OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Tpsr

TPSR is a Python research implementation of transformer-based planning for symbolic regression, combining Monte Carlo Tree Search with pretrained transformer decoding to search for symbolic equations that fit data. Its search uses fitting accuracy and equation complexity as feedback.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Tpsr — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use TPSR when you can prepare data and pretrained models locally and want to investigate equation discovery or run a configurable symbolic-regression benchmark workflow.

Who it is for

  • Researchers investigating symbolic regression or scientific equation discovery.
  • Technical teams comparing symbolic-regression methods on Feynman or PMLB benchmark datasets.
  • Python users seeking candidate equations for interpretable modeling and further analysis.

Documented uses

  • Search data for candidate symbolic expressions that balance fit and complexity.
  • Evaluate the TPSR planning method against a pretrained backbone on documented benchmark datasets.
  • Configure experiments with choices such as backbone, planning horizon, rollouts, tree width, and evaluation dataset.
Before you choose

Requirements & limits

The public repository is MIT-licensed and provides Python setup instructions using Conda and pip.

  • The documented default backbone is E2E; NeSymReS requires additional dependencies and installation steps.
  • The workflow depends on separately downloaded pretrained models and locally prepared datasets.
  • Run configuration exposes evaluation choices and MCTS parameters, including horizon, rollouts, width, and UCT algorithm.
  • The repository describes research code and benchmark procedures, not a hosted equation-discovery service.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Prepare the default E2E setup

Install the documented dependencies and place the required model and benchmark data in the repository structure before configuring an experiment.

Before you start

  • A local Python environment with Conda and pip.
  • The TPSR repository and its requirements.txt file.
  • The E2E pretrained model file.
  • A dataset such as Feynman or PMLB, if preparing a benchmark workflow.

Steps

  1. Create and activate the Conda environment with conda create --name tpsr and conda activate tpsr.

  2. Install the listed dependencies with the command below.

    Documented command
    pip install -r requirements.txt

  3. Download the E2E pretrained model and place it under symbolicregression/weights/.

  4. For benchmark data, place Feynman files under datasets/feynman/ or PMLB files under datasets/pmlb/.

Expected outcome

The documented E2E dependencies, model location, and optional benchmark dataset locations are prepared for the next experiment-configuration stage.

Guide limitations

  • This covers setup only; it does not verify installation or specify a complete execution command.
  • The README also documents a separate NeSymReS setup with additional dependencies.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
deep-symbolic-mathematics/TPSR · Official homepage title · retained in its source language
How we review sourcesSuggest a correction