Professor David Elizondo

Job: Professor in Intelligent Transport

Faculty: Computing, Engineering and Media

School/department: School of Computer Science and Informatics

Research group(s): The De Montfort University Interdisciplinary Group in Intelligent Transport Systems (DIGITS)

Address: De Montfort University, The Gateway, Leicester, LE1 9BH, United Kingdom

T: +44 (0)116 207 8471

E: Elizondo@dmu.ac.uk

W: https://www.dmu.ac.uk/digits

Close all sections

Open all sections

Personal profile

Dr. David Elizondo is a Principal Lecturer in the Department of Computer Technology at De Montfort University. After completing his BA in Computer Science from Knox College , Galesbourg, Illinois, USA, he worked as a software engineer/lab manager for a latinoamerican agronomical research and teaching institute based in Costa Rica ( CATIE ). This institute, through a Swiss project, sponsored him to do a MS in Artificial Intelligence at the Department of Artificial Intelligence and Cognitive Computing of the University of Georgia, Athens, Georgia, USA. After this he obtained a PhD in computer science from the University of Strasbourg , France in cooperation with the Swiss Dalle Molle Institute for Perceptual Artificial Intelligence (IDIAP). He then worked for Neuvoice, formerly Neural Systems, a spin off company of the University of Plymouth , UK. As a senior researcher he worked in the development of an intelligent monitoring system for the petroleum industry. This system was based on neural network techniques. Later, he worked as a software architect for ACTERNA, an international company which supplies software/hardware solutions to telecom companies. He was part of the team developing QMS, a quality of service management system for leased lines. In parallel to this work, he was a part time lecturer at the University of Plymouth where he taught database, and data structures and algorithms.

Research group affiliations

The De Montfort University Interdisciplinary Group in Intelligent Transport Systems (DIGITS)

I am also an active member of the following research groups:
(1) The Cyber Security Centre
(2) The Centre for Computational Intelligence (CCI).

I am the research leader of the CCI Neural Network subgroup, which is particularly well known internationally for the research work conducted in the area of Constructive Neural Networks and Linear Separability as evidenced by my on-going list of high quality publications in these two fields of research.

Publications and outputs

Enhancing Object Segmentation via Few-Shot Learning with Limited Annotated Data
dc.title: Enhancing Object Segmentation via Few-Shot Learning with Limited Annotated Data dc.contributor.author: García-Aguilar, Iván; Jafri, Syed Ali Haider; Elizondo, David; Calderón, Saul; Greenfield, Sarah; Luque-Baena, Rafael M. dc.description.abstract: Significant advancements in machine learning in recent years have revolutionized multiple sectors. The Segment-Anything Model (SAM) is a notable example of state-of-the-art image segmentation. Despite claims of zero-shot generalization, SAM exhibits limitations in specific scenarios like medical mammography images. SAM generates three segmentation masks per image to address this and recommends selecting the one with the highest confidence score. However, this is not always the optimal choice. This paper introduces a system that extends SAM’s segmentation capabilities by automatically selecting the correct mask, leveraging few-shot learning methods and an Out-of-Distribution threshold strategy. Several backbones were subjected to experimentation, highlighting the relationship between the support set size and the model’s accuracy.
Characterising Payload Entropy in Packet Flows—Baseline Entropy Analysis for Network Anomaly Detection
dc.title: Characterising Payload Entropy in Packet Flows—Baseline Entropy Analysis for Network Anomaly Detection dc.contributor.author: Kenyon, Anthony; Deka, Lipika; Elizondo, David dc.description.abstract: The accurate and timely detection of cyber threats is critical to keeping our online economy and data safe. A key technique in early detection is the classification of unusual patterns of network behaviour, often hidden as low-frequency events within complex time-series packet flows. One of the ways in which such anomalies can be detected is to analyse the information entropy of the payload within individual packets, since changes in entropy can often indicate suspicious activity—such as whether session encryption has been compromised, or whether a plaintext channel has been co-opted as a covert channel. To decide whether activity is anomalous, we need to compare real-time entropy values with baseline values, and while the analysis of entropy in packet data is not particularly new, to the best of our knowledge, there are no published baselines for payload entropy across commonly used network services. We offer two contributions: (1) we analyse several large packet datasets to establish baseline payload information entropy values for standard network services, and (2) we present an efficient method for engineering entropy metrics from packet flows from real-time and offline packet data. Such entropy metrics can be included within feature subsets, thus making the feature set richer for subsequent analysis and machine learning applications dc.description: open access article
Natural Language Processing Tools and Workflows for Improving Research Processes
dc.title: Natural Language Processing Tools and Workflows for Improving Research Processes dc.contributor.author: Khan, Noel; Elizondo, David; Deka, Lipika; Molina-Cabello, Miguel dc.description.abstract: The modern research process involves refining a set of keywords until sufficiently pertinent results are obtained from acceptable sources. References and citations from the most relevant results can then be traced to related works. This process iteratively develops a set of keywords to find the most relevant literature. However, because a keyword-based search essentially samples a corpus, it may be inadequate for capturing a broad or exhaustive understanding of a topic. Further, a keyword-based search is dependent upon the underlying storage and retrieval technology and is essentially a syntactical search rather than a semantic search. To overcome such limitations, this paper explores the use of well-known natural language processing (NLP) techniques to support a semantic search and identifies where specific NLP techniques can be employed and what their primary benefits are, thus enhancing the opportunities to further improve the research process. The proposed NLP methods were tested through different workflows on different datasets and each workflow was designed to exploit latent relationships within the data to refine the keywords. The results of these tests demonstrated an improvement in the identified literature when compared to the literature extracted from the end-user-given keywords. For example, one of the defined workflows reduced the number of search results by two orders of magnitude but contained a larger percentage of pertinent results. dc.description: open access article
Out-of-Distribution Detection with Memory-Augmented Variational Autoencoder
dc.title: Out-of-Distribution Detection with Memory-Augmented Variational Autoencoder dc.contributor.author: Ataeiasad, Faezeh; Elizondo, David; Ramírez, Saúl Calderón; Greenfield, Sarah; Deka, Lipika dc.description.abstract: This paper proposes a novel method capable of both detecting OOD data and generating in-distribution data samples. To achieve this, a VAE model is adopted and augmented with a memory module, providing capacities for identifying OOD data and synthesising new in-distribution samples. The proposed VAE is trained on normal data and the memory stores prototypical patterns of the normal data distribution. At test time, the input is encoded by the VAE encoder; this encoding is used as a query to retrieve related memory items, which are then integrated with the input encoding and passed to the decoder for reconstruction. Normal samples reconstruct well and yield low reconstruction errors, while OOD inputs produce high reconstruction errors as their encodings get replaced by retrieved normal patterns. Prior works use memory modules for OOD detection with autoencoders, but this method leverages a VAE architecture to enable generation abilities. Experiments conducted with CIFAR-10 and MNIST datasets show that the memory-augmented VAE consistently outperforms the baseline, particularly where OOD data resembles normal patterns. This notable improvement is due to the enhanced latent space representation provided by the VAE. Overall, the memory-equipped VAE framework excels in identifying OOD and generating creative examples effectively. dc.description: open access article
Oil spill classification using an autoencoder and hyperspectral technology
dc.title: Oil spill classification using an autoencoder and hyperspectral technology dc.contributor.author: Carrasco-Garcia, Maria Gema; Inmaculada Rodríguez-García, M.; Ruiz-Aguilar, Juan Jesus; Deka, Lipika; Elizondo, David; Turias-Domínguez, Ignacio J. dc.description.abstract: Hyperspectral technology has been playing a leading role in monitoring oil spills in marine environments, an issue of international concern. In the case of monitoring oil spills in local areas, hyperspectral technology of small dimensions becomes the ideal solution. This research explores the use of encoded hyperspectral signatures to develop automated classifiers capable of discriminating between polluted and clean water, and even distinguishing between various types of oil. The overall objective is to leverage these classifiers to be able to improve the performance of conventional systems that rely solely on hyperspectral imagery. The acquisition of the hyperspectral signatures of water and hydrocarbons was carried out with a spectroradiometer. The range of the spectroradiometer used in this study covers the ranges between [350-1000] (visible near-infrared) and [1000-2500] (short-wavelength infrared). This gives detailed information with regards to the targets of interest. Different neural autoencoders (AEs) have been developed to reduce inputs into different dimensions, from 1 to 15. Each of these encoded sets was used to train decision tree (DT) classifiers. The results are very promising, as they show that AEs performance encoded data with correlation coefficients above 0.95. The classifiers trained with the different sets provide accuracies close to 1. dc.description: open access article This work has been conducted in collaboration with the University of Cadiz, when Maria Gema Carrasco-Garcia, a PhD student at the University of Cadiz came as a visiting student to work with Dr Lipika Deka and Professor David Elizondo. The funding has come from University of Cadiz, Spain and the projects of our collaborators.
Characterising Payload Entropy in Packet Flows
dc.title: Characterising Payload Entropy in Packet Flows dc.contributor.author: Kenyon, Anthony; Deka, Lipika; Elizondo, David dc.description.abstract: Accurate and timely detection of cyber threats is critical to keeping our online economy and data safe. A key technique in early detection is the classification of unusual patterns of network behaviour, often hidden as low-frequency events within complex time-series packet flows. One of the ways in which such anomalies can be detected is to analyse the information entropy of the payload within individual packets, since changes in entropy can often indicate suspicious activity - such as whether session encryption has been compromised, or whether a plaintext channel has been co-opted as a covert channel. To decide whether activity is anomalous we need to compare real-time entropy values with baseline values, and while the analysis of entropy in packet data is not particularly new, to the best of our knowledge there are no published baselines for payload entropy across common network services. We offer two contributions: 1) We analyse several large packet datasets to establish baseline payload information entropy values for common network services, 2) We describe an efficient method for engineering entropy metrics when performing flow recovery from live or offline packet data, which can be expressed within feature subsets for subsequent analysis and machine learning applications.
2022 Index IEEE Transactions on Artificial Intelligence Vol. 3
dc.title: 2022 Index IEEE Transactions on Artificial Intelligence Vol. 3 dc.contributor.author: Aafaq, N.; Elizondo, David
Improved Flow Recovery from Packet Data
dc.title: Improved Flow Recovery from Packet Data dc.contributor.author: Kenyon, Anthony; Elizondo, David; Deka, Lipika dc.description.abstract: Typical event datasets such as those used in network intrusion detection comprise hundreds of thousands, sometimes millions, of discrete packet events. These datasets tend to be high dimensional, stateful, and time-series in nature, holding complex local and temporal feature associations. Packet data can be abstracted into lower dimensional summary data, such as packet flow records, where some of the temporal complexities of packet data can be mitigated, and smaller well-engineered feature subsets can be created. This data can be invaluable as training data for machine learning and cyber threat detection techniques. Data can be collected in real-time, or from historical packet trace archives. In this paper we focus on how flow records and summary metadata can be extracted from packet data with high accuracy and robustness. We identify limitations in current methods, how they may impact datasets, and how these flaws may impact learning models. Finally, we propose methods to improve the state of the art and introduce proof of concept tools to support this work.
Neural network models for predicting flowering and physiological maturity of soybean
dc.title: Neural network models for predicting flowering and physiological maturity of soybean dc.contributor.author: Elizondo, David; McClendon, R. W.; Hoogenboom, G. dc.description.abstract: It is important for farmers to know when various plant development stages occur for making appropriate and timely crop management decisions. Although computer simulation models have been developed to simulate plant growth and development, these models have not always been very accurate in predicting plant development for a wide range of environmental conditions. The objective of this study was to develop a neural network model to predict flowering and physiological maturity for soybean (Glycine max L. Merr.). An artificial neural network is a computer software system consisting of various simple and highly interconnected processing elements similar to the neuron structure found in the human brain. A neural network model was used because it has the capabilities to identify relationships between variables of rather large and complex data bases. For this study, field-observed flowering dates for the cultivar Bragg from experimental studies conducted in Gainesville and Quincy, Florida, and Clayton, North Carolina, were used. Inputs considered for the neural network model were daily maximum and minimum air temperature, photoperiod, and days after planting or days after flowering. The data sets were split into training sets to develop the models and independent data sets to test the models. The average relative error of the test data sets for date of flowering prediction was+0.143 days (n = 21, R2 = 0.987) and for date of physiological maturity prediction was +2.19 days (n = 21, R2 = 0.950). It can be concluded from this study that the use of neural network models to predict flowering and physiological maturity dates is promising and needs to be explored further.
A permutation entropy-based EMD--ANN forecasting ensemble approach for wind speed prediction
dc.title: A permutation entropy-based EMD--ANN forecasting ensemble approach for wind speed prediction dc.contributor.author: Ruiz-Aguilar, Juan Jesus; Turias, Ignacio; Gonzalez-Enrique, Javier; Urda, Daniel; Elizondo, David dc.description.abstract: Accurate wind speed prediction is critical for many tasks, especially for air pollution modelling. Data-driven approaches are particularly interesting but the stochastic nature of wind renders prediction tasks difficult. Therefore, a combination of methods could be useful to obtain better results. To overcome this difficulty, a hybrid wind speed forecasting approach is proposed in this work. The Bay of Algeciras, Spain, was used as a case study, and the database was collected from a weather monitoring station. The study consists of combining a pre-processing method, the empirical mode decomposition (EMD), an information-based method, the permutation entropy (PE), and a machine learning technique (artificial neural networks, ANNs), using an ensemble learning methodology. Different prediction horizons were considered: ph-hours (ph = 1, 2, 8, 24) ahead and 8-h and 24-h average. The introduction of PE significantly reduces the computational cost and the predictive risk in comparison with traditional EMD methodology, by reducing the number of the decomposed components to be predicted. Moreover, the experimental results demonstrated that the EMD–PE–ANN approach outperforms the prediction performance of the single ANN models in all the prediction horizons tested. The EMD–PE–ANN model is capable to achieve a correlation coefficient of 0.981 and 0.807 for short-term (1 h) and medium-term (24 h) predictions, respectively, significantly overcoming those obtained by a single ANN model (0.929 and 0.503). These results show that the proposed model reaches significant improvements when the prediction horizon increases, where forecasting models tend to worsen their prediction performance. Therefore, the proposed EMD–PE–ANN approach may become a powerful tool for wind speed forecasting.

Click here to view a full listing of David Elizondo's publications and outputs.

Research interests/expertise

My research interests include both work in the theory and application of Neural Networks. Application areas include transport related problems that led to the development of DIGITS (iTRAQ project).

Areas of teaching

Artificial Neural Networks and Prolog programming.

Qualifications

French Qualification: University Full Professor Qualification by the Conseille National des Universites (CNU). Artificial Neural Networks, Theory and Applications - 2008.
French Qualification: Senior Lecturer/Principal Lecturer (Maitre de Conferences) Qualification by the Conseille National des Universites (CNU) - 2003.
PhD in Computer Science from the University Louis Pasteur, Strasbourg, France and IDIAP, Martigny, Switzerland. The Recursive Deterministic Perceptron and some Strategies for Topology Reduction on Neural Networks -1998.
DEA in Computer Science from the University of Montpellier, Montpellier, France, Application of Neural Networks to a control process in a dynamic environment - 1993.
Master of Science in Artificial Intelligence from the University of Georgia, Athens, Georgia, USA, Neural Network Models to Predict Solar Radiation and Plant Phenology - 1992.
Bachelor of Science in Computer Science from Knox College, Galesburg, Illinois, USA - 1986.

Courses taught

Artificial Neural Networks and Prolog programming.

Membership of external committees

Workshop Organizer for The British Computer Society Specialist Group on Artificial Intelligence (SGAI) International Conference in Artificial Intelligence for 2010.
UK Computational Intelligence workshop (UKCI).
IEEE International Conference in Artificial Neural Networks (2004,2005, 2006,2007, 2008, 2009).

Membership of professional associations and societies

IEEE Senior Member.

Conference attendance

Organiser and chairman of the following special conference sessions:

IEEE-WCCI-2012, Brisbane, Australia. Special session on Computational Intelligence for Privacy. (http://www.ieee-wcci2012.org/)
IEEE-WCCI-2010, Barcelona, Spain. Special session on Computational Intelligence for Privacy, Security, Forensics. (http://www.wcci2010.org/)
IEEE-ICANN-2008 Prague, Czech Republic. Special session on Constructive Neural Network Algorithms (http://www.icann2008.org/ssession.php). Contacted by Springer to produce a book of extended versions of these papers. The book will be published by January 2009.). Contacted by Springer to produce a book of extended versions of these papers. The book will be published by January 2009.
IEEE-ICANN-2005 Warsaw, Poland. Special session on Knowledge Extraction (http://www.ibspan.waw.pl/ICANN-2005/SpecialSession9.pdf)

National Conference Chairman

Programme Chair Workshop on Computational Intelligence (UKCI), De Montfort University, Leicester, Sept 10-12 2008 (http://www.cci.dmu.ac.uk/ukci2008/).

Consultancy work

Large International Banana producer Company. Banana hand cut optimization using Artificial Intelligence Techniques.

Current research students

2010-2013 John North. Associating Cause and Effect: Applying Computational Intelligence to Post-Incident Security Data. De Montfort University, Symantec.

2010-2014 Harold Kimball. Adaptive Security for Mobile Devices.

2013-2016 Simon Witheridge. Integrated Traffic Management and Air Quality Control.

Externally funded research grants information

“TITLE”, SPONSOR	ROLE	AMOUNT	PERIOD
“Banana Hand cut optimization using Computational Intelligence Techniques”, Chiquita Brands International Inc., USA.	PI	£12000	June 2010
“Travel Grant, WCCI-2010, Barcelona, Spain”, Royal Academy of Engineering.	PI	£600	2010
“Dynamic Traffic Management and Passenger Guidance to Meet the Carbon Challenge”, Transport iNet HECF.	PI	£45K	2009−2010
“Travel Grant, IJCNN-2009, Atlanta, Georgia”, Royal Academy of Engineering.	PI	£800	2009
“Travel Grant, ICANN-2008, Prague, Czek Republic”, Royal Academy of Engineering.	PI	£800	2008
“Travel Grant, ICANN-2007, Porto Portugal”, Royal Academy of Engineering.	PI	£800	2007
“Design of constructive methods on neural computing systems and its application to data mining in oncology”, Spanish Research Council.	CI	£225K	2008−2012
“New strategies in the design of neurocomputing systems. Application to the process of oncology data”, Spanish Research Council.	CI	£90K	2008−2010
“Integrated Traffic Management and Air Quality Control Using Downstream Space Services”, European Space Agency.	PI	e500K (£160K for DMU)	2011
“Innovation Fellowship with the School of Pharmacy”, EMDA, UK.	PI	£15K	2011

Internally funded research project information

“TITLE”, SPONSOR	ROLE	AMOUNT	PERIOD
“Associating Cause and Effect: Applying Computational Intelligence to Post-Incident Security Data”, DMU Research Scholarship, DMU, UK, Symantec, UK.	PI	£50K	2011−2014
“Intelligent Transport Systems: Integrated Traffic Management Control”, DMU Research Scholarship, DMU, UK.	CI	£50K	2012−2015
“De Montfort Interest Group in Transport Systems (DIGITS)”, DMU RIF.	CI	£10K	Jan−Apr 2012

Professional esteem indicators

Associate editor for the IEEE Transactions on Neural Networks and Learning Systems Journal (2.95 Impact Factor and in position 12 out of 111 according to the impact factor in the area of Artificial Intelligence)
Reviewer of European FP7 research projects (2009)
Referee for the Swiss National Science Foundation (2010)
Industrial Liaison for the IEEE Computational Intelligence Society (CSI), UKRI Chapter
Workshop Organizer for The British Computer Society Specialist Group on Artificial Intelligence (SGAI)
International Conference in Artificial Intelligence for 2010
Senior Member of the IEEE
Industrial Liaison for the IEEE Computational Intelligence Society (CSI), UKRI Chapter.