Machine Learning is an algorithmic approach of creating computer models with the ability to learn and adapt from a given dataset, these models can then be used to make useful predictions of results against similar but never-seen-before data. It is often refereed to as a subset of Artificial Intelligence and forms the very base on which AI models are created.
Mimicking the Human Brain
To get you started, examples from daily life
- Siri, Alexa, OkGoogle, Personal Assistants and Chatbots on electronic devices
- Google Maps’ feature of telling in advance time to your destination and performing traffic rerouting
- Facebook’s facial recognition, Identifying you and your friends in a photograph anywhere on Facebook
- Recommender systems suggesting to you what to watch next in Netflix, Amazon and ‘People you may know’ sections in FB, LinkedIn, Instagram
- Bioinformatics & Healthcare treatments, Disease diagnosis and Biotechnological discoveries in Medical Science
- Gmail’s Spam filtering algorithms, MyAnalytics by Microsoft
- Google Self Driving Cars
- Dynamic prices in Travel websites
- Personalized Advertisement contents based on User’s search pattern
21st century is all about data, so much so that is called as the new Oil, its true and Machine Learning is that very tool which is capable of harnessing this immense power of exponential growth of technology and data we are seeing in world right now
On Industrial level, Machine Learning Algorithms are helping organisations to
- Increase productivity via Automation in manufacturing or services industries
- Models created to detect symptoms for critical illnesses at early stages, helping doctors in analyzing scan results, surgical procedures, personalized medical treatments.
- Early warning systems for storms, hurricanes, natural disasters including weather forecasting
- Identifying and controlling the spread of epidemics
- Crime investigation, control and prevention by Security agencies
- Translation of texts across languages
- Political trends in Election Campaigning
Machine learning algorithms are categorized as supervised, unsupervised, semi-supervised & Reinforcement Learning (Deep Learning)
Supervised Machine Learning
In supervised learning, the algorithm creates a mathematical model from a dataset which have both inputs and desired labeled outputs. The presence of labeled output makes the models to define the correlation between input data and their corresponding results. The model is trained repeatedly though a process in which it is fed with training data. The trained model is then tested on similar dataset to predict desired results. The training process continues until the model achieves a desired level of accuracy on the training data. Classification, Regression are examples of Supervised Machine Learning
Example Chart of Regression
Supervised machine learning is analogous to a student learning a subject by studying a set of questions and their corresponding answers. After mastering the mapping between questions and answers, the student can then provide answers to new (never-before-seen) questions on the same topic
Unsupervised Machine Learning
In contrast, Unsupervised algorithms creates models where the data used to train the model is not classified or labeled which means that the model is fed only with input values but without any information on part of output/labels.Unsupervised learning algorithms are used to find patterns or structures in the given data distribution. ‘Grouping’, ‘Categorization’ and ‘Clustering’ of data points comes under Unsupervised Machine Learning examples.
Exampe image of Unsupervised Learning
Semi-Supervised Machine Learning
Semi-supervised learning involves algorithms of mathematical models from a incomplete training dataset, where a only a portion of the sample input have labels and rest of it is unlabled or without any output results.
Example Image of Semisupervised Learning
Semi-supervised models fall somewhere in between supervised and unsupervised learning algorithms, In this method there is only small quantity of labeled data is present for training while and a large quantity of unlabeled data. The ML models which uses semi-supervised methods are capable of improving the learning accuracy by themselves. Semi-supervised learning technique is to infer labels for the unlabeled output samples, and then to train on the inferred labels to create a new model. Semi-supervised learning is most useful when acquiring labeled data is expensive but unlabeled examples are plentiful.