Python GPU Statistical Analysis

A reference implementation for statistical analysis using GPU acceleration (CuPy/CUDA) with CPU fallback (NumPy).

Features

• GPU-accelerated Monte Carlo simulations using CuPy
• Automatic CPU fallback when GPU is unavailable
• Dynamic weight optimization via backtesting
• Sliding window validation for time-series data
• Historical optimization tracking

Performance

Benchmarks on Intel Xeon E5-2676 v3 + NVIDIA GTX 1660 SUPER:

Operation	CPU	GPU	Speedup
Monte Carlo (500K)	14.4s	0.47s	30x
Rate	37K sims/s	1.1M sims/s	-

Requirements

Minimum

• Python 3.11+
• Windows 10/11 or Linux
• 8GB RAM

GPU Support (Optional)

• NVIDIA GPU with Compute Capability 7.0+
• CUDA Toolkit 12.x
• NVIDIA Driver 525+

Installation

git clone https://github.com/GuilhermeP96/python-gpu-statistical-analysis.git
cd python-gpu-statistical-analysis

pip install -r requirements.txt

# Optional: GPU support
pip install cupy-cuda12x==13.6.0

Quick Start

from src.gpu_stats import GPUStatistics

# Initialize with automatic GPU/CPU detection
stats = GPUStatistics()

# Run Monte Carlo simulation
results = stats.monte_carlo(n_simulations=500000)

# Get performance metrics
print(f"Method: {results['method']}")
print(f"Rate: {results['rate']:,.0f} simulations/second")

Project Structure

python-gpu-statistical-analysis/
├── src/
│   ├── __init__.py
│   ├── gpu_stats.py          # Core GPU/CPU statistics module
│   ├── weight_optimizer.py   # Dynamic weight optimization
│   └── backtesting.py        # Backtesting framework
├── examples/
│   ├── monte_carlo.py        # Monte Carlo simulation example
│   ├── optimization.py       # Weight optimization example
│   └── benchmark.py          # CPU vs GPU benchmark
├── tests/
│   └── test_gpu_stats.py     # Unit tests
├── requirements.txt
├── setup.py
└── README.md

Usage

Monte Carlo Simulation

from src.gpu_stats import GPUStatistics

stats = GPUStatistics()

# GPU-accelerated Monte Carlo
result = stats.monte_carlo(
    n_simulations=1000000,
    batch_size=100000
)

print(f"Elapsed: {result['elapsed']:.2f}s")
print(f"Rate: {result['rate']:,.0f} sims/s")

Weight Optimization with Backtesting

from src.weight_optimizer import WeightOptimizer

optimizer = WeightOptimizer(data)

# Find optimal weights using historical validation
weights = optimizer.optimize(
    train_ratio=0.8,
    validation_ratio=0.2
)

print(f"Optimal weights: {weights}")

CPU/GPU Benchmark

from src.gpu_stats import GPUStatistics

stats = GPUStatistics()
benchmark = stats.benchmark(n_simulations=500000)

print(f"CPU: {benchmark['cpu']['rate']:,.0f} sims/s")
print(f"GPU: {benchmark['gpu']['rate']:,.0f} sims/s")
print(f"Speedup: {benchmark['speedup']:.1f}x")

Performance

Benchmarks on Intel Xeon E5-2676 v3 + NVIDIA GTX 1660 SUPER:

Operation	CPU	GPU	Speedup
Monte Carlo (500K)	14.4s	0.47s	30x
Rate	37K sims/s	1.1M sims/s	-

Configuration

GPU Memory Management

stats = GPUStatistics(
    batch_size=50000,      # Reduce if out of memory
    memory_pool=True       # Enable memory pooling
)

CuPy Configuration

import cupy as cp

# Check available memory
mempool = cp.get_default_memory_pool()
print(f"Used: {mempool.used_bytes() / 1e9:.2f} GB")

# Clear memory
mempool.free_all_blocks()

Troubleshooting

CuPy Installation Issues

# Verify CUDA version
nvcc --version

# Install matching CuPy version
pip uninstall cupy-cuda12x
pip install cupy-cuda12x

Out of Memory Errors

Reduce batch size in GPU operations:

stats = GPUStatistics(batch_size=25000)

No GPU Detected

The system automatically falls back to CPU. Verify GPU availability:

from src.gpu_stats import GPUStatistics

stats = GPUStatistics()
print(f"GPU available: {stats.gpu_available}")
print(f"Device: {stats.device_name}")

API Reference

GPUStatistics

Method	Description
monte_carlo(n_simulations, batch_size)	Run Monte Carlo simulation
benchmark(n_simulations)	Compare CPU vs GPU performance
get_device_info()	Get GPU device information

WeightOptimizer

Method	Description
optimize(train_ratio, validation_ratio)	Find optimal weights
evaluate(weights, data)	Evaluate weight configuration
get_history()	Get optimization history

License

MIT License

Copyright (c) 2025

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.