Neural Network Multiplication Examples

Working C++ Implementations

1. CPU Version: neural_network_multiplication.cpp ⭐

• Complete implementation with enhanced features
• Features:
  • Xavier/Glorot weight initialization for better learning
  • Input normalization [0,1] improves training stability
  • Learning rate scheduling for optimal convergence
  • Comprehensive evaluation and visualization
• How to use:

  g++ -O3 -o neural_network neural_network_multiplication.cpp
  ./neural_network

2. GPU Version: neural_network_cuda.cu

• Working CUDA implementation (compiles and runs successfully)
• GPU-accelerated neural network with parallel computation
• Avoids header conflicts with custom math functions
• How to use:

  nvcc -O2 -o neural_network_cuda neural_network_cuda.cu -lcudart
  ./neural_network_cuda

Build System

CPU Build: Makefile

make                    # Build CPU version
make run               # Build and run CPU version

GPU Build: Makefile_cuda

make -f Makefile_cuda neural_network_multiplication    # Build CPU
make -f Makefile_cuda neural_network_multiplication_cuda  # Build GPU (if headers work)
make -f Makefile_cuda run-gpu                         # Run GPU version

Benchmark Tool

benchmark_cpu_gpu.cpp

• Performance comparison between CPU and GPU implementations
• How to use:

  g++ -O3 -o benchmark_cpu_gpu benchmark_cpu_gpu.cpp
  ./benchmark_cpu_gpu

Core Files Summary

File	Type	Status	Purpose
neural_network_multiplication.cpp	C++	✅ Working	CPU implementation with enhanced features
neural_network_cuda.cu	CUDA	✅ Working	GPU implementation
benchmark_cpu_gpu.cpp	C++	✅ Working	Performance comparison
Makefile	Build	✅ Working	CPU build system
Makefile_cuda	Build	✅ Working	GPU build system

Quick Start

Test CPU Learning:

g++ -O3 neural_network_multiplication.cpp -o neural_network
./neural_network

Test GPU Acceleration:

nvcc -O2 neural_network_cuda.cu -lcudart -o neural_network_cuda
./neural_network_cuda

Requirements

• CPU: GCC/G++ compiler
• GPU: NVIDIA GPU with CUDA toolkit

  # Install CUDA (Ubuntu/Debian)
  sudo apt install nvidia-cuda-toolkit nvidia-cuda-dev