OOP-Brain: Rust GUI & ANN Edition

OOP-Brain is a robust, from-scratch feedforward neural network and matrix computation engine engineered in memory-safe Rust with an integrated graphical user interface (eframe/egui).

Features

• From-Scratch Architecture: Implements matrix operations, linear algebra transformations, and backpropagation without relying on heavy deep-learning frameworks.
• Multi-Class Classification: Supports ReLU, Sigmoid, and Softmax activations paired with Categorical Cross-Entropy and MSE loss functions.
• Tabular Data Processing: Built-in CSV dataset parser with automated dataset splitting and shuffling mechanisms.
• Interactive GUI Dashboard: Real-time training parameter tuning (Epochs, Learning Rate, Batch Size), live loss plotting via egui_plot, and inference tools.
• Model Persistence: Save and load trained synaptic weights and biases securely.

Project Structure

• matrix.rs: Heap-allocated safe matrix math engine supporting transposition, broadcasting, and slicing.
• activations.rs: Mathematical activation functions and numerical stability routines.
• traits.rs: Polymorphic core abstractions for Layer and Activation.
• network.rs: Sequential pipeline execution and gradient descent optimization.
• dense_layer.rs: Fully connected linear layers with thread-safe interior mutability (Mutex).
• loss.rs: Evaluation metrics, loss derivatives, and classification accuracy calculations.
• dataset.rs: Data preprocessing, one-hot encoding, and train/test partitioning.
• io.rs: Safe file stream management for CSV ingestion and model weight exports.
• main.rs: Immediate-mode desktop graphical user interface.

Getting Started

Ensure you have Rust and Cargo installed locally. Clone the repository and run the application via Cargo:

cargo run

[dependencies]
eframe = "0.24"
rand = "0.8"
winapi = { version = "0.3.9", features = ["winuser", "windef"] }
egui_plot = "0.24"