Machine learning is the subfield of computer science that, according to Arthur Samuel in 1959, gives “computers the ability to learn without being explicity programmed. Evolved from the study of pattern recognition and computational learning theory in artificial intelligence, machine learning explores the study and construction of algorithms that can learn from and make predictions on data. Such algorithms overcome following strictly static program instructions by making data-driven predictions or decisions, through building a model form sample inputs. Machine learning is employed in a range of computing tasks where designing and programming explicit algorithms with good performance is difficult or unfeasible; example applications include email filtering, detection of network intruders or malicious insiders working towards a data breach, optical character recognition (OCR), learning to rank and computer vision.
Machine learning is studying how the computer to simulate or to realize the study behavior of human being. The aim is to obtain the new knowledge of the skill, organize the knowledge structure, which can make progeressive improvement of it’s own performance. The machine learning research establishes the computation model or the understanding model, according to the study mechanism of humanity through the physicology, the cognitive science, develops each kind of study theory and the study method, studies the general algorithm and carriies on the theoritically analysis, and establishes study system that has the specific application facing the duty.
The learning system basic structure of machine leaning is “environtment -> learning -> knowledge base -> execution” and for the learning is connected with execution. Machine learning is a fine tuning a system with tunable parameters. It has numerous applications and provides solutions to many real-world problems. Some of the applications include:
- Face detection and recognition
Cameras can detect when someone smiles more accurately now better than it used to before because of advances in machine learning. Similiarly because of machine learning, an individual’s photo can be indentified due to a computer program.
2. Visual perception
Analyzing and interpreting visual information surrounding us sums up the visual perception of an individual. This has two more sub-categories:
- Pattern recognition
- Scene analysis
The modelling algorithms used in machine learning help in segregating the piece of informationn received based on the content it has. It is based on training set of data containing observations that leads to classification according to the problem asked for.
4. Adaptive systems
Adapting behaviour based on previous experiences and developing rules according to that, refers to adaptive systems. This includes:
- Cybernetics : communication between automatic control systems.
- Conceptual clustering : models of concept formation that increments and clusters accordiing to that.
To predict the behavior and relationship between real-world objects or entities, set of transformational rules have been written.
- Problem solving systems
- Hobot world modeling (perceptual and functional representations)
6. Speech and image processing
Deep learning, another subcategory of machine learning plays an integral role in speech recognition and image classification and processing. Machine learning also helps in:
- Language and speech understanding
- Semantic information processing
- Retrieval of information
A combination of most or all of the above abilities with the ability to move over terrain and manipulate objects.
- Industrial automation
- AI in household (smart homes)
8. Solving problems
Ability of planning a slution on the basis of formulation of the given problem.
- Interactive problem solving
- Heuristic seacrh
Clustering algorithms or data mining are used in genetic to help finding genes assiciated with a particular disease.
10. Anomaly detection
Insider trading in a stock market can be detected. Fraudulent transaction in high volume business can be tracked because of machine learning.
Translating the rules into a structure that helps in reaching adequate level of performance. Games like chess and bridge.
There are three aims in the Machine Learning:
- General learning algorithm
This direction research is the theoritical analysis duty and the development uses in the non-usable learning duty the algorithm. There is no limit to the algorithm type. The algorithm nt necessarily is similiar the method which uses in the humanity. Some person thought studies the knowledge structure which produces to be supposed to be similiar humanity’s knowledge structure at least, even if the learning process is different. At present, some scientists are researching the possible learning algorithm the theory space.
2. Cognitive model
This direction is a studying human’s learning computation theory and an experimental model. Not only had this kind of research vital significance of humanity education, but also of developing the machine learning system.
3. The goal of the project
This direction is aimed at solving the special actual problem, and developing to accomplish these tasks the project system. Not only do these questions often concern on the learning but also on other questions, for example, input signal by reasonable explanation or development question special-purpose data conversion.
Two of the most widely adopted machine learning methods are supervised learning and unssupervised learning. Most machine larning-about 70 percent-is supervised learning. Unsupervised learning accounts for 10 to 20 percent. Semi-supervised and reinforcement learning are two othe technologies that are sometimes used. Supervised learning algorithms are trained using labeled examples, such as input where are desired output is known. For example, a piece of equipment could have data points labeled either “F” (failed) or “R” (runs). The learning algorithm receives a set of inputs along with the corresponding correct outputs, and the algorithm learns by comparing its actual output with correct outputs to find errors. Unsupervised learning is used against data that has no historical labels. The system is not told the “right answer.” The algorithm must figure out what is being shown. The goal is to explore the data and find some structure within. Unsupervised learning works well on transactional data. For example, it can identify segments of customers with similar attributes who can then be treated similarly in marketing campaigns. The difference between supervised and unsupervised learning is whether you have a labeled training set to work with or not. The labels you apply to data are simply the outcomes you care about. Maybe you care about identifying people in images. Maybe you care about identifying angry or spammy emails, which are all just unstructured blobs of text. Maybe you’re looking at time series data — a stream of numbers — and you care about whether the next instances in the time series will be higher or lower.