Labeled Samples are the input for machine learning algorithms. An ML algorithm, for example, an artificial neural network, learns by analyzing examples that are pre-categorized by a human editor. Showing examples of apples and melons to the neural network allows it to figure out what distinguishes apples from melons and ultimately enables it to classify them correctly. The input is called “labeled samples” because it isn’t sufficient to provide 100 images of apples and 100 images of melons. Instead, each of the images needs to be manually pre-categorized (labeled), so that the machine can learn. Labeling is often called “annotating”. The manual labelling of millions of input samples is what makes neural networks so cumbersome to train.