Near Unsupervised Learning for Early Discovery of Novel Patterns: Methods, Scalability and Label Dependability

点击翻译按钮获取中文摘要

This project aims to predict the unknown class labels using the existing small number of class labels. The outcomes of the project have direct relevance to the economy, environment, energy and health sectors due to the abundance of data coming out of these areas. For example, if an oncogene, a gene that can cause cancer when mutated can be found using data with only few labels and a large amount of unlabelled data, the costs and time needed for lab experimentation can be greatly reduced enabling pharmaceutical companies to develop corresponding medicines quicker. It will not only save more lives but also generates millions of dollars of income.