Let us discuss what is Machine Learning…
Machine learning is a concept by which machines like computers get new knowledge without directly being programmed on what the next instruction is. Main aim of this is to allow machines learn by themselves without any human intervention. It is closely related with the concept of Artificial Intelligence.
Machine Learning is actually a concept where the machines gather new knowledge using existing data. It is done using machine learning algorithms. These machine learning algorithms build models based on sample data. Then those models are used in making actionable insights. There is no need for the programmers to create new instructions for all the changing situations. Instead of that, machine itself creates new rules or models to match with the changing situations or changing outcomes and improves its’ efficiency.
This process is mostly similar to the learning process by a human brain. Always the human brain learn with examples and experience. In the same way machine learning algorithms learn more and more with the experience they gather. The accuracy of prediction made for any previously unknown area will be less than the accuracy of a well known area by the machine.
Here, this training data is the set of data used to initially train our model. After that we use testing data to test our model.
Machine Learning algorithms are categorized as follows.
- Supervised learning algorithms
- Unsupervised learning algorithms
- Reinforcement learning algorithms
Supervised learning algorithms
In supervised learning, there is an existing set of labeled data. This is much similar to the learning process carried out under a supervisor or a teacher. Under supervision, training data is accompanied by labels indicating the class of the observations. Here, class labels of each training data tuple is provided. After that new data is classified based on this training set. This is a high accurate method.
Supervised learning allows to solve many real world problems with previous experience and data. But there can occur problems in classifying those big data very efficiently. That means this may take a lot of computational time. Below are some supervised learning types.
Unsupervised learning algorithms
Under this model, there is no need of any supervision. And also class labels of the training tuples are not previously known. That means algorithms are used against unlabeled data. This can be used to cluster data based on their statistical properties. Unsupervised learning is complex with compared to supervised learning. This method is less accurate with compared to supervised learning. This unsupervised learning can find all types of unknown patterns in data. Below are some of the unsupervised learning types.
- Anomaly detection
- Neural networks
Reinforcement learning algorithms
Reinforcement learning is a method of machine learning concerned with how software agents learn to take actions in an environment to maximize the reward. Under this, the agent learns with an interactive environment by trial and error method. During each trail it tries to maximize its’ reward.
Under this model there is no any supervisor and has only a reward signal or a real number. It mainly interacts with the environment while working. This algorithm makes decisions sequentially. When using this method, the parameters may sometimes affect the speed of learning. Below are some types of reinforcement learning.