Intelligence Series

Machine learning (ML) is defined as the scientific study of algorithms and statistical models that computer systems use to perform a specific task without using explicit instructions. Instead, the system relies on using patterns and inference instead. Machine learning can often achieve in days or hours what might take humans years to program. They may be able to find patterns within data that a human analyst could not discern.

Artificial Intelligence (AI) is the emulation of human intelligence by machines. The traditional areas of AI include reasoning, planning, learning, natural language processing, perception and the ability to move and manipulate objects. AI methods can be placed in two broad categories: The symbolic (top-down) approach focuses on analysing cognition independent of the brain’s structure, in terms of symbol processing. The connectionist (bottom-up) approach involves creating artificial neural networks which imitate the brain’s structure.

Digital Twins are virtual representations of a physical asset, process or system. Sensor data is fed into a 3D model allowing for the asset to be simulated and monitored remotely. Digital Twins have been closely related to Building Information Modelling (BIM) systems but are slightly different.

Big data denotes data that has either a high velocity, a high variety or high volume. For instance, data collected from sensors in real time, data combining text, images and sounds on social media, and the dozen exabytes (10^18 bytes) of data contained in Google’s data centres, are typical examples of big data. The storage and processing of big data requires specific architectures, and advanced analytic techniques must be employed for their exploitation, enabling predictive analytics, the optimisation of complex systems, and the use of data mining to discover new trends and correlations.

Network optimisation algorithms are sets of rules for improving the overall performance of a system by adjusting the flow of network traffic. These algorithms include the use of various techniques and mediums such as predictive analytics, machine learning and neural networks.

Image recognition systems utilise various image processing techniques (image manipulation to enhance image quality, restoration, compression and decompression) and machine learning, to accurately identify objects or people from photos or video feeds. Typically, they consist of a high resolution CCTV system combined with pattern recognition computer software.

Neural networks are a form of computing systems inspired by the human brain. Neural networks are often applied to classic machine learning problems, such as pattern recognition, rule learning or unsupervised data classification. 

Predictive analytics uses historic data to determine patterns and calculate the likelihood of future events. Many industries use this method to determine the risk of detrimental events and apply preventative measures. 

Swarm robotics uses many simple robots to perform tasks. Individual agent capabilities are low, but algorithms rely on the ‘collective intelligence’ that emerges from agent interaction. This can reduce cost of automation and simplify robot design.