A Simple Perceptron in C

This repository contains a simple perceptron implementation in C for binary classification. Created for a uni class and scored 100/100.

Overview

The perceptron implementation includes:

• Data loading from CSV files
• Training with configurable learning rate and iteration count
• Testing and evaluation

Requirements

• C compiler (gcc recommended)
• CSV data files with the following format:
  • First column: x1 (first feature)
  • Second column: x2 (second feature)
  • Third column: label (1 or -1)
  • First row should contain headers

Usage

Compilation

Compile the program using:

gcc -Wall -o perceptron perceptron.c

Running the Program

Run the compiled program with the following arguments:

./perceptron <learning_rate> <max_iterations> [debug_prints]

Parameters:

• learning_rate: Step size for weight updates (e.g., 0.01)
• max_iterations: Maximum number of training iterations
• debug_prints: Optional (0 or 1) to enable verbose output

Example:

./perceptron 0.01 1000 0

Input Data

The program expects two CSV files in the same directory:

• training_data.csv: For training the perceptron
• test_data.csv: For evaluating the trained model

Implementation Details

Data Structures

• Data: Stores input features and labels
• Perceptron: Contains weights and bias for the model

Key Functions

• init_perceptron(): Initializes the perceptron with random weights
• train_perceptron(): Implements the perceptron learning algorithm
• predict(): Makes predictions using the trained model
• eval_predictions(): Calculates accuracy on test data

Output

The program outputs:

• Training information (accuracy and iterations)
• Final model parameters (weights and bias)
• Test accuracy

Memory Management

The program handles memory allocation and cleanup automatically to prevent leaks.

Example Output

=== Starting new perceptron training run ===
--- ---

Train data successfully read.
Rows: 100
---

Training phase completed
Accuracy=87.0000 
Total Training Iterations=15

Learning Rate: 0.000500
Max Iterations: 15
Weights: 0.0790898331, 0.1935892719
Bias: 0.3837106444
---

Test data successfully read.
Rows: 100
-----

Test phase ended
Prediction Accuracy = 87.0000

--- ---