Digit-Recognizer-MNIST-Data

• link of the competition - https://www.kaggle.com/c/digit-recognizer
• Used deep learning techniques to train a digit classifier on popular MNIST Data
• Developed a ConvNet with 6 Convolutional Layer and 3 Fully Connected Layer
• Used ReLU, softmax and sigmoid for activation
• Used SGD and Adadelta for optimization
• Achieved an accuracy of 98.2% on test data by using adadelta and sigmoid
• Increased the accuracy to 99.39% by using augmented data
• Achieved top 9% rank in Digit Recognizer Competition by Kaggle