turingbot 3.2.0

Python interface for TuringBot, a symbolic regression software that discovers mathematical formulas from data

TuringBot Python Library

Python interface for TuringBot, a symbolic regression software that discovers mathematical formulas from data.

TuringBot uses simulated annealing to search the space of possible mathematical expressions, finding accurate and compact formulas that describe your data. It supports 15 search metrics, train/test splits, and a wide range of mathematical functions.

Installation

pip install turingbot

You also need the TuringBot desktop application installed. Download it from turingbotsoftware.com/download.html.

Quick start

import time
from turingbot import simulation

sim = simulation()

# Start a symbolic regression search
# TuringBot is detected automatically if installed in the default location
sim.start_process(
    input_file="your_data.csv",  # CSV with input columns + target column
    search_metric=4,             # RMS error
    threads=4,
)

# Let TuringBot search for formulas
time.sleep(60)

# Read the best formulas found so far
sim.refresh_functions()
for f in sim.functions:
    print(f)

# Stop the search
sim.terminate_process()

Or using the context manager:

import time
from turingbot import simulation

with simulation() as sim:
    sim.start_process(
        input_file="your_data.csv",
        search_metric=4,
        threads=4,
    )

    time.sleep(60)
    sim.refresh_functions()

    for f in sim.functions:
        print(f)

If TuringBot is installed in a non-default location, pass the path explicitly:

sim.start_process(
    input_file="your_data.csv",
    path="/custom/path/to/TuringBot",
)

Default paths checked automatically:

OS	Path
Windows	C:\Program Files (x86)\TuringBot\TuringBot.exe
macOS	/Applications/TuringBot.app/Contents/MacOS/TuringBot
Linux	/usr/lib/turingbot/TuringBot

Using numpy arrays and pandas DataFrames

You can pass numpy arrays or pandas DataFrames directly instead of file paths:

import numpy as np
from turingbot import simulation

# Generate some data: y = 2*x1 + x2^2
data = np.column_stack([
    x1 := np.random.rand(100),
    x2 := np.random.rand(100),
    2 * x1 + x2 ** 2,
])

with simulation() as sim:
    sim.start_process(
        input_file=data,
        column_names=["x1", "x2", "y"],
    )
    # ...

If column_names is not provided, TuringBot will use its default names (col1, col2, ...).

Pandas DataFrames work the same way, and column names are preserved automatically:

import pandas as pd
from turingbot import simulation

df = pd.read_csv("your_data.csv")

with simulation() as sim:
    sim.start_process(input_file=df)
    # ...

Input format

The input file should be a CSV or TXT file where the last column is the target variable. Example:

x1,x2,y
0.1,0.2,0.24
0.3,0.4,0.55
0.5,0.6,0.81

Search metrics

ID	Metric
1	Mean relative error
2	Classification accuracy
3	Mean error
4	RMS error (default)
5	F-score
6	Correlation coefficient
7	Hybrid (CC + RMS)
8	Maximum error
9	Maximum relative error
10	Nash-Sutcliffe efficiency
11	Binary cross-entropy
12	Matthews correlation coefficient
13	Residual sum of squares
14	Root mean squared log error
15	Percentile error

Parameters

See the full list of parameters in the start_process docstring, including:

• threads -- Number of threads
• train_test_split -- Train/test split ratio
• maximum_formula_complexity -- Max complexity of formulas
• integer_constants -- Restrict to integer constants
• normalize_dataset -- Normalize input data
• allowed_functions -- Restrict which math functions to use
• custom_formula -- Provide a formula template

Documentation

Full documentation: turingbotsoftware.com/documentation.html

License

MIT