### Download Example Data

Source: https://context7.com/sj001/ai-feynman/llms.txt

Downloads example datasets from the MIT Feynman Symbolic Regression Database. Useful for testing and experimentation with the library.

```python
import aifeynman

# Download example data files from MIT server
aifeynman.get_demos("example_data")

# Downloads files to ./example_data/ directory:
# - example1.txt (I.8.14: kinetic energy)
# - example2.txt (I.10.7: gravitational potential)
# - example3.txt (I.50.26: etc.)

# Run regression on downloaded example
aifeynman.run_aifeynman(
    pathdir="./example_data/",
    filename="example1.txt",
    BF_try_time=60,
    BF_ops_file_type="14ops.txt",
    polyfit_deg=3,
    NN_epochs=500
)
```

--------------------------------

### Install AI Feynman Environment

Source: https://github.com/sj001/ai-feynman/blob/master/README.md

Commands to create a virtual environment and install the necessary Python packages.

```bash
virtualenv -p python3 feyn
source feyn/bin/activate
pip install numpy
pip install aifeynman
```

--------------------------------

### Complete Symbolic Regression Workflow

Source: https://context7.com/sj001/ai-feynman/llms.txt

Demonstrates the full pipeline including directory setup, synthetic data generation, model execution, and result parsing.

```python
import numpy as np
import os
import aifeynman

# 1. Create working directories
os.makedirs("./my_experiment/data/", exist_ok=True)

# 2. Generate synthetic data for unknown equation
# True equation: f(a,b,c) = sqrt(a^2 + b^2) * exp(-c)
np.random.seed(42)
n_points = 5000
a = np.random.uniform(0.1, 5.0, n_points)
b = np.random.uniform(0.1, 5.0, n_points)
c = np.random.uniform(0.0, 2.0, n_points)
f = np.sqrt(a**2 + b**2) * np.exp(-c)

# Add small noise (optional, tests robustness)
noise = np.random.normal(0, 0.01 * np.std(f), n_points)
f_noisy = f + noise

# 3. Save data file
data = np.column_stack([a, b, c, f_noisy])
np.savetxt("./my_experiment/data/mystery.txt", data, fmt="%.10f")

# 4. Run AI Feynman
aifeynman.run_aifeynman(
    pathdir="./my_experiment/data/",
    filename="mystery.txt",
    BF_try_time=120,           # 2 minutes per brute force call
    BF_ops_file_type="14ops.txt",
    polyfit_deg=4,
    NN_epochs=2000,
    test_percentage=20
)

# 5. Load and display results
print("\n=== Discovered Equations (Pareto Frontier) ===\n")
try:
    results = np.loadtxt("results/solution_mystery.txt", dtype=str)
    if results.ndim == 1:
        results = results.reshape(1, -1)

    for i, row in enumerate(results):
        print(f"Solution {i+1}:")
        print(f"  Expression: {row[-1]}")
        print(f"  Test Error: {row[0]}")
        print(f"  Complexity: {row[3]} bits\n")
except Exception as e:
    print(f"Error loading results: {e}")
```

--------------------------------

### Execute AI Feynman via Terminal

Source: https://github.com/sj001/ai-feynman/blob/master/README.md

Example of invoking the symbolic regression solver directly from the command line.

```bash
python ai_feynman_terminal_example.py --pathdir=../example_data/ --filename=example1.txt
```

--------------------------------

### Run AI Feynman Symbolic Regression

Source: https://github.com/sj001/ai-feynman/blob/master/README.md

Python script to download example data and execute the symbolic regression solver.

```python
import aifeynman

aifeynman.get_demos("example_data") # Download examples from server
aifeynman.run_aifeynman("./example_data/", "example1.txt", 60, "14ops.txt", polyfit_deg=3, NN_epochs=500)
```

--------------------------------

### Run AI Feynman Symbolic Regression

Source: https://context7.com/sj001/ai-feynman/llms.txt

Primary entry point for symbolic regression. Requires data file and configuration for search parameters. Results are saved to results/solution_<filename>.txt.

```python
import aifeynman
import numpy as np

# Create sample data for the equation: f(x,y) = x^2 + y
# Data format: columns are [x, y, f(x,y)] separated by spaces
data = np.column_stack([
    np.random.uniform(0.1, 5.0, 1000),  # x values
    np.random.uniform(0.1, 5.0, 1000),  # y values
])
data = np.column_stack([data, data[:,0]**2 + data[:,1]])  # add f(x,y) column
np.savetxt("./my_data/", data)

# Run AI Feynman symbolic regression
aifeynman.run_aifeynman(
    pathdir="./my_data/",           # Directory containing the data file
    filename="mystery.txt",          # Name of the data file
    BF_try_time=60,                  # Brute force search time limit (seconds)
    BF_ops_file_type="14ops.txt",    # Operations file (7ops, 10ops, 14ops, or 19ops)
    polyfit_deg=4,                   # Maximum polynomial degree to try
    NN_epochs=4000,                  # Neural network training epochs
    vars_name=[],                    # Optional: variable names for dimensional analysis
    test_percentage=20               # Percentage of data for test set
)

# Results saved to results/solution_mystery.txt
# Each row: [test_error] log_err log_err_cumulative complexity error symbolic_expression
# Example output row: -15.2 -12000.5 25 1.2e-5 x0**2 + x1
```

--------------------------------

### Prepare Input Data

Source: https://context7.com/sj001/ai-feynman/llms.txt

Generates a space-separated text file containing independent variables followed by the dependent variable.

```python
import numpy as np

# Example: Create data for f(x1, x2, x3) = sin(x1) + x2*x3
n_samples = 10000
x1 = np.random.uniform(-np.pi, np.pi, n_samples)
x2 = np.random.uniform(0.1, 10, n_samples)
x3 = np.random.uniform(0.1, 10, n_samples)
y = np.sin(x1) + x2 * x3

# Stack columns: [x1, x2, x3, y]
data = np.column_stack([x1, x2, x3, y])

# Save with spaces (also supports tabs and commas)
np.savetxt("./data/my_function.txt", data, fmt="%.8f")
```

--------------------------------

### Run AI Feynman Regression

Source: https://context7.com/sj001/ai-feynman/llms.txt

Executes the symbolic regression process on a specified data file. Ensure the input file follows the required column-based format.

```python
aifeynman.run_aifeynman(
    pathdir="./physics_data/",
    filename="gravity.txt",
    BF_try_time=120,
    BF_ops_file_type="14ops.txt",
    polyfit_deg=4,
    NN_epochs=4000,
    vars_name=["G", "m1", "m2", "r", "F"],  # Must match units.xlsx
    test_percentage=20
)
```

```python
import aifeynman
aifeynman.run_aifeynman(
    pathdir="./data/",
    filename="my_function.txt",
    BF_try_time=60,
    BF_ops_file_type="14ops.txt",
    polyfit_deg=3,
    NN_epochs=2000
)
```

--------------------------------

### Configure Operations Files for Brute-Force Search

Source: https://context7.com/sj001/ai-feynman/llms.txt

Selects the set of mathematical operations for the brute-force search. Different files offer trade-offs between search speed and complexity.

```python
import aifeynman

# 7ops.txt: Basic operations (fastest)
# Symbols: + * D > ~ R 0
# Operations: add, multiply, divide, greater-than, negate, sqrt, zero
aifeynman.run_aifeynman("./data/", "file.txt", 30, "7ops.txt")

# 14ops.txt: Standard operations (recommended)
# Symbols: + * - D > < ~ I R P S C L E
# Operations: add, multiply, subtract, divide, >, <, negate,
#            inverse, sqrt, power, sin, cos, log, exp
aifeynman.run_aifeynman("./data/", "file.txt", 60, "14ops.txt")

# 10ops.txt and 19ops.txt also available for intermediate/extended searches
```

--------------------------------

### Enable Dimensional Analysis with Variable Names

Source: https://context7.com/sj001/ai-feynman/llms.txt

Provides variable names and units to AI Feynman for dimensional analysis, reducing problem complexity by identifying dimensionless combinations. Requires data and a units.xlsx file.

```python
import aifeynman
import numpy as np

# Data for gravitational force: F = G*m1*m2/r^2
# Variables: G (gravitational constant), m1, m2 (masses), r (distance), F (force)
G_vals = np.full(1000, 6.674e-11)
m1_vals = np.random.uniform(1e20, 1e25, 1000)
m2_vals = np.random.uniform(1e20, 1e25, 1000)
r_vals = np.random.uniform(1e6, 1e9, 1000)
F_vals = G_vals * m1_vals * m2_vals / r_vals**2

data = np.column_stack([G_vals, m1_vals, m2_vals, r_vals, F_vals])
np.savetxt("./physics_data/gravity.txt", data)

# Provide variable names matching entries in units.xlsx
```

--------------------------------

### Parse Regression Results

Source: https://context7.com/sj001/ai-feynman/llms.txt

Reads the generated solution file to extract and display Pareto-optimal equations and their associated metrics.

```python
import numpy as np
import aifeynman

# Run regression
aifeynman.run_aifeynman("./data/", "example.txt", 60, "14ops.txt", test_percentage=20)

# Read and parse results
results = np.loadtxt("results/solution_example.txt", dtype=str)

# With test_percentage > 0, columns are:
# [test_error, log_error, cumulative_log_error, complexity, error, expression]
for row in results:
    test_err = float(row[0])      # Error on held-out test set
    log_err = float(row[1])       # Mean log2 error (bits)
    cum_log_err = float(row[2])   # Cumulative log2 error
    complexity = float(row[3])    # Expression complexity (bits)
    error = float(row[4])         # Training error
    expression = row[5]           # Symbolic expression (e.g., "x0**2 + sin(x1)")

    print(f"Expression: {expression}")
    print(f"  Complexity: {complexity:.1f} bits")
    print(f"  Training error: {error:.2e}")
    print(f"  Test error: {test_err:.2e}")
    print()
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.