Machine Learning

Introduction

Machine Learning is a great introduction to machine learning and artificial intelligence. It provides useful articles on how this technology can affect your work and life. The non-technical manual of the site on artificial intelligence is also especially useful for those who are interested in mastering machine learning, but who do not have any technical background.

Supervised learning

Supervised learning is the task of inferring a function from the labelled training data. When approaching a marked training set, we want to find the most optimal model parameters for predicting unknown marks on other objects (test set). If the label is a real number, we call the task regression. If a label refers to a limited number of values ​​where these values ​​are disordered, then this is a classification.

Regression

Similarly, in the case of supervised learning, you give specific examples known to the computer. You say that for the given characteristic value x1 the output is y1, for x2 it is y2, for x3 it is y3, and so on. Based on these data, you let the computer discover an empirical relationship between x and y.

Once the machine is trained in this way with a sufficient number of data points, it would now ask the machine to predict Y for a given X. Assuming that you know the actual value of Y for this given X, you will be able to deduce whether the machine’s prediction is correct.

Therefore, it will test whether the machine has learned using the known test data. Once you are satisfied that the machine can make the predictions with a desired level of accuracy (say 80 to 90%), you can stop training the machine any further.

Now, you can safely use the machine to make predictions at unknown data points, or ask the machine to predict Y for a given X for which you do not know the actual value of Y. This training comes under the regression of the that we talk about earlier.

Classification

You can also use machine learning techniques for classification problems. In classification problems, classify objects of a similar nature into a single group. For example, in a set of 100 students, you can group them into three groups based on their height: short, medium, and long. By measuring the height of each student, you will place them in a suitable group.

Now when a new student comes in, they will place him in an appropriate group by measuring his height. By following the principles of regression training, you will train the machine to classify a student according to their characteristic: height. When the machine learns how groups are formed, it will be able to classify any new unknown student correctly. Again, you would use the test data to verify that the machine has learned its classification technique before putting the developed model into production.

Supervised learning is where AI really started its journey. This technique was successfully applied in several cases. You used this model while doing handwritten recognition on your machine. Various algorithms for supervised learning have been developed.

Unsupervised learning

With uncontrolled training, we have less information about objects, in particular, the train is not marked. What is our goal now? You can observe some similarities between groups of objects and include them in the corresponding clusters. Some objects can be very different from all clusters, so we assume that these objects are anomalies.

Reinforcement learning

Enhanced learning is not like any of our previous tasks, because we don’t tag or tag datasets here. RL is a machine learning area related to how software agents need to take action in some environments to maximize some understanding of the total reward.

Consider training a pet dog, we train our pet to bring us a ball. We throw the ball some distance away and ask the dog to return it to us. Every time the dog does this well, we reward the dog. Gradually, the dog learns that doing the job correctly gives him a reward, and then the dog begins to do the job correctly every time in the future. Exactly, this concept is applied in the “reinforcement” type of learning. The technique was initially developed for machines to play games. The machine receives an algorithm to analyze all possible movements at each stage of the game. The machine can select one of the movements at random. If the movement is correct, the machine is rewarded, otherwise it can be penalized. Little by little, the machine will begin to differentiate between correct and incorrect movements and after several iterations it will learn to solve the game puzzle with greater precision. The accuracy of winning the game would improve as the machine plays more and more games.

Deep learning

Deep learning is a model based on Artificial Neural Networks (ANN), more specifically Convolutionary Neural Networks (CNN). There are several architectures used in deep learning, such as deep neural networks, deep belief networks, recurrent neural networks, and convolutional neural networks.

These networks have been successfully applied to solve problems of computer vision, voice recognition, natural language processing, bioinformatics, drug design, medical image analysis, and gaming. There are several other fields in which deep learning is applied proactively. Deep learning requires huge processing power and huge data, which are generally readily available these days.

Deep reinforcement learning

Deep Reinforcement Learning (DRL) combines reinforcement and deep learning techniques. Booster learning algorithms like Q-learning now combine with deep learning to create a powerful DRL model. The technique has been highly successful in the fields of robotics, video games, finance, and healthcare. Many problems that previously could not be solved are now solved by creating DRL models. There is a lot of research going on in this area and this is very active by industries.

So far, you have a brief introduction to various machine learning models, now let us explore a little more deeply into various algorithms that are available under these models.