Highly commended paper awarded to undergraduate student
This year The Undergraduate Awards received 6472 papers to the programme. Academics from around the globe assess the work received and through a rigorous judging process find the best, most innovative, and creative undergraduate coursework. Just the top 10% of papers in each of the 25 categories are Highly Commended
Jack’s report titled 'PCA and Competitive learning' explores how various techniques could be used to reduce the number of dead units in a neural network. By reducing the number of dead units, you are optimising the number of neurons in the network that contribute to the output. The impact of this means that neural networks can make better use of the data they are trained on. The report highlights how the techniques improve a simple clustering algorithm for grouping different handwritten letters.