Machine Learning
"Machine learning is a sub-field of computer science that evolved from the study pattern recognition and computational learning theory in
artificial intelligence". The main aim of machine learning is developing machines to be independent learners that can solve problems and interpret data without human interference.
This can be achieved using three methods:
1- Supervised learning
2-Unsupervised learning
3- Reinforcement learning
In supervised learning, The trial and error method is mainly used , where we test computers to give us a known answer, We first enter values
and compare them with the correct answer. This way computers can gain experience so that they avoid previous mistakes and adjust their answer
for the next values so they would improve the results gradually until they give a precise answer.
For the unsupervised learning, the same method used for supervised learning is applied except that the computers are not given the correct answer,
they are left to find structure on its own and find out the hidden patterns in order to come to a correct logical answer.
As for reinforcement learning, a computer should find out how to function by interacting with its surroundings to achieve a certain goal, such as driving a car, or playing chess.
Beside these three main learning methods, there is the semi-supervised learning, which merges supervised and unsupervised learning so that the computer is given an incomplete training signal.
After going through these learning phases, computer become able to "generalize", where they are able to come up with solutions that they haven't encountered before.
This is done through a theory known as Computational Learning Theory, giving probability bounds at which the computer would give the correct answer.
Machine learning and AI are approached through many different fields that aim to make machines as independent and human like as possible.
Example of these fields are neural networks and deep learning where scientists try to create an artificial brain for the computer. This is done using perceptrons
as the simplest unit. Its idea is giving the perceptron some values, and through supervised learning and some algorithms, the perceptron is able to calculate the
error in the answer and keep repeating the process until it gives out a satisfying answer. And then, on a larger scale, the computer forms a series of perceptrons
and other forms of network to interact with the world around him and be able to process data on its own.
other examples for these fields are : Inductive logic programming, Support vector machines, Clustering, Bayesian networks, and many others that helped increasing the
efficiency of machine learning.
References:
https://en.wikipedia.org/wiki/Machine_learning
http://natureofcode.com/book/chapter-10-neural-networks/
Questions:
1- What are the ultimate achievements that could be gained through machine learning ?
2- Is there a new potential fields that could rise and improve Machine Learning in a remarkable way?
3- What are the main obstacles facing developments in this field ?
Revised Essay:
"Machine learning is a sub-field of computer science that evolved from the study pattern recognition and computational learning theory in artificial intelligence".
The main aim of machine learning is developing machines to be independent learners that can solve problems and interpret data without human interference.
This can be achieved using three methods:
1- Supervised learning
2-Unsupervised learning
3- Reinforcement learning
In supervised learning, the trial and error method is mainly used, where we test computers to give us a known answer, We first enter values and compare them with the correct answer.
This way, computers can gain experience so that they avoid previous mistakes and adjust their answer for the next values so they would improve the results gradually until they give a precise answer.
For the unsupervised learning, the same method used for supervised learning is applied except that the computers are not given the correct answer; they are left to find structure on its own and find out the hidden patterns in order to come to a correct logical answer.
As for reinforcement learning, a computer should find out how to function by interacting with its surroundings to achieve a certain goal, such as driving a car, or playing chess.
Beside these three main learning methods, there is semi-supervised learning, which merges supervised and unsupervised learning so that the computer is given an incomplete training signal.
After going through these learning phases, computers become able to "generalize", where they are able to come up with solutions that they haven't encountered before.
This is done through a theory known as Computational Learning Theory, giving probability bounds at which the computer would give the correct answer.
(https://en.wikipedia.org/wiki/Machine_learning)
Machine learning and AI are approached through many different fields that aim to make machines as independent and human like as possible.
Example of these fields are neural networks and deep learning where scientists try to create an artificial brain for the computer. This is done using perceptrons as the simplest unit. Its idea is giving the perceptron some values, and through supervised learning and some algorithms,
the perceptron is able to calculate the error in the answer and keep repeating the process until it gives out a satisfying answer. And then, on a larger scale, the computer forms a series of perceptrons and other forms of network to interact with the world around him and be able to process data on its own.
(http://natureofcode.com/book/chapter-10-neural-networks/)
Other examples for these fields are : Inductive logic programming, Support vector machines, Clustering, Bayesian networks, and many others that helped increasing the efficiency of machine learning.
Questions:
1- What are the ultimate achievements that could be gained through machine learning ?
2- Are there new potential fields that could arise and improve Machine Learning in a remarkable way?
3- What are the main obstacles facing developments in this field ?