M is for Machine Learning

This post is part of a continuing series, “Digital: from A to Z,” that explores what it means to be “digital” from A to Z, broken down into individual blog posts diving deeper into various subjects. Check back regularly to see continuing posts as I work my way through the alphabet and let me know: What’s in your A to Z of digital? You can find me on twitter @Max_Hemingway or leave a comment below.

Machine Learning (ML) allows a computer to learn and act without being explicitly programmed with that knowledge. For example, if you get a computer to recognise a picture of a car and show it some examples of a car, it will then be able to recognise cars going forward and apply what it has learnt against new pictures shown.

Machine learning tasks are typically classified into three broad categories, depending on the nature of the learning signal or feedback available to a learning system. These are

  • Supervised learning: The computer is presented with example inputs and their desired outputs, given by a “teacher”, and the goal is to learn a general rule that maps inputs to outputs.
  • Unsupervised learning: No labels are given to the learning algorithm, leaving it on its own to find structure in its input. Unsupervised learning can be a goal in itself (discovering hidden patterns in data) or a means towards an end (feature learning).
  • Reinforcement learning: A computer program interacts with a dynamic environment in which it must perform a certain goal (such as driving a vehicle or playing a game against an opponent). The program is provided feedback in terms of rewards and punishments as it navigates its problem space.

Source: https://en.wikipedia.org/wiki/Machine_learning 

Machine Learning has opened a lot of uses and applications within business and industry, such as a manufacturing process that looks for defects in products by telling the computer what “good” looks like, with imperfections being identified for further investigation.

One common place for interacting with Machine Learning is through the use of chat bots. You may have used a chat bot without knowing it, with Machine Learning helping to provide the answers to your queries.

An interesting chat bot to try out through Facebook Messenger is Keiko, a people search droid. The search can provide a series of questions to help narrow the search to who you are looking for. Yes, you could do this through a normal search engine, but the thing I like about Keiko is the ability to interact and respond to the query with additional questions or suggestions for the search. Keiko can provide other functions other than just searching for a person. Search engines themselves have a lot of Machine Learning involved in ensuring that the results returned are the best match to the person searching and the search criteria.

If you want to have a go with a Machine Learning program, try Google’s Autodraw. In this application you can draw an object and Google will try and suggest clip art that is similar to your drawing. As the program is used, it learns from the drawings and selections that people make.

Further Reading

Join me next time as I look at “N is for networks” in my Digital A-Z series.  See my last post, L is for legal.

This entry was originally posted in Max’s blog.

Max Hemingway is a senior architect for DXC in the United Kingdom. With more than 25 years of experience, he has a broad and deep range of technical knowledge and is able to translate business needs into IT-based solutions. Currently the chief architect of the BAE Systems account in the UK, Max has a proven track record acquired through continual client engagement and delivery of leading edge infrastructures, all of which have delivered positive results for end-clients, including IT cost reduction, expansion of service capability and increased revenues.


  1. If a machine acts according to a set of rules/procedures/processes and provides answers based on the knowledge that is defined by humans then it is not machine learning, it is only automation. A learning machine must be capable to learn from its sensors or from its given data, through training and retraining of data, to get insight from the data, to generate knowledge from the data and build its own knowledge base. It must also knows how to reason to get such conclusion base on its knowledge base.

  2. Why I think future will be totally on the machine.


  1. […] A neural network is a network that is modeled on a biological brain and nervous systems pathways (synapses) that allows computations to take place at speed and across many nodes. Neural networks have been used across a vast number of tasks but are probably best recognised today as the underlying network of AI (Artificial Intelligence) and machine learning. […]

Speak Your Mind


This site uses Akismet to reduce spam. Learn how your comment data is processed.