srinadhu / convolutional_nn

Implemented fully-connected DNN of arbitrary depth with Batch Norm and Dropout, three-layer ConvNet with Spatial Batch Norm in NumPy. The update rules used for training are SGD, SGD+Momentum, RMSProp and Adam. Implemented three block ResNet in PyTorch, with 10 epochs of training achieves 73.60% accuracy on test set.

http://cs231n.github.io/assignments2017/assignment2/

cs231n-assignment fully-connected-deep-neural-network batch-normalization dropout convolutional-neural-networks pytorch-cnn-classification sgd-optimizer sgd-moment rmsprop adam-optimizer cifar-10 pytorch

Details about this assignment can be found here.

About

Implemented fully-connected DNN of arbitrary depth with Batch Norm and Dropout, three-layer ConvNet with Spatial Batch Norm in NumPy. The update rules used for training are SGD, SGD+Momentum, RMSProp and Adam. Implemented three block ResNet in PyTorch, with 10 epochs of training achieves 73.60% accuracy on test set.

http://cs231n.github.io/assignments2017/assignment2/

cs231n-assignment fully-connected-deep-neural-network batch-normalization dropout convolutional-neural-networks pytorch-cnn-classification sgd-optimizer sgd-moment rmsprop adam-optimizer cifar-10 pytorch

Languages

Language:Jupyter Notebook 89.9%Language:Python 10.0%Language:Shell 0.0%