WHAT IS MACHINE LEARNING ..?

                                    MACHINE    LEARNING

                          

MACHINE LEARNING can be considered as study of statistical models where we use computer systems rather more to perform any task. It mainly focuses on making predictions using computers. Algorithms in machine learning build a mathematical model of sample data, and to make predictions or decisions without any explicit programs to perform task. Mostly these algorithms are used in some of the applications of email filtering, detection of network intruders and computer vision. [1]

Machine learning is application of Artificial Intelligence that provide system an ability to automatically learn and improve from experience without being explicitly programmed. 

Machine learning focus on the development of computer programs that can access data and use to learn for themselves.[3]

Its primary aim to allow the computers to learn automatically without any human intervention or assistance and adjust its actions accordingly. Systems can learn from data, identify patterns and make decisions with minimal human intervention. [5]

Machine learning is probably used dozen times a day without knowing that we are using it. Some examples where we make of use machine learning in early times:fraud detection, spam filtering, network security threat detection, predictive maintenance and self-driving cars, practical speech recognition, effective web search, and this improved understanding between human and machines. [2]Process of learning through machines begins with observations or previous data analysis, It allows software applications to become more accurate in predicting outcomes using previous outcomes and use statistical analysis to predict an output.[4]

Machine learning is termed as subset of ARTIFICIAL INTELLIGENCE. [1] Mostly it is best way to make progress towards human-level Artificial Intelligence.

By new computing technologies, machine learning today is totally different from the past. It is born from pattern recognition and the theory that computers can learn without being programmed to perform specific tasks, researchers interested in artificial intelligence wanted to see if computers could learn from data. They learn from previous computations to produce reliable, repeatable decisions and results. [5]

Machine Learning is a process in which machine can learn and process data by previous feedback to machine. It is a feature of a computer to obtain results for a given input by the experiences done previously. It consists of many algorithms and models, depending on the complexity of problem the appropriate algorithms and models are selected. [6]

Data mining is a field of study in machine learning, it mainly focuses on exploratory data analysis through unsupervised learning. It is also referred as predictive analytics. [1]

Statistical pattern recognition with emphasis on statistical decision and estimation. Pattern recognition problems are discussed in terms of the eigenvalues and eigenvectors. It is recognition of patterns through statistic predictions.

 

 

Arthur Samuel coined machine learning in 1959. [1]

Tom M. Mitchell quoted about algorithms in machine learning: “A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E”. [1]

According to all these Alan Turing’s proposed “Computing Machinery and Intelligence”“Can machines think?”[1] and do what we think to do It means that can machine do same as what we can do, which further lead to explanations and classifications, much more depth learning about machines.

Generally, humans learn from their past experiences, if human can train a machine to follow certain instructions and predicts outcomes accurately through its previous sessions experiences. This kind of training to machine can be reffered as machine learning. As mentioned before depending upon the complexity of problem to provide output or to generate a choice refer to it. There will be algorithm selected so that problem can generate the correct output within the polynomial time. The previous experiences for the computers are nothing but data. More the data, better the model and higher the accuracy.

 

 

There are some special algorithms to solve problems by predicting and classifying them into different categories.

 

 

Machine Learning categories

11)      Supervised Learning

a)      Classification Algorithm

b)      Regression Algorithm

            Semi Supervised Learning

22)      Unsupervised Learning

a)      Dimensional Reduction

b)      Association

c)      Clustering

33)      Reinforcement Learning

  

Machine learning tasks are classified into several broad categories.

 

(i)                 Supervised learning

(parametric/non-parametric algorithms, support vector machines, kernels, neural networks).

Supervised learning consists of features and labels. When we feed data into the machine it predicts which feature is associated with which labels it accordingly maps data and can be able to predict the values or data that which we enter next further and provides corresponding output when we request for an input which is somehow related to previous data accordingly to its assumptions from previous data remained in itself. So, if we request for any other object feature as an input then it predicts the feature and gives the required label as output.

Supervised algorithms require a data scientist or data analyse with machine learning skills to provide both input and desired output for any data, so that they are able to provide feedback about the accuracy of predictions during algorithm training. Data scientists determine which variables or features that the model should be used to develop predictions. Once training is completed the algorithm will be applied to new data. [4]

For example, when we provide system with set of fruits we refer each fruit with some features like apple-red, round shaped, banana-yellow, curve shaped, similarly for remaining fruits so that next time when we request for apple it will predicts its color shape size etc., in similar way when we request for any other fruits it will predict the values that usually from data that we have entered before. Thus machine learns the things from training data that we have given before and then apply the knowledge to next data.[6]

In supervised learning, the algorithm build a mathematical model from a set of data that contains both the inputs and the desired outputs.

For example, if task for determining whether an image contain a certain object or not, the training data for a supervised learning algorithm would include images with and without that object and each image would have a label designating whether it contained the object. In special cases, the input may be only partially available, or restricted to special feedback.

The learning algorithm receive a set of inputs along with the corresponding correct outputs and the algorithm learn by comparing its actual output with correct outputs to find out errors. It then modifies the model accordingly. Through methods like classification, regression, prediction and gradient boosting, supervised learning uses patterns to predict the values of the label on additional unlabeled data. [5]

Supervised learning is commonly used in applications where historical data predicts likely future events. For example, it can anticipate when credit card transactions are likely to be fraudulent or which insurance customer is likely to file a claim.   

Semi Supervised Learning

Semi-supervised machine learning algorithm is in between supervised and unsupervised learning, they use both labelled and unlabeled data, it is in between both of them. The system which use this method are able to improve learning accuracy. [3]

Semi-supervised learning algorithms develop mathematical models from incomplete training data, where a portion of the sample input doesn't have labels. where a portion of the sample input doesn't have labels. [1]

Semi supervised learning is used for the same applications as supervised learning. But it uses both labelled and unlabeled data for training. This type of learning can be used with methods such as classification, regression and prediction. Semi-supervised learning is useful when the cost associated with labeling is too high to allow for a fully labelled training process. Example for this is for identifying a person face on a web cam.[5]                      

Supervised learning types:

 1) Classification algorithms

2) Regression algorithms

 

 Classification algorithms are used when the outputs are restricted to certain limited set of values. For a classification algorithm that filters emails, the input would be an incoming email, and the output would be the name of the folder in which to file the email. For an algorithm that identifies spam emails, the output would be the prediction of either "spam" or "not spam" represented by the boolean values true and false. [1]

 A classification problem is when the output variable is a category like red or blue.[6]

Regression algorithms are named for their continuous outputs, meaning they may have any value within a range. Examples of a continuous value are the temperature, length, or price of an object.[1]

A regression problem is when the output variable is a real value like weight or height.[6]            

 

(ii)               Unsupervised learning                 

(clustering, dimensional reduction, recommended systems, deep learning)

Unsupervised learning identifies the patterns. There are no labels associated with this type of learning. The learning with unlabeled data is unsupervised learning. It contains set of data of inputs and no desired outputs labels. These contains algorithms which can determine its data patterns on its own.

In unsupervised learning an algorithm that builds a mathematical model from a set of data which contains only inputs and no desired output labels. [1]

 Unsupervised learning algorithm can be used to find structure in the data like grouping or clustering of data points. Unsupervised learning can be able to discover patterns in the data and can group inputs into different categories as similar to feature learning.

Unsupervised machine learning algorithm are used when the information is neither classified nor labelled. Unsupervised learning explains how systems can refer a function to describe a hidden structure from unlabeled data. The system doesn’t find out the right output, but it explores the data and can get some information from datasets to describe hidden structures from unlabeled data.[3]

Here the task of machine is to group unsorted information according to their similarities patterns and differences without any training of previous data. Unlike supervised learning no training will be given to the machine previous inputs or outputs about the data. It should be capable for recognizing and sorting data according to its own assumptions.

Unsupervised algorithms are not trained with desired outcome data. Instead they use an iterative approach called deep learning to review data and arrive at conclusion of data. Unsupervised learning algorithms are also called as neural networks are used for more complex tasks than supervised learning systems.

Some of examples are  image recognition, speech-to-text and natural language generation. These neural networks work by combining a millions of examples of training data and automatically identify relations between many variables. Once trained the algorithm can use its data of associations to predict new data. These algorithms have only become feasible in the age of big data, as they require massive amounts of training data.

For example, it can be used to identify segments of customers with similar attributes who can be treated similarly in marketing sectors or else it can find the main attributes that separate customer segments from each other. Popular some of techniques are self-organizing maps, nearest-neighbor mapping, k-means clustering and singular value decomposition. These algorithms are also used to segment text topics recommend items and identify data outlines.

Unsupervised data is classified into two categories

1)       Clustering: A clustering problem is when you want to discover the inherent group in data, such as group customers in a shopping mall by their purchasing behavior and predict their purchasing capabilities.

 

2)       Association:  An association learning problem is where you want to discover rules that describe large portion of data, like in some cases people that buy X also tend to buy Y.in such cases we use this association type of unsupervised learning.

 

3)       Dimensional reduction can also be used for reducing number of dimensions or states so that complexity of problem can be reduced and solved further easily.

 

Reinforcement Learning

 

Reinforcement learning is a reward based learning. It works on the principle of feedback. If an input is given to model of machine, then if the output produced is correct then the user will take output. If in case the output generated is wrong the machine learns from the feedback obtained due to wrong and can correct its mistakes for next time onward. The correct answer is stored whenever there is output generated is related to previous inputs received and from feedback then it can take input from feedback given to it and process the output.

 

Reinforcement machine learning algorithms is a learning method that interacts with its environment by producing actions and discovers errors or rewards. Trial and error search and delayed reward are the most relevant characteristics of reinforcement learning. This method allows machines and software agents to automatically determine the behavior within a specific context in order to maximize its performance. Simple reward feedback is required for the agent to learn which action is best this is known as reinforcement signal.

This area of deep learning involves models iterating over many attempts that to complete a task. Task that produce correct outcomes are rewarded and task that produce incorrect outcomes are iterated until the algorithm learns the optimal process.

Reinforcement learning is an area of Machine Learning. Reinforcement. It is about taking suitable action to maximize reward in a particular situation. It is employed by various software and machines to find the best possible behavior or path it should take in a specific situation. Reinforcement learning differs from the supervised learning in a way that in supervised learning the training data has the answer key with it so the model is trained with the correct answer itself whereas in reinforcement learning, there is no answer but the reinforcement agent decides what to do to perform the given task. In the absence of training data set, it is bound to learn from its experience.

Main points in Reinforcement learning –

·         Input: The input should be an initial state from which the model will start

·         Output: There are many possible output as there are variety of solution to a particular problem

·         Training: The training is based upon the input, The model will return a state and the user will decide to reward or punish the model based on its output.

·         The model keeps continues to learn.

·         The best solution is decided based on the maximum reward.

 

Reinforcement learning is often used for robotics, gaming and navigation. With reinforcement learning, the algorithm discovers through trial and error which actions yield the greatest rewards. This type of learning has three primary components: the agent (the learner or decision maker), the environment (everything the agent interacts with) and actions (what the agent can do). The objective is for the agent to choose actions that maximize the expected reward over a given amount of time. The agent will reach the goal much faster by following a good policy. So the goal in reinforcement learning is to learn the best policy.

Reinforcement learning algorithms are given feedback in the form of positive or negative reinforcement in a dynamic environment

 

Reinforcement machine learning algorithms is a learning method that interacts with its environment by producing actions and discovers errors or rewards. Trial and error search and delayed reward are the most relevant characteristics of reinforcement learning. This method allows machines and software agents to automatically determine the ideal behavior within a specific context in order to maximize its performance. Simple reward feedback is required for the agent to learn which action is best; this is known as the reinforcement signal.

 

 

References:

[1]- @    https://en.wikipedia.org/wiki/Machine_learning

[2]- @    https://www.coursera.org/learn/machine-learning

[3]-@     https://www.expertsystem.com/machine-learning-definition/

[4]-@     https://searchenterpriseai.techtarget.com/definition/machine-learning-ML

[5]-@    https://www.sas.com/en_in/insights/analytics/machine-learning.html

[6]-@    https://www.geeksforgeeks.org/machine-learning/