How t-thirupathi/neural-network-implementation works
This is a from-scratch, small neural-network experiment: running the script creates and normalizes sample data, splits it, trains a model, and plots training and prediction results. The model uses activation functions, backpropagation, an optimizer, and a stopping criterion. The README also describes experimenting with alternative activations and optimizers. The graph reflects the shown script; these modules are tightly coupled in one runtime rather than separate services.
Main entry points: neural_network.py Neural solver · activation.py Activations · optimizer.py Optimizers
Components 7
Data and results
- Prepare data
neural_network.py - Evaluate and plot
neural_network.py
Model training
- Neural solver
neural_network.py - Stop criteria
neural_network.py
Training methods
- Activations
activation.py - Optimizers
optimizer.py
Other components
- Experimenter
Connections 7
- Experimenter → Prepare data: runs script
README - Prepare data → Neural solver: trains model
neural_network.py - Neural solver → Activations: uses forward/backward
neural_network.py - Neural solver → Optimizers: requests updates
neural_network.py - Neural solver → Stop criteria: checks stopping
neural_network.py - Prepare data → Evaluate and plot: supplies evaluation data
neural_network.py - Neural solver → Evaluate and plot: returns loss history
neural_network.py