Convolutional Neural Networks (CNN)

A class of deep learning best suited for analyzing visual imagery. They are heavily influenced by how humans see the world. We analzye images in layers of increasing complexity. The first layer distinguishes basic attributes like lines and curves.

Feature detection is based on scanning the input with the filter of a given size and applying matrix computations in order to derive a feature map

Activation function returns 0 for every negative value in the input image while it returns the same value for every positive value

Pooling provides spatial variance which means that the system will be capable of recognizing an object as an object even when its appearance varies in some way. It will perform an down sampling operation along the spatial dimensions (width, height), a pooling_size of (2,2) will halve the width and height

Flatten the output of the last convolution layer and connect every node of the current layer with the other node of the next layer. Neurons in a fully connected layer have full connections to all activations in the previous layer.

Validation accuracy is the accuracy on the validation dataset where as the accuracy score is the correct score that includes the training data

Learning rate is a hyperparameter that controls how much to change the model in response to the estimated error each time the model weights are updated. A value too small may result in a long training process that could get stuck whereas a value too large may result in a sub-optimal set of weights

Often in real life you would have skewed data to fix this you can use undersampling (deleting records), oversampling (adding records) or generate synthetic records.

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/29aa1aa8-a5f2-48e0-b146-de3589d0343c/Untitled.png

Gatsby