By Nicholas Brown
A machine learning model is an algorithm built to establish relationships between different data points so it can then be trained to perform useful activities such as object detection, language translation, conversational apps that assist customers, computer vision systems for self driving cars, data analysis, and more.
Machine learning models are trained by feeding them data that will help them to recognize and establish relationships on their own.
Common examples of machine learning models:
- Object Detection Models: An object detection model can be trained to detect cats by supplying it with many pictures of cats and a label indicating that they are cats (this is called a dataset). After processing those images, the machine learning model will then be able to detect cats that were not included in that original dataset, because they share physical features found in the other cats. This is the software equivalent of training a baby by saying: ‘This is a picture of a cat’, and doing it thousands of times with many different cat pictures so that it becomes good at recognizing cats. One such model is ‘coco-ssd’.
- Natural Language Processing: You can feed common phrases to a natural language processing model (train it) and tell it what each of them is until it can detect variants of those phrases on its own. To be specific: The point is for it to detect variants that you did not enter. Like the object detection model mentioned above, the objective here is to provide it with enough phrases with the same meaning so that it can detect the countless other ways in which the phrase might be worded.
Machine learning is a subset of artificial intelligence and provides the foundation of modern AI projects.