Foundation technologies

We develop models, techniques and methods that can be used to enhance practices in artificial intelligence (AI).

Our research is based on over a decade of experience in:

  • cognitive science
  • computational neuroscience
  • cognitive psychology

The core of the technology we have built uses self-learning (unsupervised learning techniques), that dynamically self-evolve to represent new situations, such as:

  • incremental (ongoing) learning from real-time data
  • data integration or fusion
  • making sense of unstructured data such as text
  • capturing sequences in data
  • efficient processing of large data sets

We encourage deep learning, while laying the foundation of brain and mind inspired and self-learning techniques. Our work on developing systems that can generate abstract understanding from detailed information is inspired by the columnar architecture and the multi-layered functionality of the human neocortex. Current uses of our algorithm include:

  • health data mining,
  • smart electricity meter stream mining,
  • text mining, and
  • social media data analysis.

Unsupervised machine learning

We have developed The Growing Self-Organising Map (GSOM), that is:

  • self-generating,
  • data-driven,
  • unsupervised and,
  • a neural network.

It works through self-organization but its key advantages are:

  • its ability to structurally adapt to the underlying topology of data,
  • two-dimensional map generation,
  • providing intuitive visualisation of the data,
  • its ability to separate data into clusters,
  • and represent the data at different levels of abstraction as cluster hierarchies

The GSOM has been used in fields such as biology, text mining, engineering, health and social sciences.

Text comprehension

Several innovative technologies and algorithms for text mining have been developed by the group. Key novelties and advantages are;

  • faster processing,
  • new feature extraction integration of NLP
  • capturing semantics,
  • data enrichment,
  • and visualization.

Techniques have been applied to clustering large volumes of text data – web articles, research papers, pathology reports, Twitter and other social networks.

A very innovative research direction under this topic is the extending of sentiment analysis into further understanding of user emotions and emotional states.

It is expected that this work will provide organizations such as healthcare practices, governments, educational institutions a much deeper and detailed understanding of the thinking and ‘feelings’ of their clients.

Analytical capabilities

Upon the foundation technologies, our analytical capabilities primarily focus on;

  • Predictive analytics and Forecasting
  • Exploratory Analytics
  • Data warehousing and information management
  • Visual Analytics
  • Multimodal Big Data Analytics (Text, Image, Video)
  • Event monitoring, detection and prediction
  • Real-time insights
  • Data stream mining
  • Social Media Analytics

Data stream mining

Whenever someone uses a point of sale system (POS) at a store, operates a smart meter or uploads an image to social media they are contributing to a data stream.

In the age of Big Data, we ensure proper management and analysis of these streams. By doing this we are ‘mining‘ the data streams with algorithms. To get relevant information from data streams we have developed:

  • integrated online learning, and
  • adaptive learning algorithms

Data visualisation

As volumes of data generated by individuals, businesses and societies increase, it is necessary to conceptualise, design and develop novel means of managing these large datasets. Our research lab experiments with a number of conventional and new technologies in data and information management, such as:

  • relational databases,
  • data warehouses,
  • NoSQL databases and
  • data lakes.

We possess expertise in the live implementation of information management systems based on databases and data warehouses for both research and industry-based projects. One of our current research projects focuses on customisation of data lakes to suit growing needs of predictive and prescriptive analytics.

It is imperative to visualise managed datasets and analytics outcomes. To this end, we have developed expertise in data visualisation techniques. This ranges from skills and experience in current visual analytics tools to custom visualisation methods for multi-dimensional datasets.