This diagram gives an overview of the machine learning process which consists of the training phase and the inference phase. The training phase takes in data and builds a model by rapidly trying many different parameters. This data could be from user-generated data or from data downloaded online. In order to generate an optimal model, this process is repeated thousands or even millions of times to find the optimal weights and biases that give the best results. Once the model is finalized, then it moves to the inference phase where new unseen data is run through the model and it generates a prediction. Image-based identification is a good example. Assuming that the designer wants to identify cat pictures in the database, the designer can select some cat pictures from the database and train this model many times with many different types of cat photos so that the model can identify the cat pictures with a high confidence. Then in the inference phase, if the designer has a camera input, the model will be able to identify when it is looking at a photo of a cat. The NXP eIQ™ platform is what allows the designer to do this inference phase on NXP microcontrollers and microprocessors.