Machine learning is a subset of the Artificial Intelligence (AI) field used in which we can build a model or algorithm for specific purposes based on given data using special techniques such as programs and statistical computation.
The main types of ML include:
- Supervised learning;
- Unsupervised learning;
- Reinforcement learning; and
- Semi-supervised learning.
1- Unsupervised Learning
This type of ML contains data without labels.
Training set = {(X1), (X2), (X3), (Xn)}.
The main aim is to understand the structure of the data given by algorithms. Types of unsupervised learning include clustering and association.
2- Reinforcement Learning
In this type of ML, the main goal is to maximize the reward action when an agent is performing the right path (increase the behavior); otherwise, a decrease in the behavior is considered to be a result of punishment. The next step of the agent path occurs according to the kind of reward of the previous step if it is positive and continues on the same path; otherwise, it changes the path.
Types of Reinforcement Learning include:
1. Positive Reinforcement; and
2. Negative Reinforcement.
The main factors of RL are:
1. Reward (R): denoting the feedback signal indicating how well the step that the agent makes in a particular time;
2. Action (A): the function of the reward (R) and state (S);
3. State (S): describing the environment;
4. Policy (P): the transfer from the environment to the action being (P);
5. Value Function (V): a measurement tool to indicate how good the step is; and
6. Model (M): the demonstration of the agent’s environment.
3- Semi-Supervised Learning
There are techniques that combine supervised and unsupervised learning to build a model which is able to predict a large number of unlabeled data. Supervised learning uses a small number of label data initially to train the model with a known target to build the model. Unsupervised learning is used by unlabeled data for the same model which is trained prior to predict this kind of data. This operation is named semi-supervised learning.
This technique is used in web mining, text mining, and video mining in which there are huge numbers of unlabeled data and a small number of labeled data.
and difficult. Therefore, by using this method, it will make use of a large number of unlabeled data as well as increase the model accuracy.
4- Supervised Learning
Also called “learning with a master,” this type of ML contains data and labels.
Training set = {(X1, Y1), (X2, Y2), (X3, Y3), (Xn, Yn)}.
The main target is to build a model0 that maps X to the Y label.
Types of supervised learning include classification and regression.
Classification is supervised learning that predicts categories using specific algorithms and labels data. The classification problem is the idea of categorizing data points or instances into a set of labels.
In binary classification, each input image will be classified into one of two classes (such as predicting whether an animal is a dog or a cat or whether mail is spam or non-spam).
On the other hand, in multi-classification, an input image is classified into one of the numbers of classes (such as classifying an MNIST handwritten digit or DR levels).
Some samples of classification include speech recognition, handwriting recognition, biometric identification, image recognition, and so on.
The most commonly used algorithms for image classification are:
- Support Vector Machine;
- Decision Tree;
- Feed-Forward Neural Network;
- Back-propagation network; and
- Deep Learning.
0 Comments:
Post a Comment