Learning to learn and extend with meta-learning

Traditionally, machine learning algorithms are designed to receive a fixed dataset as an input and after training the model will stay the same. However, in real world applications, the machine learning models often need to evolve with a constantly changing dataset that increases with new training examples. The main challenge is to tackle catastrophic forgetting; […]

Continue reading

Enabling NAO robots to learn (Nao programming)

The NAO humanoid robot [1] is a high-end robot (about the size of a small child) used in education and research and also other areas such as banking. It offers opportunities to explore areas such as 3-D imaging, speech recognition and production, vision processing, face tracking, complex limb motions, and artificially intelligent and emotional interaction. The […]

Continue reading

Improving Anonymity in Web Browsing

Our behaviour online is increasingly being tracked by a variety of third parties, including advertisers or providers of free WiFi services, with these data freely traded. In particular, browsing histories can reveal a lot about a person, and allow third parties to de-anonymise people and learn specifics about them. There is an undergraduate project which […]

Continue reading

Freeing Neural Training Through Surfing

Neural network training surfaces are conventionally highly complex and fixed for a sizeable data set, commonly involving ravines and local minima, and which slow or prevent learning as a result [1], [2]. A new method has been developed to make the training surface dynamic and simpler, as though it is being surfed. This design is […]

Continue reading

Improving the prediction of cancer outcomes

Genes control how our cells work by making proteins that have specific functions and act as messengers for the cell. Each gene must have the correct instructions for making its protein. This protein will then be able perform the correct function for the cell. All cancers begin when one or more genes in a cell […]

Continue reading

Neuro–Robotic Juggling using Webots

Robots are increasingly used in unstructured environments for increasingly complex tasks. To be able to cope in these more challenging situations, a self-altering internal neural control architecture has been created [1] that enables both adaptation and learning to take place simultaneously in a symbiotic combination. A first version of the architecture has been tried out […]

Continue reading