Alan Turing once said: “It seems probable that once the machine thinking method had started, it would not take long to outstrip our feeble powers… They (the machines) would be able to converse with each other to sharpen their wits.”
Turing’s prediction isn’t too far off the mark and today it has manifested itself as Artificial Intelligence (AI) and Machine and Deep Learning. It’s already all around us, without many people even realising. For instance, if you subscribe to Netflix, then you’ll already be experiencing Machine Learning as it makes personalised recommendations based on your viewing habits.
Machine Learning is a type of AI that enables software applications to become intelligent and accurate in predicting outcomes without human intervention.
It builds algorithms that allow the machine to learn based on the data it is fed and researchers at Birmingham City University are currently exploring the capabilities of this technology. The University encourages forward-thinking businesses to come and find out more about this cutting-edge technology and how collaborating with us could help.
On the Edge
With a particular focus on Edge Machine learning, (enabling small devices to make intelligent decisions) Birmingham City University’s research is being led by Mohamed Gaber, a professor in big data analytics.
Mohamed’s interest in Edge Machine Learning came from his early work in data mining and machine learning on high performance computers where he began investigating if machine learning models could operate on small devices. Fast forward to today and these small edge devices have entered our everyday lives with more and more people using wearables to track their sleep patterns, fitness levels and remind them to drink water. Soon, we will all be surrounded by multiple, small connected, intelligent devices – at home or work, in our cars and in our cities.
These small devices typically have limited memory and weak processing power and use sensors to monitor their environment and interact with their user. To make predictions or intelligent decisions, they rely on Machine Learning models to interpret signals from their sensors from the cloud. However, accessing data via the cloud can be impractical due to bandwidth, connectivity and privacy issues, and Birmingham City University’s research into Edge Machine Learning is focused on overcoming this by enabling these small Internet of Things (IoT) devices to make intelligent decisions for themselves.
Mohamed and his team at Birmingham City University are keen to share their knowledge and help businesses understand more about this technology and its potential for helping make the most of the ever-increasing amounts of data being generated. One particular area of interest is the potential of Machine Learning to predict serious health problems and Mohamed believes it could not only help save lives but reduce costs in the NHS too.
Edge devices and health care
Birmingham City University research is looking at how Edge Machine Learning could be applied to small devices for people with serious health complaints. These small devices could be worn by people with a variety of health conditions and be used to monitor their vital signs and alert them if they need to take action, or even call for an ambulance or doctor on their behalf if it detected an urgent situation. Mohamed firmly believes that this continual monitoring of a variety of conditions could ultimately reduce pressure on the NHS as patients learn to rely on their smart device and reduces the need for them to seek medical attention in non-urgent situations.
Exciting discoveries about the application of these technologies are happening all the time and it was recently announced that a Machine-Learning system had been developed that is as good as human experts at detecting eye problems and prescribing appropriate courses of treatment. The new system demonstrated that it was able to refer patients with over 50 different types of eye diseases for treatment with the same level of accuracy as world leading eye specialists.
Deep learning
Another area of research that Mohamed is currently looking into is Deep Learning, a subset of Machine Learning. It functions in a similar way but essentially its capabilities are different and uses algorithms that can learn features, instead of handcrafting them. Based on a layered structure of algorithms, the Deep Learning model draws conclusions in much the same way as a human. It’s these models that are at the core of driverless cars which learn to distinguish between different scenarios, hazards and road signs.
Deep learning models can only be successful when fed with the correct data and an example of this is Google’s AlphaGo, a game similar to chess only much more difficult. A Deep Learning model based virtual competitor continually played against professional Go players without being instructed on when to make a move and eventually became so sophisticated at the game that it managed to defeat one of the world’s greatest players.
Ensuring that Deep Learning doesn’t draw incorrect conclusions is work in progress. Birmingham City University’s research into it hopes to contribute to the accuracy of these conclusions and the University invite any businesses looking to understand how this technology could be applied to industry challenges and drive innovation to get in touch and find out more.
The importance of data
To work effectively, Machine and Deep Learning needs to be fed accurate, real time data and at Birmingham City University, Mohamed and his team work closely with Dr Adel Aneiba who brought The Things Network to Birmingham. The network is based on LoRaWAN™ technology – low energy, long range and low bandwidth, and enables Internet connectivity without 4G or WiFi. The use of sensors enables data to be collected in real time and accurately and energy efficiently send it across the network for analysis. Mohamed and Adel believe that the ability for data to be collected in this way opens up a host of opportunities for local businesses, giving them the ability to harness data effectively and use it to increase understanding of behavioural patterns and predictive analytics.
Birmingham City University students are currently working on a project with a financial organisation to analyse data relating to customer behaviours and how any learnings taken from the data could be applied to improve customer service. Any organisation thinking about machine and deep learning needs to think about what their business could predict in the future, what could happen in various scenarios, and how these predictions could enable them to work smarter, better or make business improvements.
From security surveillance to predicting a person’s likelihood of developing cancer, the potential for this technology is immense. Birmingham City University welcomes enquiries from local businesses who would like to know more about Machine and Deep Learning and what it could help their business to predict in the future.
