Optimal Design of Machine Learning

Modern real-world problems often require the design of complex Machine Learning structures such as Deep Neural Networks. This design may require some the work of human experts. In this project you will investigate automatic design of machine learning algorithms, that is the emerging trend called Neural Architecture Search. The latter is a challenging modelling and optimisation problem. The project requires a moderate amount of programming and an understanding of basic AI techniques (Neural Networks and Evolutionary Algorithms).

Novel Training Techniques for Machine Learning

Large Neural Systems often require to be trained to solve some problems, such as big data classification. The training in machine learning can be seen as a complex optimisation problem. This project is about investigating novel algorithmic solution to train large Neural Systems, e.g. in Deep Learning.  

Fitness Landscape Analysis for Numerical Optimisation

Fitness Landscape Analysis (FLA) is a set of techniques consisting of analysing optimisation problems with the aim of exploiting the collected information to design an effective Optimiser. The present project consists of the design of a testing technique and/or of an optimisation algorithm using one FLA technique. Landscape features like multimodality and epistasis. The student will be provided with  existing knowledge/software and will be tasked with the developement of some conceptual or software aspects of it. The project requires a moderate amount of programming and a certain degree of mathematical curiosity (basic algebra and calculus).

Studies on P Systems

P Systems are models of computation and in many implementations are equivalent to the Turing Machine. Since P Systems can be used as number generators, they have been used a sampling mechanism of Optimisation Algorithms. In this project you will focus focus on a specific type of P System such as Spiking Neural P Systems (SNPS) or Numerical P System to use it as an Optimiser. The student will be provided with an implementation of a P System and will be asked to develop some theoretical and/or software aspect of the system. For example, the project may include the implementation of a bank of parallel P Systems and/or the design of a novel  supervising algorithm, namely Guider Algorithm. The project requires a substantial amount of programming and an interest for automata theory and optimisation.




The following project themes are suitable for Year 3 and  Year 4 Undergraduate students as well as for MSc students for the School of Computer Science, University of Nottingham, UK. Specific arrangements will be made with the individual students to meet the background, set of skills and ambition of each student.