Diverse Forms of Data for Training Machine Learning and Deep Learning Models

Welcome to the 23rd episode of my Engineering Exploration series. In this article, we will look in detail at various forms of data that can be used to train Machine Learning and Deep Learning models.

When it comes to data, it refers to any information that can be captured, measured, or observed — yes, anything, such as stock market trends, images, sound, or speed. But a critical question arises: Can you directly use the captured data in your Machine Learning and Deep Learning models? The answer is NO. These models require the data to be in a specific format. For example, if you are training a Deep Learning model with an image dataset, the model expects all of your images to be of the same size. Although you might collect images from the internet or other sources that vary in shape and format, as a DL/ML engineer, you must ensure consistent input to the model. The process of resizing images to a uniform size is known as image preprocessing.

While the division of datasets into smaller subsets (like training, testing, and validation sets) is important, it falls outside the scope of this article. Our primary focus here is to comprehensively explore, with examples, the various forms of data that can be either used directly or transformed into a format suitable for machine learning and deep learning techniques.

It’s important to note that in the realm of Machine Learning, the terms ‘type of data’ and ‘data type’ do not refer to the same concept, though they are related. “Types of Data” refers to the nature or category of the data used in a model, while “Data Type” pertains to the specific format of kind of data representation in a programming or mathematical context.

We can classify various types of data based on different criteria. Below are the most common types of data based on various criteria.

This article has been moved here.