CHEAT SHEET

ConvNet Cheat Sheet

Algorithm Name Convolutional neural networks

Description A special type of neural network that is well suited for image data.

Applicability Any supervised learning problem (classification or regression).

Assumptions The input is a 2D image of pixels (although 1D, 3D, and 4D variants exist).

Underlying Perceptron
Mathematical Convolution layers
Principles
Pooling layers
Batch normalization

Open Source PyTorch

Implementations TensorFlow

Additional Details • Training is usually done using SGD

• Use cross entropy loss for classification and MSE for regression
• Batch normalization is applied to make training easier
• Convolution layers exploit local structures in images

CIS537: Deep Learning

and Neural Networks
1
© 2019 eCornell. All rights reserved. All other copyrights, trademarks,
trade names, and logos are the sole property of their respective owners.
Computing and
Information Science