Site is loading

Machine Learning, AI and deep learning seem separate in many respects but are inextricably linked. It is namely a subset of each other. In the first part of this five-part, examples are explained about Machine learning, AI and deep learning.

Let’s start with the basics: what are machine learning, AI and deep learning, and how are they used in practice? If you have read this article, you have the required knowledge to understand the next part of the series.

Machine learning, deep learning and AI are related to each other. The three topics have to do with each other. They are dependent on each other. The topics are discussed from each other so that you have a better picture of these terms.

With machine learning, you can develop an algorithm through big data that can make predictions. Machine learning is therefore dependent on data, and with this data, you can predict a particular output from a dataset. However, you must indicate once what is right or wrong within the data. It is best to explain this through a practical example.

For a brand fashion website, a customer wants to classify new products in a category on the site automatically. In this way, fewer FTEs are needed to assign products to a category manually, so a considerable saving for the organization. An algorithm can do this. We had loaded the 40 GB data set, so big data. The data must now be trained. Training data means that the algorithm must learn to which category a product belongs. In this case, the algorithm must know that skinny jeans fall under Jeans-> Skinny jeans. When the data was trained, the machine learning algorithm was ready. The most famous fashion products are now automatically classified in the correct category.

The customer then had a new supplier who delivers their product names in English. Now the algorithm cannot recognize whether the English products fall under the correct category, and the data must be re-trained for English products. This is exactly what machine learning is. The algorithm needs instructions and cannot recognize new patterns. That brings us to the possible solution; deep learning.

Deep learning solves the problem of machine learning. With machine learning, you have to give algorithm instructions to classify an item in a category. In the example above at the Fashion website, you saw that the machine learning algorithm could not deal with new (English) data. However, deep learning can do this by recognizing specific patterns. If we change our algorithm to a deep learning algorithm, then the algorithm is smart enough to know that a “dress” and a “dress” are precisely the same.

AI is simply that the computer can make intelligent decisions independently. With machine learning and deep learning, you partly have AI, but not entirely. Scientists have not yet succeeded in implementing the same capacities as the human brain in an algorithm. The algorithms are based on calculus, statistics and linear algebra. You often hear that deep learning and machine learning are AI, but that is actually not true. For our fashion website, a deep learning model is best for classifying categories because we cannot yet make the human brain digital.

An algorithm is developed in practice with Python. It is recommended to write machine learning applications in Python because the libraries for machine learning are based on this language. As most people think, you don’t have to write the algorithms yourself. Within Python, there is a library called sci-kit learn. Here you can call various algorithms. It is essential to know which algorithm to use.

When you get started with machine learning, you quickly discover that you have to visualize data in a coordinate system. In this way, you know what the correlations are between specific data units. Visualizing data can be done with Python through Jupyter Notebook. This is a tool to visualize data. When you have visualized the data, you know approximately which algorithm you can use. In the example below, data is plotted in a coordinate system:

You see that the information correlates with each other. Based on this, you can use a linear regression algorithm to make predictions. If someone has been a member for five years, you can find out how much they spend on average.

With this, the differences between machine learning, deep learning and AI are discussed. You have read that these topics are related to each other. AI is still in the future as scientists are busy digitizing the human brain. In practice, deep learning is used, such as self-driving cars. It depends on you whether you interpret this as AI or deep learning.

The following article examines the algorithms in more detail and follows linear regression and classification problems. Next, several practical examples will follow on how this can be implemented.