm 8, longevity 93
- inputs: {"age_days": 1314, "days_push": 193, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1971, forks 264 (observed 2026-08-28T04:06:00.866190+00:00)

## What it is
PhySO is a Python library for physical symbolic optimization that uses deep reinforcement learning to discover analytical physical laws from data via symbolic regression. It leverages physical unit constraints and dimensional analysis to reduce the equation search space, and supports fitting a single functional form across multiple datasets.

## Use cases
- discover physical equations from experimental data
- perform symbolic regression with unit constraints
- fit one analytical formula across multiple datasets
- recover equations like a damped harmonic oscillator from data points
- run symbolic regression robust to noisy measurements
- benchmark equation discovery on the Feynman lectures dataset

## When to choose
- you need interpretable analytical formulas rather than black-box models
- your data has known physical units that can constrain the search
- you want state-of-the-art symbolic regression performance under noise
- you work in Python with PyTorch and want pip/conda installability

## When to avoid
- you need general-purpose symbolic regression outside physics or without unit information
- you need a non-Python or GPU-cluster-scale production system
- you want purely numerical curve fitting without symbolic expressions

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, reinforcement-learning, data-science
- domain: machine-learning, deep-learning, data-science
- platform: python, cross-platform
- tags: symbolic-regression, equation-discovery, dimensional-analysis, physics, reinforcement-learning, pytorch, scientific-computing, algorithms

## Member repositories
- WassimTenachi/PhySO (main) score 52

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:00.866190+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:05:03.973163+00:00, confidence not recorded.
  - readme: https://github.com/WassimTenachi/PhySO (fetched 2026-08-28T04:06:00.866190+00:00, sha 704def8fbdb5)
  - registry_pypi: https://pypi.org/pypi/physo/json (fetched 2026-08-29T10:44:57.092234+00:00, sha 4726c700ebf8)
- Data as of 2026-08-30T08:39:29.467469+00:00.