Research Section | Analytics Tools | Blog | Scholarly Works | Machine Learning |Data Science
AI and Machine Learning Section
Digital Research

In this brief lesson, we would learn the following concepts

1. What is Machine Learning?

To understand the basics of machine learning, think of how how normal computer programs works. You write a program, and feed it into the computer. Then you provide some input data to the program. Some calculation or processing with the input data and then you have your output. This is illustrated in Figure 1.

With Machine Learning algorithms, you provide an input dataset along with the outputs for the data. The algorithm would then examine this data and build a model it would use to deduce the outputs of new inputs.

The input and output dataset is called Training Data and the process where the algorithm examines the data to build a model(program) is called Training or Learning.

Figure 1: Maching Learning

2. Supervised and Unsupervised Learning

Machine Learning is divided into two broad paradigms

Supervised Learning: In this variation, the system is provided with a given set of feature/lable pairs and the learning algorithm would have to deduce the implicit relationship between the pairs. That is find a rul that predics the label associated with a previously unseen input.

Unsupervised Learning: In this variation a given set of features (feature vector) without labels, the task of the algorithms is to find a way to create a label for the group. Maybe by grouping the data into some kind of natural clusters.

There is a third variation called Reinforcement Learning. This variation is has to to with finding a suitable action to take in a given situation with the goal of maximizing certain reward for the action. This technique was highlighted in 1998 by Sutto & Barto. Here the algorithm is not given optimal inputs, but must deduce via a trial and error process.

Figure 2: Variations of Machine Learning

3. Introduction to Neural Networks

Between the two more topics I would like to introduce: Neural Networks and Deep Learning, I choose to start with Neural Networks since you need some basics of knowledge of Neural Network as a prerequisite for Deep Learning.

In the heart of machine learning is Artificial Neural Network. This is simply a machine learning approach of mimicking the working of the human brain. This is such that attempts are made to deduce the mathematical representation of the way the human brain processes information.

A Neural Network is a network of nodes and edges where the nodes are called Neurons. A neuron is a single processing unit and has an input and an output. The input and output of a neuron are connected to other neurons as well, to form a network. However the arrangement of the connection are arranged in a layered form as shown in the Figure 3.

Figure 3: Structure of a Neural Network

What is a Perceptron? A Perceptron is also a neural network but with just one neuron! Read more on Perceptrons

4. What is Deep Learning?

Deep Learning is a subset of Machine Learning which based on complex neural networks. Deep Learnng works with Neural Networks that have several hidden layers where hidden layers are layers between the input layer and the output layer.

More on Difference Between Machine Learning and Deep Learning.

Follow @datarmatics

Privacy Policy | Terms of Use | Site Map
Copyright 2014. Datarmatics Integrated Services Limited. All Rights Rerserved. Designed by KTTPRO and KTGEN