Machine Learning seems obscure for many people. Only a small fraction of people really knows what it stands for. However, let’s try to bridge this gap and understand a bit about what ML is.
What is Machine Learning?
It is a kind of Artificial Intelligence(AI) which provides computers with the ability to learn without being explicitly programmed.It mainly focuses on the development of computer programs which can be changed when exposed to new data.
Machine Learning in brief
It deals with the systems that can learn from the data. ML works with data and processes: The data, to discover patterns which can later be used to analyse new data.
ML can be referred to:
- The branch of Artificial Intelligence(AI).
- The methods used in this field.
Its tasks are classified into 3 main categories, that are
- Supervised ML
- Unsupervised ML
- Reinforcement Learning
Supervised Learning is the ML task inferring from the labelled training data.
Suppose we want to make the computer learn, how to distinguish pictures of Cats and Dogs.
First, tell your friends to send you pictures of Dogs and Cats adding a tag ‘Cat’ or a ‘Dog’.
Generally labelling is done by human writers to ensure a high quality of data.
So, now we know the true labels of the data and can use this to ‘supervise’ our algorithm for learning the right way to classify images.
It is a type of Machine Learning algorithm which is used to draw inferences from datasets consisting of input data without any labelled responses.
Let us continue the above example.
Now, let us suppose that your friends forgot to label the images of Cats and Dogs which they sent you.
But, you still want to categorize this data into 2 categories.
In this case you can use the technique called Clustering to separate two images into groups based on some inherent features of the picture.
It is a branch of AI which will allow the machines and software agents to automatically learn the ideal behaviour within the specific context, to maximize its performance.
Let us take an example of playing chess
In a game of chess, the ML doesn’t have every move in the game labelled as successful or unsuccessful but only has the result of the whole game at the end.
Machine Learning Workflow
Reason for using ML
As we know that ML relies on data, the most important necessity of using ML is having the data you can use to train a ML model.The amount of data needed always depends on what you are looking for and how complex your problem is. But, collecting more data is always a good decision.
One thing you need to keep in mind is that, the data you want to train your ML model on should be similar to the one on which you want to make predictions later on.
Types of Machine Learning Algorithms
Decision Tree/ Classification Tree Learning
It is a tree in which non-leaf node is labeled with an input feature.
Arcs coming out from a node which is labeled with a feature are labeled with each of the possible value of the feature.
Each leaf of the tree is labeled with a class or a probability distribution over the class.
Association Rule Learning
It is a rule-based ML method for discovering relations between variables in large database.
Artificial Neural Networks(ANNs)/ Connectionist System
It is a computational model used in ML, Computer Science and many other research areas which is dependent on large collection of connected units known as Artificial Neurons.
In this, the inputs are transformed through more layers.At every layer, the signal is transformed by a processing unit, who’s parameters are learned through training.
Inductive Logic Programming
It uses Logic Programming as a uniform representation for examples, background knowledge and hypothesis.
Support Vector Machines
It is a discriminative classifier. In other words, with the given labelled training data the algorithm outputs an optimal hyperplane which categorizes new examples.
Allocation of a set of observation into clusters(subsets) so that the observations in the same subset are similar in some sense.
Clustering is a method of Unsupervised Learning.
It represents a set of random variables and their conditional dependencies via a Directed Acyclic Graphs(DAG).
Representation Learning/ Feature Learning
A transformation of raw data input to a representation that can be effectively exploited in ML tasks.
Similarity & Metric Learning
It is a class of Supervised Machine Learning. It is mainly related to regression and classification.
Sparse Dictionary Learning
Its aim is to find the sparse representation of the input data in the form of linear combination of basic elements as well as those basic elements themselves. The elements are called atoms and they compose a dictionary.
It is a method used for solving both Constrained and Unconstrained optimization problem which are based on natural selection. This algorithm repeatedly modifies the population of individual solutions.
Rule-Based Machine Learning
It is a term in computer science which is intended to bound any ML method that identifies, learns or evolve ‘RULES’ to store, manipulate or apply.
It is able to achieve this by making a classified decision based on the value of a linear combination of the characteristics. Object’s characteristic is typically presented to the machine in a vector called feature vector.
Application of Machine Learning(ML)
Application ML is in various fields among which some are mentioned below:
These services don’t require any training. They are available for integration into apps and workflows.
- Speech to text
- Text to speech
- Natural Language Processing
- Natural Language Classifier
- Conversation/Chat box
- Language Translation
- Tone Analyzer
- Visual Analyzer
- Recommendation System
- Search Engine
Text & Language
- Predicting Emergency Rooms Wait Times
- Predicting Psychopaths
- Identifying Heart Failure
- Predicting Strokes & Seizures
- Diagnosing Cancer
- Predicting Hospital Readmissions
- Identify Skin Lesions
- Managing Diabetes Patients
Business & Legal
- Machine Translation
- Handwriting Generation
- Text Generation
Image Recognition & Manipulation
- Find Tax Deduction
- Calculate Insurance Pay-out
- Predict Successful Product Launches
- Predict Trade Prices
- Understanding Legalese & Contract Law
- Outsmart the Other Litigator
- Handle a Bankruptcy Legal Practice
- Prevent Money Laundering
- Anomaly Detection
- Improve Customer Service
- Predict the auction price of Heavy Element
- Determine if a Car is a Lemon
Other Applications of Machine Learning
- Automatically adding sound to Silent Movies
- Image Caption Generation
- Caption Videos
- Colourizing Black & White Images
- Focus attention on Objects in Images
- Answer Questions about objects in Photograph
- Turn Sketches into Photos
- Create Styled images from Rough Images
- Search Images by Contents
Predictions of Machine Learning in 2017 & in Coming Years
- Write Music
- Employee Access Control
- Fine-Tune Security Screening
- Protecting Animals
- Determining Bird Species by Audio Recording
- Stopping Span & Malware
- Improve Cyber Security
- Compete Intelligently
- Write Code
- Jam with the Machine
- Inspect Cell Phone Towers & Identify Problems
Some Brands which use Machine Learning (ML)
- Big-Data Soul Searching leads to the gates of Machine Learning
- VCs investing in algorithm-based start-ups are in for a surprise
- ML talent arbitrage will continue at full speed
- Top down ML initiatives built on PowerPoint slides will end with a whimper
- Deep Learning Commercial Success will be few and far in between
- Exploration of Reasoning and Planning under uncertainty will pave the way to new Machine Learning heights
- Humans will still be central to decision making despite further Machine Learning adoption
- Data Scientist or Developers will introduce Machine Learning into their companies
- Hyatt Hotels
- Under Armour
- The North Face
- Macy’s and many more…