Neural Network Classifier (N2N) 🐾✨

This project implements a two-layer artificial neural network (ANN) for binary classification using NumPy, with dazzling training visualizations powered by Matplotlib and Seaborn. Whether you prefer static 📊 or live 🎥 updates, this code tracks loss, accuracy, and even throws in a confusion matrix to show off its skills. The dataset—featuring adorable cats 🐱 and dogs 🐶—is loaded from HDF5 files via a handy function in utilities.py.

The neural network rocks:

An Input Layer (sized by the dataset features)
A Hidden Layer (you pick the neuron count! 🧠)
An Output Layer (cat or dog? Binary magic! 🎯)

Features 🌟

Forward & backward propagation for training 🚀
Log-loss cost function 📉
Gradient descent optimization ⚙️
Accuracy tracking + confusion matrix fun 🎨
Loss & accuracy visuals over epochs (static or live!) 🌈
Dataset preprocessing (reshaping & normalizing) 🔧

Prerequisites ✅

You’ll need Python 3.6+ installed 🐍. Check requirements.txt for the full list of goodies.

Installation 🛠️

Clone or grab this repo:

git clone https://github.com/Adamo08/TwoLayer-PropNet.git
cd TwoLayer-PropNet

Install the dependencies:
```
pip install -r requirements.txt
```
Make sure datasets/ has trainset.hdf5 and testset.hdf5 ready to roll! 🐾

Requirements 📦

Here’s what’s in requirements.txt:

numpy
matplotlib
seaborn
scikit-learn
tqdm
h5py

Get them with:

pip install matplotlib seaborn scikit-learn tqdm numpy h5py

File Structure 📂

├── datasets/
│   ├── trainset.hdf5    # Training dataset (cat & dog pix! 🐱🐶)
│   └── testset.hdf5     # Test dataset (more furry friends!)
├── utilities.py         # Helper functions (e.g., load_data) 🛠️
├── app.py               # Main script with ANN awesomeness 🌟
├── requirements.txt     # Required Python modules 📋
└── README.md            # You’re here! 👋

Dataset 🐱🐶

The dataset is stored in HDF5 format and packed with images of cats and dogs for binary classification (cat = 0, dog = 1, or vice versa—your call!). It includes:

Training Set: X_train (images) & y_train (labels)
Test Set: X_test (images) & y_test (labels)

Loaded via load_data() in utilities.py. 😻

Usage 🚀

Check that all files are in place.
Tweak hyperparameters in app.py if you’re feeling fancy:
- n1: Hidden layer neurons (default: 3) 🧠
- alpha: Learning rate (default: 0.01) ⚡
- epochs: Training rounds (default: 100) ⏳
Fire it up:
```
python app.py
```

Training Options 🎛️

N2N_NORMAL: Trains and shows static plots—loss, accuracy, and a confusion matrix—when done. 📈
N2N_LIVE: Trains with live-updating loss & accuracy plots, plus a confusion matrix at the end. 🎬

Switch it up in app.py:

# Static vibes
params = N2N_NORMAL(X_train_reshape, y_train, 3, 0.01, 100)

# Live action
params = N2N_LIVE(X_train_reshape, y_train, 3, 0.01, 100)

Outputs 🎉

Console: Final model accuracy 🏆
Plots:
- Loss & accuracy curves (saved during training) 📉📈
- Confusion matrix (saved as confusion_matrix.png) 🎨
Sample Images: A cute grid of cat & dog pics with labels before training starts! 🖼️

Code Overview 🔍

Key Functions

initialize: Sets up weights & biases ⚖️
Forward_Propagation: Runs data through the network ➡️
log_loss: Measures the cost 📊
Back_Propagation: Calculates gradients ⬅️
update: Tweaks params with gradient descent 🔧
predict: Makes cat-or-dog calls 🐾
Confusion_Matrix: Plots the results 🎯
N2N_NORMAL / N2N_LIVE: Trains with style! 🌟

Preprocessing 🛠️

Reshaping: Flattens images (e.g., (samples, height, width) → (samples, features)) 📏
Normalization: Scales pixels from [0, 255] to [0, 1] 🌈

Example 🌟

Say your dataset has 1000 cat & dog pics (64x64 grayscale):

X_train shape: (1000, 64, 64)
After reshaping: (1000, 4096)
After transposition: (4096, 1000)
y_train shape: (1, 1000)

Run:

python app.py

Trains with 3 hidden neurons, a 0.01 learning rate, and 100 epochs. 🐱🐶

Notes 📝

Built for binary classification (cat vs. dog, n2 = 1) 🐾
For bigger datasets, bump up n1, alpha, or epochs 🔧
Live plotting (N2N_LIVE) might lag a bit due to real-time updates ⏱️

License 📜

Unlicensed—free to use or tweak for fun & learning! 🎓

Acknowledgments 🙌

Inspired by classic neural network ideas and powered by Python’s awesome data science tools. 😎