Master 2 MIAGE
Méthodes Informatiques Appliquées à la Gestion des Entreprises
UE : FST (Fouille de Données Statistiques)
University of Rennes 1 - France
Author : Pierre Delaunay
Contact : pierre.delaunay@etudiant.univ-rennes1.fr
The best entry (using LeNet v5 described below) scored an accuracy of 99.50% on the test set given by Kaggle.
| Model | Train Accuracy (%) | Validation Accuracy (%) | Final Score (%) |
|---|---|---|---|
| LeNet-5 | 98.09 | 96.48 | 99.50 |
ConvNet --> Pool --> ConvNet --> Pool --> ConvNet --> (Flatten) --> FullyConnected --> Softmax
epochs - 100 loss - 0.0509 train_accuracy - 0.9809 val_loss - 0.0955 val_accuracy - 0.9648
LeCun et al., Gradient-based learning applied to document recognition (1998)