LATEST NEWS

.
.
.
.
.
.
.
.
.
.
.
.


15/08/18

What is Machine Learning?

Machine Learning, ML, is the science and method of getting computers to obtain new information and act just like humans. This learning occurs through statistical techniques and identification of patterns to make decisions and learn.

ML is a sector of Artificial Intelligence (AI). The key thing about ML is that human input is restricted as far as possible.

How does Machine Learning work?

ML uses a pair of algorithms, both basic and advanced, to build machine learning models. ML algorithms are usually categorized as unsupervised or supervised.
Unsupervised learning has no outcomes as just the input is being analysed. Supervised Learning has both input variables and an output variable.
The name Supervised Learning comes from the fact that we know the correct answers. However, in Unsupervised Learning, there are no correct answers.
.

ML Process
.
Let's imagine that you are playing Flappy Birds. Your bird is in an environment with several obstacles. Now your bird wants to go past the obstacles.
So, your bird makes an action and flies past the obstacles. While this happens the environment changes to a new state. Moreover, as your bird flies, the environment rewards the bird.

That is how RL works. The bird represents the agent and the goal of the agent is to maximise the reward.

Learning Algorithms include:
  • Neural Networks- these include different layers for analysis and learning of data. The image below is an example using an Audi A7 and the stages Neural Networks take
.

​​​​​​​Neural Network Link
.
  • Local search optimization techniques (e.g., genetic algorithms) - This is a way to solve difficult optimization problems.
Genetic algorithms, in brief, are functioned to provide an effective local search as well as to locate high performance regions of large and complex search spaces in a short period of time.

Why use Machine Learning?

Firstly, it completes tasks that are too computationally challenging for humans to code directly. It would be time-consuming and impractical for humans to directly code for tasks with a large amount of data.

Also, with computational processing that is cheaper and more powerful, it is clearly ideal for building precise models to identify opportunities to increase revenue.

However, that's not all.

There are several other benefits of ML, such as:
  • Tasks are completed at a faster pace at a higher accuracy thus increasing productivity
  • It can be applied to a self-driving google car
  • Increase security in cyber fraud detection
  • Email spam filtering
  • Pattern and image recognition
Now you must be convinced that ML does and will continue to have an impact.

Traydstream

ML is also involved in Trade Finance.
Traydstream aims to revolutionise Trade Finance. The first step is to convert paper documents to digital information by using Optical Character Recognition (OCR engine). Click the link below to read about what OCR is and why it is important.
​​​​​​​
This is when ML plays its part.
The next stage is to check that the documents comply to hundreds of thousands of rules. Checks will occur via algorithms, such as those discussed above, and ML. Checks are made to ensure that rules and regulations are met in the UCP 600 and the ISBP as well as other country requirements.
ML does checks at a higher accuracy and a faster speed than when done manually. We shall wait and see what other applications ML will have in the future.
Scroll to top