classificatore_immagini/AGENTS.md

# AGENTS.md - Agentic Coding Guidelines

## Project Overview
C-based neural network implementation from scratch for image classification on CIFAR-10 and MNIST datasets.

## Build Commands

### Compilation
```bash
# Main classifier (MNIST) - compiles the single-classifier training program
gcc -o classificatore_singolo_mnist classificatore_singolo.c -lm

# XOR test/demo - simple test to verify neural network works
gcc -o codice_ricordo codice_ricordo.c -lm

# Visualizer (requires Allegro library)
gcc -o visualizzatore visualizzatore.c -lalleg -lm

# Older multi-layer implementation
gcc -o rete_neurale rete_neurale.c -lm
```

### Running Tests
```bash
# Run XOR validation test (quick sanity check for neural network)
./codice_ricordo

# Run full training loop with existing compiled binary
./classificatore_singolo_mnist

# Automated training with pauses (prevents overheating)
./addestratore.sh

# Memory leak detection (valgrind)
valgrind --leak-check=full ./codice_ricordo
```

### Running a Single Test
```bash
# Compile and run XOR test (single test case)
gcc -o codice_ricordo codice_ricordo.c -lm && ./codice_ricordo

# Compile and run classifier with specific category
gcc -o classificatore_singolo_mnist classificatore_singolo.c -lm && ./classificatore_singolo_mnist
```

## Code Style Guidelines

### Language
- Use **Italian** for comments and variable names (maintain consistency)
- Use **English** for struct/type definitions and technical terms

### Formatting
- Indent with 4 spaces (no tabs)
- Opening braces on same line: `if (cond) {`
- Always use braces for control structures, even single lines
- Line length: ~80-100 characters preferred
- One space after keywords (if, for, while, return)
- No space between function name and opening parenthesis
- Blank line between functions

### Naming Conventions
- Functions: `snake_case` (e.g., `inizializza_rete_neurale`)
- Structs: `PascalCase` (e.g., `ReteNeurale`)
- Constants: `UPPER_SNAKE_CASE` (e.g., `EPOCHE`)
- Global variables: file scope preferred
- Types: use `typedef` for structs

### Types
- Use `byte` (typedef for `unsigned char`) for values 0-255
- Use `double` for all floating-point calculations
- Prefer explicit types over implicit conversions
- Use `size_t` for sizes and indices where appropriate

### Imports
- Standard library headers first (`<stdio.h>`, `<stdlib.h>`, `<math.h>`, `<time.h>`)
- Project headers after (use `"quotes"`)
- No include guards needed for header-only library
- Group related includes together

### Memory Management
- Always check `malloc` return values
- Free memory in reverse allocation order
- Use `perror()` for error messages before exiting
- Avoid memory leaks in loops

### Error Handling
- Check file operations with `if (!file)` pattern
- Return `NULL` on failure for functions returning pointers
- Exit with `EXIT_FAILURE` on unrecoverable errors
- Validate function inputs at entry points

### Key Constants (from percettroni.h)
- `LRE = 0.1` (learning rate - was 1.414, now 0.1)
- `soglia_sigmoide = 0.5` (sigmoid threshold for binary classification)
- `file_pesi = "rete_pesi.bin"` (model weights file)

### Dataset Configuration
In `percettroni.h`, uncomment the desired dataset section:
- XOR: `#include "xor_manager.h"` (for testing - currently active)
- MNIST: Uncomment mnist includes and set file paths
- CIFAR-10: Uncomment cifar-10 includes and file paths

## Testing
No formal test framework. Use these approaches:
1. `codice_ricordo.c` - XOR validation (4 inputs, quick convergence test)
2. Visual inspection of weight outputs via `stampa_pesi_rete()`
3. Monitor epoch error rates in training output
4. Check for memory leaks with valgrind

## Project Structure
- `percettroni.h` - Core neural network (header-only library with implementations)
- `classificatore_singolo.c` - Single-category classifier main program
- `codice_ricordo.c` - XOR test/demo
- `xor_manager.h` - XOR dataset for testing
- `mnist/mnist_manager.h` - MNIST dataset loader
- `cifar-10/cifar10_manager.h` - CIFAR-10 dataset loader
- `rete_pesi.bin` - Saved model weights
- `addestratore.sh` - Training automation script

## Neural Network Architecture
- Activation: sigmoid function
- Training: backpropagation with gradient descent
- Configurable: layer count and perceptrons per layer
- Learning rate: controlled via `LRE` constant (default 0.1)
- Binary threshold: 0.5 for classification decisions

## Development Workflow

### Adding a New Dataset
1. Create manager header (e.g., `custom_manager.h`)
2. Define `N_INPUTS` constant for input size
3. Implement `get_dataset()` returning `Dataset*`
4. Update `percettroni.h` includes and file paths

### Debugging Tips
- Call `stampa_pesi_rete(rete)` to inspect weights
- Reference `codice_ricordo.c` for minimal working example
- Verify dataset loading before training loop
- Check epoch timing to monitor training progress

### File I/O
- Weights saved as binary in `rete_pesi.bin`
- Use `salvaReteNeurale()` and `caricaReteNeurale()` for persistence
- Dataset files must match expected binary format

## Performance Notes
- Training is CPU-intensive (minutes per epoch expected)
- Use `addestratore.sh` with sleep intervals to prevent overheating
- Memory allocated dynamically based on network architecture
- No GPU acceleration - pure CPU implementation

## Language Reference
Technical terms (English): activation function, gradient descent, sigmoid, neural network, backpropagation

Italian terms: pesi (weights), bias, livello (layer), percettrone (perceptron), addestramento (training), epoca (epoch), errore (error), classificazione (label/classification), previsione (prediction), istanza (instance)