Areas of Expertise
The sheer range of skills offered by Ascent is made possible by close collaboration and excellent working relationships with many different institutions and individual academics. As with all Ascent work, a key part of our role is to maintain tight quality control and to ensure delivery of key business objectives.
Making computers behave more like humans is one of the most demanding tasks in IT. Our work in AI does not usually involve stand-alone systems – the goal is more often to improve existing applications, databases, and environments. What you’ll see is a system that responds better to users and applications vary widely from medical uses and science through to stock trading, the law, speech recognition and factory scheduling.
As a subfield of artificial intelligence, machine learning allows programs to acquire new understanding by ‘learning’ from large data sets. Systems become more efficient over time and machine learning is a valuable alternative approach to some classic AI challenges such as pattern recognition.
Dealing with vast and growing amounts of data is made even more complex by different formats and databases. Data mining means you can analyse information from many different sources including both operational or transactional data and accounting or non-operational data. Sophisticated tools take very disparate inputs and give you useful information that can help with real business issues, revealing whole new patterns and relationships.
Expert systems are designed to simulate a real person’s judgements and behaviour. They are used when human expertise would usually be needed and are based on the work of a knowledge engineer, who analyses and translates human behaviour into computer-friendly rules. Ascent participates in projects designed to produce a range of different systems. These systems can aid or sometimes replace experts in fields such as medicine, accounting, financial services, production and others.
A mathematical approach to problem solving, fuzzy logic arrives at conclusions based on vague or incomplete information. It is an important tool when trying to emulate the way in which many human judgements are made and there is a very wide range of applications ranging from consumer products and fuzzy control to medical diagnostic systems and fraud detection. Compared with conventional approaches, fuzzy logic can help to bring better products to market more quickly and at lower costs.
Neural networks learn by example, using a computational model based on biological neural networks. Ascent can develop networks configured for specific applications across a wide range of requirements and the technology has proved very effective in detecting hidden relationships within a dataset – including stock market and sales data.
Support Vector Machines or SVMs are algorithms that learn by example to assign labels to objects. The technology is often used in demanding applications such as text categorization, character or object recognition, image classification and bioinformatics. As one example, an SVM can spot fraudulent credit card activity by analysing trends across many fraudulent and genuine card activity reports.
Finding an algorithm that can solve an optimization problem within realistic time limits is a demanding challenge. Ascent can develop approximation algorithms that meet a ‘performance guarantee’, bringing intractable problems under control and delivering faster answers with a small, known sacrifice in accuracy. Several different general-purpose techniques are available.