Data processing

In the DL life cycle, the data (inputs and outputs) serves important functions such as defining the goal of the problem, training the algorithm, evaluating the performance of the trained model, and building baselines for model monitoring and so this is considered as the most important phase of the DL life cycle. The data processing phase can be subdivided into data collection and data wrangling, which in turn divides into data processing and feature engineering, as depicted here:

Figure 9.3 – Different subphases of data preprocessing

As shown here, the data collection phase mainly includes identifying data resources and the accessibility of data. The data wrangling phase includes data preprocessing and feature engineering. Let’s discuss each of the phases in detail in the following section.

Data collection

Data collection is technically the first step of the DL life cycle. Without data, there is no model. Data collection...