How alexxdceni/ml works
This repository is a from-scratch C++ neural-network framework with a task-selecting executable. The main workflow selects a task, loads a model and training or evaluation data, then trains or predicts; models can be saved, and MNIST and interactive point/function experiments are additional capabilities. The graph separates task orchestration, network computation and persistence, data utilities, and experiments. Edges reflect calls or references shown in excerpts, the source index, or the README; several internals are unsampled, so their relationships are omitted.
Main entry points: main.cpp Task runner · neural_network.hpp Network engine · mnist_util.hpp MNIST utilities · nn_signal_handler_util.hpp Interrupt autosave
Components 13
Application workflow
- Task runner
src/main.cpp - Interrupt autosave
src/include/utils/nn_signal_handler_util.hpp
Neural network
- Network engine
src/include/neural_network.hpp - Saved models
models - JSON support
src/include/external/json.hpp - Earlier network version
src/include/utils/v0-neural-network.hpp
Data handling
- MNIST utilities
src/include/utils/mnist_util.hpp - MNIST files
data/mnist - Input and targets
data/io
Experiments and conversion
- Image conversion
src/include/utils/file_convertor.hpp - Image decoder
src/include/external/stb_image.h - Point and function tasks
src/include/utils/functions_util.hpp
Other components
- Task selector
Connections 14
- Task selector → Task runner: selects task
src/main.cpp - Task runner → Network engine: trains or predicts
src/main.cpp - Task runner → Interrupt autosave: registers handler
src/main.cpp - Task runner → MNIST utilities: loads or tests MNIST
src/main.cpp - Task runner → Point and function tasks: runs experiments
src/main.cpp - Task runner → Input and targets: loads task data
src/main.cpp - Interrupt autosave → Saved models: saves on interrupt
src/include/utils/nn_signal_handler_util.hpp - Network engine → Saved models: loads and saves
src/include/neural_network.hpp - Network engine → JSON support: uses JSON configuration
src/include/neural_network.hpp - MNIST utilities → Network engine: fills training data
src/include/utils/mnist_util.hpp - MNIST utilities → MNIST files: reads dataset
src/include/utils/mnist_util.hpp - MNIST utilities → Image conversion: uses image utilities
src/include/utils/mnist_util.hpp - Image conversion → Image decoder: uses image decoder
src/include/utils/file_convertor.hpp - Point and function tasks → Network engine: trains and predicts
src/include/utils/functions_util.hpp