Deep Neural Network in C with 2 layers

Class: COMP 605: Scientific Computing

Zack Humphries, Anuradha Agarwal, Thomas Keller

1. Files and folder in this directory:

• main.c: serial code for the deep neural network
• evaluation.h: contains all the functions used in the serial code
• /openmp: contains all the files related to openMP
• /datasets: contains all the datasets used in this project
• /comparision: contains the forward propagation code used to compare results
• /CUDA: contains all the files related to CUDA

2. How to run...

2a. Serial Code

There are 3 command line arguments: number of inputs(30 or 4800), learning rate, and number of epochs

• How to compile:

	gcc main.c -o main -lm

• How to run:

	./main <numberofInputs> <learningRate> <numberofEpochs>

2b. OpenMP Code

There is only one command line arguments: number of threads

• How to compile:

	gcc -g -Wall main.c -o main -lm -fompenmp  

• How to run:

	./main <numberOfThreads>

2c. CUDA Code (requires prerequisite Nvidia CUDA Toolkit downloaded)

There are no command line arguments

• How to compile:

	nvcc main.cu -o main -lm

• How to run:

	./main

3. Dataset Used

https://www.kaggle.com/datasets/uciml/breast-cancer-wisconsin-data?resource=download

4. Compared to

https://github.com/nihil21/parallel_nn

5. Report

The file final_project_report.pdf contains an in-depth analysis of the paralleled neural networks and algorithms.

Licence:

MIT