Supperised Learning uses labeled training datasets to understand the relationship between input and output data. The data scientist manually creates a training dataset containing the input data along with the corresponding labels. Monitored learning trains the model and applies the correct output to the new input data in the actual use case.
During training, the model's algorithm processes large datasets to investigate potential correlations between inputs and outputs. The performance of the model is then evaluated on the test data to see if it was trained successfully. Cross-validation is the process of testing a model using different parts of a dataset.
A gradient descent family of algorithms that include stochastic gradient descent (SGD) is the most commonly used optimization or learning algorithms when training neural networks and other machine learning models. Model optimization algorithms evaluate accuracy via loss functions, i.e. equations that measure the inconsistency between the model's prediction and actual values.
The gradient, or slope, of the loss function, is the main metric of model performance. The optimization algorithm descends the slope to minimize its value. Throughout training, the optimization algorithm updates the model's parameters (operational rules or “settings”) to optimize the model.
