At 6S Global we offer a range of services related to security, especially cybersecurity. One of the tools we use to provide such services is Machine Learning. That can sound a little daunting if you are not a regular user of information technology. It is actually not anything troublesome, and can, on the contrary, give great power to our systems, so we thought you might like to know a bit more about it.
What machine learning is NOT!
The robots, or the computers, taking over the world, to the detriment of humans. Humans set the context of machine learning. Humans devise the decisions that the computers look for. Humans train the technology systems what is success or failure, and check that they head towards the successful outcome. Humans decide which data the systems work on.
So what is machine learning?
As Wikipedia succinctly puts it “machine learning involves computers discovering how they can perform tasks without being explicitly programmed to do so”. The core of machine learning is the efficiency of the exact rules or algorithms required by the computer being developed by the computer itself, based on previous attempts.
Although human control is still very much setting the overall direction, there is no requirement for the detailed rule structure to be derived by human operators. The machine creates its own and improves it by repetition of the task.
How does this happen?
There are a number of different strategies used in machine learning. Common ones are:
– Supervised learning
– Unsupervised learning
– Reinforcement learning
– Deep learning
These are all approaches to how the computer system is directed to create algorithms. They can be used in combination to achieve the best results. Let’s look at them in more detail.
In this approach, the system is given examples of data input and their output classification. On the base of those examples, the system generates a specific set of rules to yield correct outputs suitable for new inputs.
No guiding examples are given to the system in this approach, but it will derive algorithms from the (usually large) database which can indicate patterns not readily seen by human viewing.
A combination of supervised and unsupervised learning, where some data inputs are loosely labelled is commonly used and is very efficient.
In this mode, the system has desired outcome goals defined for it, then amends its algorithms based on approval or disapproval from the teacher. These triggers allow the algorithms to be refined and improved as the process continues.
This is becoming more common recently. It uses artificial neural networks (which connect elements in a way similar to the human brain). This allows multiple layer processing of complex data. This is particularly helpful for computer vision and speech recognition purposes.
Best of all, why not talk to us at 6S Global about it, and we can advise how you can get the most from machine learning.