Hello and welcome to this video on machine learning problems. In this video, we will talk about the different types of machine learning, including supervised, unsupervised and reinforcement learning. On this course, we'll primarily focus on supervised learning, which is important to understand what each of these is. Let's start with supervised learning. In supervised learning, an algorithm learns from labeled data. For example, let's say we're training an image recognition algorithm to distinguish cats from dogs. Each training data point will be a pair of an image, that's the training data and the label, specifying whether it's a cat or a dog. In a similar way, if we are training a translation engine, we would provide both input and output sentences and ask the algorithm to learn the function that connects them. Conversely, in unsupervised learning, data comes without labels and the task is to find similar data points in the data set, in order to identify any underlying higher order structure. 

For example, think of a data set that contains purchased references from e-commerce users. These users would like a form cluster, that behaving in a similar way. They may have similar purchase behavior. They may spend the same amount of money or the similar amount of money or buy similar type of objects. You can think of these different groups as different tribes with different preferences. Once these tribes are identified, we can describe each data point, which is each user, in terms of the tribe, it belongs to. Getting a deeper understanding of the data and also reducing the number of data points, that we have to deal with. Finally, reinforcement learning is similar to supervised learning, but in this case, the algorithm is training an agent to act in an environment. The agent can take some action and know the state of the environment and these actions lead to an outcome. The outcome is attached to a score and the algorithm tries to maximize the score. 

Typical examples are the algorithms, that learn to play games, like chess or Go. And the main difference with supervised learning is that the algorithm does not really see a label for each action, but may receive the score after a sequence of actions. In 2016, a software train with this technique beat the World Go Champion and marked the new milestone in the race to World's Artificial Intelligence. So, in conclusion, we have seen there's many types of machine learning problems and deep learning has been successfully applied to both supervised, unsupervised and reinforcement learning problems. We will focus on supervised learning problems in the rest of the course. Thank you for watching and see you in the next video.