Training the model

Using the training data subset, on the platform, we train ML models to learn the relationships between the target and the features. ML model training is an iterative process: it starts from an assumed model with initial parameters and continues the learning process until it fits the training dataset. Figure 4.1 shows a sample ML model training process, where we have selected a linear regression model (z=wx+b) and chosen the initial parameters (w and b). We calculate the model predict-error – the gap between the model output and the actual data label – this step is called forward propagation. If the error is not optimized (the accuracy is not within the specified range), we will need to move back and adjust the model’s parameters (w and b) – this step is called backward propagation. We will then go forward to recalculate the error again. This model training process repeats the steps of forward propagation, backward propagation, and forward...