How robinramdin13/digit_recognizer works
This is a small MNIST digit-recognition script, not a multi-service application. Its documented workflow loads Kaggle CSV data, prepares and augments training examples, trains a one-hidden-layer neural network, and evaluates accuracy. The README also describes saved sample plots and comparisons across noise levels. Source excerpts are partial, so the graph omits uncertain wiring for test prediction and output generation.
Main entry points: main.py Script entry point · plots Saved sample plots
Components 13
Data workflow
- Script entry point
main.py - Load and split
main.py - Noise augmentation
main.py - Pixel normalization
main.py - Accuracy evaluation
main.py
Neural network
- Gradient descent
main.py - Forward propagation
main.py - Backpropagation
main.py - Parameter updates
main.py
Visual outputs
- Sample plotting
main.py - Saved sample plots
plots
Other components
- Project user
- MNIST CSV data
Connections 12
- Project user → Script entry point: runs
- Script entry point → Load and split: calls
- Load and split → MNIST CSV data: reads
- Load and split → Noise augmentation: passes training data
- Noise augmentation → Pixel normalization: returns augmented data
- Pixel normalization → Gradient descent: passes inputs
- Gradient descent → Forward propagation: calls
- Gradient descent → Backpropagation: calls
- Gradient descent → Parameter updates: calls
- Script entry point → Accuracy evaluation: evaluates
- Script entry point → Sample plotting: plots samples
- Sample plotting → Saved sample plots: saves