Creative Writing and Drawing via Deep Neural Intelligence

Primary supervisor

Additional supervisors

  • Gavin Brown

Additional information

Contact admissions office

Other projects with the same supervisor

Funding

  • Competition Funded Project (European/UK Students Only)
This research project is one of a number of projects at this institution. It is in competition for funding with one or more of these projects. Usually the project which receives the best applicant will be awarded the funding. The funding is available to citizens of a number of European countries (including the UK). In most cases this will include all EU nationals. However full funding may not be available to all applicants and you should read the full department and project details for further information.

James Elson studentship advertisement image

Project description

Recent advances in deep neural networks have enabled the successful training of computers to solve supervised tasks, such as speech recognition, intelligent game playing, and image scene/object classification. How to enable A.I. to perform at levels of competence more similar to humans on creative tasks, such as image and text generation, is still a very open question. This has been of great interest to both computer scientists, and A.I./machine learning companies, but so far, although there has been reasonably good progress in language generation (e.g., machine translation, image caption generation), image generation has been of very limited success.

This adventurous Ph.D. project lies in the intersection of the computer vision and natural language processing fields, and targets the advancement of A.I. on the creative tasks of writing and drawing. One objective is to deepen our understanding on how neural networks can perceive and encode image/text information. The principal outcome will be an A.I. system, which, given a topic described in the format of either an image or text, will be capable of producing a story or generating a drawing of a picture, that are semantically relevant. The comprising neural networks will need to encode knowledge jointly learned from subsets of existing text (e.g., literature) and from existing artworks (e.g., sketches), and generate the aforementioned creative output (the new text and image) based on this encoded knowledge. Building upon current deep learning designs, this project will explore new neural network architectures, training objective functions and optimization algorithms that are suitable for solving the intended learning tasks.

Candidates are expected to have an excellent undergraduate degree in Computer Science or Mathematics or related discipline, and preferably, a relevant M.Sc. degree. Very good experience with computer programming of mathematical models and algorithms (in Matlab, R, or Python) is necessary; also, strong report writing and presentation skills, as well as communication skills and ability to work well with fellow students and colleagues. This is a novel project for dedicated students, with strong personal drive towards scientific innovation who have interest in the fields of machine learning, computer vision and natural language processing.

Key words: Applied mathematics, Artificial Intelligence, Computer Science, Data Analysis, Statistics.

This project is eligible for The James Elson Studentship Award in Artificial Intelligence. The James Elson Studentship will provide an outstanding candidate with fees and an enhanced stipend to carry out a 3-year PhD research project relating to artificial intelligence. The School of Computer Science offers this prestigious PhD studentship for September 2017 entry, for students from the UK and EU who are eligible to pay 'Home' fees.

The deadline for applications for this studentship is Friday 17th March.

▲ Up to the top