Machine Learning is a fascinating field of artificial intelligence research and practice, which is now intensively investigated by experts. Also, it’s becoming increasingly more significant part of the world of software development, being used in a number of tech solutions. Today, let me briefly introduce you to some basic info about ML before we cover more specified topic in further articles.
Sign up for Untitled Kingdom newsletter
Machine learning — basic definition
The most common definition of machine learning stays as follows:
Machine learning is a method of data analysis that automates analytical model building. Using algorithms that iteratively learn from data, machine learning allows computers to find hidden insights without being explicitly programmed where to look.
Alright then… but what does it really mean? How does it work? What is it about? Let’s have a deeper look and make things a little bit clearer.
Basically, machine learning is a technology which studies computer algorithms for learning to do stuff. We might, for instance, be interested in learning how to make accurate predictions or categorise some particular elements. The learning is always based on some observations or data, such as examples, instruction or direct experience.
We may assume machine learning is about training to do better in the future basing on what was experienced in the past. Its main goal is to devise learning algorithms automatically without human intervention or assistance.
Examples of machine learning
Instead of programming the computer to solve some task directly, in machine learning, we seek methods, thanks to which the computer will come up with its own problem-solving program basing on previously provided examples.
Thanks to machine learning we can e.g., make the computer identify the particular type of handwriting, or tell the difference between actual e-mail and spam message. It’s possible thanks to analysing the previously exposed data and finding the pattern, which enables the program to identify the elements basing on their features. Obviously, the technology is used in many other cases, such as:
- Face detection — finding faces in images Topic spotting: categorise news articles according to their topic
- Understanding spoken language: classifying the words and their meaning
- Medical diagnosis: diagnosing a patient as a sufferer or non-sufferer of some disease
- Customer segmentation: grouping the customers e.g., according to their behaviour. This technique lets us e.g., to predict customer’s response to a particular promotion.
Goals of machine learning
The primary goal of machine learning research is to develop efficient general purpose algorithms of practical value. These learning algorithms should be as general-purpose as possible, as they should be easily applied to a broad class of learning problems.
Machine learning process is meant to set a prediction rule that is as accurate as possible in its the predictions. Also, in some contexts (such as medical diagnosis), these rules need to be easily understandable by human experts.
What’s the advantage of machine learning over direct programming?
Shortly saying, In some cases, using “traditional” programming is just a waste of time, and writing your own algorithms seems pointless.
The results of using machine learning are also usually more accurate than in case of direct programming. The reason is that machine learning algorithms are data-driven, and can examine large amounts of data.
Another advantage is that thanks to so-called theoretical machine learning, we might answer some previously inscrutable questions, e.g:
Which properties of a given learning problem make it hard or easy to solve?
How much do you need to know ahead of time about what is being learned to be able to learn it effectively?
In theoretical machine learning the goal stays pretty much the same: We still are interested in designing machine learning algorithms, but at the same time, we hope to analyse them mathematically to understand their efficiency.
The parts of machine learning
There are three major parts of machine learning:
- Model — the system responsible for making predictions or identifications.
- Parameters (features): — the signals or factors, which are used by the model to form decisions.
- Learner: the system which adjusts the parameters by looking at differences between estimated predictions the actual outcome.
Identifying objects: exemplary machine learning problem
Computers can’t see as humans can. So how does the computer learn to identify objects?
Let’s say you want to identify a frog. The first step is to build a model of factors that might be helpful in identifying what’s a frog (images, colors, shapes, etc.) Then you feed in a training set pictures of frogs and see how well the model works.
Every time you use some picture, the learner makes adjustments, and the training cycle goes on. The process of identifying the objects is complicated — there are a lot of parameters (and even parameters within parameters) designed to translate pictures into patterns that the system can match to objects. In other words, in the context of machine learning, the image is really an illustration of the type of patterns that the computer is looking for.
To be continued
This is just an introduction and the basic information about machine learning technology. As machine learning is one of the main directions that Untitled Kingdom’s team develops in, we’re publishing more detailed and in-depth machine learning articles.
Read more from Untitled Kingdom: