One-Hot Encodings for Neural Network Inputs - Deep Learning Dictionary
One-hot Encoded Labels - Deep Learning Dictionary
During the supervised learning task of training a network to classify categorical data, we pass the samples and their corresponding labels to the network. The network uses numerically encoded values of the labels to calculate the loss between its predicted labels for the samples and the true labels.
If the dataset we're training on has no order, which is often times the case for many image classification tasks, then this data is considered nominal.
Committed by on