Surrey Computer Science Blog

The blog from the Department of Computer Science at the University of Surrey

Surrey CS PhD student presented research at IEEE SSCI 2016 in Greece

Author: Andreas Antoniades (PhD student)

The 2016 IEEE Symposium Series on Computational Intelligence (IEEE SSCI 2016), co-organised by the Department’s Prof Yaochu Jin as the General Co-Chair, was held in Athens, Greece, a city full of arts, philosophy and historical attractions, between the 6th and 9th of December. Bringing together more than 20 different symposiums relevant to computational intelligence, SSCI was a great opportunity cross-fertilisation and collaboration.

The conference and more specifically the symposium on Computational Intelligence for Big Data (CIBD), was a great opportunity for me to present and promote my work. Throughout the course of the conference, I had the opportunity to discover new applications in the topic of computational intelligence and attend numerous keynote presentations from pioneers in the field.

My paper titled “An improved mini-batching technique: Sample-and-Learn” revolved around Big Data and it’s effects on machine learning and data analytics. The days where data was scarce and difficult to find are long gone. Now thanks to the Internet of Things, unprecedented volumes of data are created daily. Our algorithms need to adapt in order to cater for large scale analytics. It is in this spirit that a reservoir sampling algorithm is proposed to alleviate the computational load of models trained with big data. The resulting algorithm was proven to decrease training times of neural networks while also providing equivalent or better accuracy than state-of-the-art.Andreas at SSCI 2016

Computer Science PhD student/Research Fellow at MLSP 2016

Author: Santosh Tirunagari (PhD Student and Research Fellow)

The 26th IEEE International Workshop on Machine Learning for Signal Processing (MLSP 2016) took place at Lloyd’s Baia Hotel in Salerno, Italy from 13 Sept to 16 Sept 2016. MLSP is organised annually by the IEEE Signal Processing Society with a primary focus on  bringing Information Technologies and Databases & Information Systems professionals on to a common platform where they can exchange ideas and opinions to advance knowledge for Signal Processing, Machine Learning, Information Processing, Data Analysis and Pattern Recognition.

The conference included oral and poster presentations on the most recent and exciting advances in machine learning for signal processing through keynote talks, tutorials, as well as special and regular single-track sessions. Attending this conference was a great opportunity for me, as I met several top researchers in my field of  work i.e., in machine learning and signal processing. The conference had special sessions on Bayesian Machine Learning for Neural Signal Processing, Advances in Gaussian Processes for Machine Learning and Signal Processing and Computational Methods for Audio Analysis. All these sessions are of interest to me.

My paper entitled “Automatic Classification Of Irregularly Sampled Time Series With Unequal Lengths: A Case Study On Estimated Glomerular Filtration Rate”, was accepted  at this conference and I was invited to present this work as a 15 minute  lecture presentation.  This paper was produced as part of MRC CKD project which essentially deals with modelling Chronic Kidney Disease (CKD, which  is considered as significant cause of morbidity and mortality across the  world. Patients with CKD have increased risk of death from cardiovascular disease and end stage kidney failure, leading to dialysis and kidney transplant. Indeed, according to an NHS Kidney Care report in 2012 , CKD was estimated to cost £1.45 billion in 2009-10; 1.8 million people were diagnosed with CKD in England; and, there were potentially 900,000 to 1.8 million people with undiagnosed CKD. Therefore, the importance and urgency of managing CKD cannot be over-emphasized. In this paper using the state-of-the-art machine learning techniques and methodologies we automatically screen patients’ data to identify those whose kidney function is at risk of deteriorating into more severe stages of CKD, thus enabling the clinicians to monitor their patients remotely.

This work was supported by the Medical Research Council under grant number MR/M023281/1.  The project details can be found at http://www.modellingckd.org/.

dsc09722 dsc09744 dsc09894

Surrey CS PhD student at IJCNN 2016

Author: Andreas Antoniades (PhD student)

The flagship conference for neural networks, 2016 International Joint Conference on Neural Networks (IJCNN 2016), was held as part of the IEEE World Congress on Computational Intelligence (WCCI 2016) in the beautiful city of Vancouver, British Columbia, Canada. The conference took place between the 24th to 29th July. The objective of this bi-annual congress is to bring together the latest advancements in Computational Intelligence.

The conference was a great opportunity for me to present and promote my work, as well as meet some of the pioneers in the area of machine learning. Throughout the week, I was able to benefit from a number of parallel sessions and discuss possible collaborations with academics from different institutions and different research areas.

My presented work revolved around the concept of feature selection for cancer in children. Given a dataset of patient records, it is important to understand which genes are most important for the detection and treatment of Neuroblastoma, Rhabdomyosarcoma, non-Hodgkin lymphoma and the Ewing family of tumors. Clinicians often have to consider thousands of genes in order to correctly diagnose and treat a patient. The proposed algorithm utilised the latest advancements in Autoencoders to select the most important genes in an automatic manner, substantially reducing diagnostic waiting times.

PhD student Santosh Tirunagari won Best Presentation Award at ICICV 2016

Author: Santosh Tirunagari (PhD student)

The 2016 IEEE International Conference on Image, Vision and Computing Conference (ICIVC) took place in Portsmouth, UK, from August 3rd to 5th August 2016, held at the Conference Auditorium, University of Portsmouth. The main objective of the ICIVC 2016, is to bring together innovative academics and industrial experts in the field of Image, Vision and Computing to a common forum. I attended this conference in order to present a article entitled: “Can DMD obtain Scene Background in Color?” which is available at https://arxiv.org/pdf/1607.06783.pdf. This is an excellent opportunity because this three-day conference brought together various academics working particularly in the field of computing. Therefore, this provided me with an interactive and friendly platform to present my work, discuss my work with fellow PhD students.

The main track of the Conference was composed of three types of session, namely: 1) Oral presentation sessions (1,2),  2)Plenary session  3) keynote session.  Keynote and plenary Speakers in ICIVC 2016 included Prof. Ezendu Ariwa (University of Bedfordshire, UK),  Dr. Branislav Vuksanovic (University of Portsmouth, Portsmouth, UK), Prof. Jenny Benois, (University of Bordeaux, France), Dr. Hui Yu (University of Portsmouth, UK) and Dr. Huseyin Seker, The University of Northumbria at Newcastle, UK. My presentation was in session 2 and Dr. Branislav Vuksanovic was the chair. My presentation was selected as the best presentation from session 2 and was awarded a certificate at the end of this session.

Santosh Tirunagari receiving ICIVC 2016 Best Presentation Award

Dr Norman Poh gave a talk at Workshop on Big Data: Modelling, Estimation and Selection

This two-day workshop (https://indico.math.cnrs.fr/event/830/timetable/#20160609) was organised by CNRS and took place from 9-10 June 2016 in Lille, France, gathered researchers from the industry and the academia working in the area of big data. While the talks on the first day were tutorials targeting the general audience; on the second day, talks were focused around technical and mathematical details such as alternative methods to improve gradient-descend type of optimization. Dr Norman Poh’s talk was arranged at the beginning of the second day in order to link the high-level tutorial on the first day and the technical talks which followed after that.

Dr Poh’s talk was entitled ‘What could we learn from millions of patient records? A machine-learning perspective’. The talked provided the healthcare context, justifying why healthcare records are a big data problem and motivated the need to develop novel machine-learning algorithms that are more adapted to modelling the temporal dynamics, potentially over the life course of a patient, defined on a large concept space, which is spanned by hundreds of thousands of clinical concepts. In addition, the population denominator, which is in the order of millions of patients, thus qualifies the problem of modelling healthcare records as ‘big data’.

By way of using Chronic Kidney Disease (CKD) as a disease of interest, Dr Poh’s talk highlighted how classical machine learning tasks such as classification, regression, and clustering can be applied to modelling CKD. The slides can be found here (http://personal.ee.surrey.ac.uk/Personal/Norman.Poh/data/2016-06-10_BigData_Lille_compact.pdf).

Dr Poh’s work is funded by MRC project: Modelling the Progression of CKD (www.modellingCKD.org).

Talk abstract

Increasing healthcare cost coupled with an ageing population in both developing and developed worlds means that it is important to understand disease demographic profiles in order to better optimize resources for quality health and care. By using Chronic Kidney Disease (CKD) as a case study, I will present challenges that are related to understanding, modelling and predicting the progression of CKD; and how machine learning techniques can be used to solve them. Examples include calibration of estimated Glomerular Filtration Rate (eGFR), modelling of eGFR, automatic selection clinically relevant variables, and non-linear dimensionality reduction for data discovery.

2015 British Renal Society (BRS) Conference

By Santosh Tirunagari

The 2015 British Renal Society (BRS) Conference took place in Leeds, UK, from 30th June 2015 to 2nd July 2015, held in the ‘University of Leeds, Conference Auditorium’. The main objective of the BRS conference is to promote effective patient-centred multi professional care in order to improve quality of life for people with kidney failure. My supervisor Dr. Norman Poh and I attended this conference in order to present our poster entitled : “AUTOMATIC CLASSIFICATION OF LONG-TERM KIDNEY FUNCTION FOR CKD PATIENTS USING MACHINE LEARNING TECHNIQUES: CLASSIFYING EGFR TRENDS” Which is available at [1]. Another reason for attending BRS 2015 is to find potential collaborators who work on renal diseases from the medical field. This is an excellent opportunity because this three-day conference brought together various medical professionals, clinicians, medical device exhibitors and academics working particularly on renal diseases. Therefore, this provided us with an interactive and friendly platform to present our posters, discuss our work with the clinicians, share ideas with each other, and gauge their interest in automated algorithms to solving their problems. More importantly, it was a good opportunity for us to understand the broader technical and clinical problems in managing patients with renal diseases.

1 2 3 13