SANTTUcurriculum vitae
.01

ABOUT

PERSONAL DETAILS
Hounslow, UK
mapiconimg
tsantosh7@gmail.com
07914*****7
Hello. I am a Data Scientist and a Programmer. I am passionate about coding, cooking as well as driving motor bikes. Welcome to my Personal and Academic profile Available as freelance

BIO

ABOUT ME

I'm an experienced Data Scientist with a PhD in AI / machine learning, with 8+ years background in predictive analytics, data driven modelling, data visualisation, multivariate data analysis, feature extraction, natural language processing (NLP), computer vision (CV), software / web development and cloud computing.

My present work at Synoptica focuses on developing NLP based novel software-as-a-service AI tools , which enable sales and marketing teams to save their time and sell their products better and faster.

Previously, I was a Research Fellow in "machine learning for healthcare" at the University of Surrey (MRC) - developing ML based AI algorithms and software aimed at improving understanding of chronic diseases such as diabetes and Chronic Kidney Disease (CKD). These models, validated by expert clinicians, give doctors access to actionable predictions about their patients’ chronic conditions. For this work, I was recognised with the CogX 2018 award

In general, I have a broad experience of machine learning, AI and data science across a variety of domains as well as significant leadership and mentoring experience in these areas. Furthermore, I have commercial software engineering experience of putting prototypes into production, liaising with dev and front end teams and just making it work.

PhD

INTERESTS

My PhD was about machine learning, signal processing and computer vision, and was fully sponsored from the University of Surrey (£87,000). During my PhD, I was recognised with three awards including the "Excellent Oral Presentation" from IEEE ICIVC , news articles on the same can be found here and I also won the "Best Research Potential" and "Outstanding Service" awards.

Skills

FACTS ABOUT ME

I have significant experience in Python, MATLAB, Apache Solr/Lucene, SQL, Linux, GIT, Agile, AWS and web/software application development. I also write grant proposals and research articles https://goo.gl/5gdZ4U.

.02

RESUME

EDUCATION
  • 2014
    2016
    Surrey, United Kingdom

    Computer Science - PhD

    University of Surrey

    Graduated with Doctor of Philosophy in Computer Science from University of Surrey, UK. Topics of research were Signal Processing, Machine Learning and Computer Vision. Attended several international conferences. Served as a Session Chair at SSCI CICARE 2017. Received Full Studentship Grant, from the University of Surrey (England) for PhD in Computer Science.
  • 2010
    2012
    Espoo, Finland

    Computer Science - MSc(Tech.)

    Aalto University - School of Science

    Graduated with Master of Science in Technology (Msc.(Tech.)) from Aalto University, School of Science and Technology, Finland. Majored in Machine Learning and Data Mining (MACADAMIA), minored in Computational Fluid Dynamics (CFD). Master Thesis Intership at Kotka Maritime Research Centre. Received Research Funding from University of Aalto (2011-2012), (WP1) Future Combustion Engine Power (FCEP)- Wartsila, Finland Oy (2011) and Kotka Maritime Research Center (KMRC) - CAFE Project (2012). Received Scholar Ship from Merenkulun saatio (2012) and a Travel Grant, from Department of Marine Technology - Aalto University, School of Engineering for attending a conference.
  • 2005
    2009
    Hyderabad, India

    Computer Science - B. Tech

    Jawaharlal Nehru Technological University

    Graduated with Bachelor of Technology in Computer Science and Engineering (B.Tech (CSE)) from Jawaharlal Nehru Technological University, Hyderabad, India. Web Development Tutor and Undergraduate Student Mentor. Developed Malware Doctor v1.0 for detecting malware in the University computers. Received 1st prize in student project displays at CIENCIA 2K8. Developed a game called "One Way Out" which was played by more than 200 users at CIENCIA 2K8.
ACADEMIC AND PROFESSIONAL POSITIONS
  • Jul-17
    Present
    Guildford, United Kingdom

    Data Scientist

    Synoptica

    Work involves developing artificial intelligence algorithms using state-of-the-art machine learning techniques that help B2B companies automate research and prioritise leads based on company indicators made up of deep web and proprietary data sources. This enables sales and marketing teams to save time and sell faster. Moreover, I contributed to a proposal, that attracted £50,000 grant from Innovate UK to develop an AI based web application using Machine Learning for effectively managing their data.

    Programming: Python 3, Keras, TensorFlow. Machine/Deep learning: SVMs, Kernels, Similarity Metrics, Deep Auto-encoders, Marginalised Stack Denoising Auto-encoders, CNNs.

  • Dec-17
    Present
    Middlesex, United Kingdom

    Visiting Researcher

    Middlesex University

    Supervising PhD students.
  • Oct-15
    Dec-17
    Surrey, United Kingdom

    Research Fellow (Data Science)

    University of Surrey

    My research focused on modelling biomedical changes over time with a view to forming actionable predictions using electronic medical records. I have been involved in developing machine learning algorithms to classify clinical time series trends and automatically identify acute events, along with algorithms to identify and automatically correct errors in medical records. As the biomedical measurements used are irregularly sampled, I also developed a method of re-sampling time series in order to enable them to be used when training a classifi er. This has involved working closely with a core set of computational and clinical researchers, but also forming ad-hoc working relationships with an extended group of external clinical advisers and contributors.

    Programming & Tools: Matlab Methods : Gaussian Process Regression, Probabilistic Broken-sticks Modelling, Bayesian Modelling, Auto-encoders and t-SNE Clustering.

  • Jan-14
    Dec-16
    Guildford, United Kingdom

    Doctoral Researcher (Data Science)

    University of Surrey

    My thesis work expanded the method of Dynamic Mode Decomposition (DMD) to be used for solving novel problems in the fields of Signal Processing, Machine Learning and Computer Vision. The results from my thesis presented DMD as a promising approach for applications that require feature extraction, including: (i) trends and noise from signals, (ii) micro-level texture descriptor from images, and (iii) coherent structures from image sequences/videos, as well as applications that require suppression of movements from dynamical spatio-temporal image sequences.
  • Jan-13
    Dec-13
    Helsinki, Finland

    Data Scientist

    LokalHouse

    Predictive analytics for real estate industry. Work involved politely crawling real estate websites to collect data and later convert to insights. Provide one search for all the real estate listings in India and business analytics to the realtors and big investors.

    Tools and functions utilised were: Python - Scrappy, Django framework. Machine learning Methods utilised were Multilinear Regression Analysis and Exponential Smoothing Techniques.

  • Dec-10
    Dec-12
    Espoo, Finland

    Research Assistant (Data Science)

    Aalto University School of Science

    From Oct 2010 - May 2011. I worked at Information and Computer Science lab. Work involved document clustering using different similarity metrics over different dimensionality reduction methods. Methods used were Principle Componant Analysis (PCA), Singular Value Decomposition (SVD) and K-means Clustering. Programing languages used were Python, Java, shell scripting and Matlab.

    From June 2011- May 2012, I worked as an intern at Department of Energy Technology. Work involved extracting coherent structures using machine learning algorithms in turbulent jets and sprays. Methods used were Proper Orthogonal Decomposition (POD), Image Analysis, Dynamic Mode Decomposition (DMD) and programmed in Matlab.

    From Jun 2012- Dec 2012, I worked at Department of Marine Technology. Work involved extraction of causal relations, human and organisational factors present in the accident investigation reports. I used Text Mining, Natural Language Processing (NLP), Information Extraction (IE), Information Retrieval (IR), Regular Expression Grammar, Named Entity Recognition (NER). Parts of Speech (PoS) Tagging. Parsing, Chuncking, Chinking, Binary Classification. Methods used for classification were Support Vector Machines (SVM), Naive Bayes Classifier. Tools and Languages used were Natural Language Tool Kit (NLTK), Apple Pie Parser, Python, Core-Java, and Matlab.

  • Mar-10
    Aug-10
    Hyderabad, India

    Specialist GIS

    Rofous Software Ltd

    Manual Development and Quality Testing of Google Maps of European countries including Belgium, Netherlands and Finland. I was working at Google India Pvt Ltd. representing Rofous S/W Ltd.
  • May-09
    Mar-10
    Hyderabad, India

    Web Programmer

    Medecode Solutions

    Worked on Joomla, WordPress extensions development and content management system for developing web portals. Programming languages used were PHP and MySQL.
.03

PUBLICATIONS

PUBLICATIONS LIST
06 Aug 2013

Data Mining of Causal Relations from Text: Analysing Maritime Accident Investigation Reports

Master Thesis, Aalto University


Machine Learning Santosh Tirunagari

Data Mining of Causal Relations from Text: Analysing Maritime Accident Investigation Reports

Santosh Tirunagari
Machine Learning
About The Publication

Text mining is a process of extracting information of interest from text. Such a method includes techniques from various areas such as Information Retrieval (IR), Natural Language Processing (NLP), and Information Extraction (IE). In this study, text mining methods are applied to extract causal relations from maritime accident investigation reports collected from the Marine Accident Investigation Branch (MAIB). These causal relations provide information on various mechanisms behind accidents, including human and organizational factors relating to the accident. The objective of this study is to facilitate the analysis of the maritime accident investigation reports, by means of extracting contributory causes with more feasibility. A careful investigation of contributory causes from the reports provide opportunity to improve safety in future.

Two methods have been employed in this study to extract the causal relations. They are 1) Pattern classification method and 2) Connectives method. The earlier one uses naive Bayes and Support Vector Machines (SVM) as classifiers. The latter simply searches for the words connecting cause and effect in sentences.

06 Dec 2016

Local Binary Patterns as a Feature Descriptor in Alignment-Free Visualisation of Metagenomic Data

IEEE CIBCB 2017 Proceedings


HealthcareMachine Learning S Kouchaki, S Tirunagari, A Tapinos, DL Robertson

Local Binary Patterns as a Feature Descriptor in Alignment-Free Visualisation of Metagenomic Data

S Kouchaki, S Tirunagari, A Tapinos, DL Robertson
HealthcareMachine Learning
About The Publication

Shotgun sequencing has facilitated the analysis of complex microbial communities. However, clustering and visualising these communities without prior taxonomic information is a major challenge. Feature descriptor methods can be utilised to extract these taxonomic relations from the data. Here, we present a novel approach consisting of local binary patterns (LBP) coupled with randomised singular value decomposition (RSVD) and Barnes-Hut t-stochastic neighbor embedding (BH-tSNE) to highlight the underlying taxonomic structure of the metagenomic data. The effectiveness of our approach is demonstrated using several simulated and a real metagenomic datasets.

20 Dec 2016

Dynamic mode decomposition for computer vision and signal processing

PhD Thesis, University of Surrey


BiometricsHealthcareMachine Learning Santosh Tirunagari

Dynamic mode decomposition for computer vision and signal processing

Santosh Tirunagari
BiometricsHealthcareMachine Learning
About The Publication

The method of Dynamic Mode Decomposition (DMD) was introduced originally in the area of Computatational Fluid Dynamics (CFD) for extracting coherent structures from spatio-temporal complex fluid flow data. DMD takes in time series data and computes a set of modes, each of which is associated with a complex eigenvalue. DMD analysis is closely associated with spectral analysis of the Koopman operator, which provides linear but infinite-dimensional representation of nonlinear dynamical systems. Therefore, by using DMD a nonlinear system could be described by a superposition of modes whose dynamics are governed by the eigenvalues. The key advantage of DMD is its data-driven nature which does not rely on any prior assumptions except the inherent dynamics which are observed over time. Its capability for extracting relevant modes from complex fluid flows has seen significant application across …

17 May 2016

Classification of irregularly sampled time series with unequal lengths

IEEE MLSP


HealthcareMachine Learning S Tirunagari, S.Bull, N.Poh

Classification of irregularly sampled time series with unequal lengths

S Tirunagari, S.Bull, N.Poh
HealthcareMachine Learning
About The Publication

A patient’s estimated glomerular filtration rate (eGFR) can provide important information about disease progression and kidney function. Traditionally, an eGFR time series is interpreted by a human expert labelling it as stable or unstable. While this approach works for individual patients, the time consuming nature of it precludes the quick evaluation of risk in large numbers of patients. However, automating this process poses significant challenges as eGFR measurements are usually recorded at irregular intervals and the series of measurements differs in length between patients. Here we present a two-tier system to automatically classify an eGFR trend. First, we model the time series using Gaussian process regression (GPR) to fill ingaps’ by resampling a fixed size vector of fifty time-dependent observations. Second, we classify the resampled eGFR time series using a K-NN/SVM classifier, and evaluate its performance via 5-fold cross validation. Using this approach we achieved an F-score of 0.90, compared to 0.96 for 5 human experts when scored amongst themselves.

01 Jan 1970

Can DMD obtain a Scene Background in Color?

Image, Vision and Computing (ICIVC),


Machine Learning S Tirunagari, N Poh, M Bober, D Windridge

Can DMD obtain a Scene Background in Color?

S Tirunagari, N Poh, M Bober, D Windridge
Machine Learning
About The Publication

A background model describes a scene without any foreground objects and has a number of applications, ranging from video surveillance to computational photography. Recent studies have introduced the method of Dynamic Mode Decomposition (DMD) for robustly separating video frames into a background model and foreground components. While the method introduced operates by converting color images to grayscale, we in this study propose a technique to obtain the background model in the color domain. The effectiveness of our technique is demonstrated using a publicly available Scene Background Initialisation (SBI) dataset. Our results both qualitatively and quantitatively show that DMD can successfully obtain a colored background model.

14 Sep 2016

” Flow Size Difference” Can Make a Difference: Detecting Malicious TCP Network Flows Based on Benford’s Law

IEEE Transactions on Information Forensics and Security


Security A Iorliam, S Tirunagari, ATS Ho, S Li, A Waller, N Poh

” Flow Size Difference” Can Make a Difference: Detecting Malicious TCP Network Flows Based on Benford’s Law

A Iorliam, S Tirunagari, ATS Ho, S Li, A Waller, N Poh
Security
About The Publication

Statistical characteristics of network traffic have attracted a significant amount of research for automated network intrusion detection, some of which looked at applications of natural statistical laws such as Zipf’s law, Benford’s law and the Pareto distribution. In this paper, we present the application of Benford’s law to a new network flow metric” flow size difference”, which have not been studied before by other researchers, to build an unsupervised flow-based intrusion detection system (IDS). The method was inspired by our observation on a large number of TCP flow datasets where normal flows tend to follow Benford’s law closely but malicious flows tend to deviate significantly from it. The proposed IDS is unsupervised, so it can be easily deployed without any training. It has two simple operational parameters with a clear semantic meaning, allowing the IDS operator to set and adapt their values intuitively to adjust the overall performance of the IDS. We tested the proposed IDS on two (one closed and one public) datasets, and proved its efficiency in terms of AUC (area under the ROC curve). Our work showed the” flow size difference” has a great potential to improve the performance of any flow-based network IDSs.

06 Dec 2016

Visualisation of Survey Responses using Self-Organising Maps: A Case Study on Diabetes Self-care Factors

IEEE Computational Intelligence (SSCI)


Healthcare S Tirunagari, S Bull, S Kouchaki, D Cooke, N Poh

Visualisation of Survey Responses using Self-Organising Maps: A Case Study on Diabetes Self-care Factors

S Tirunagari, S Bull, S Kouchaki, D Cooke, N Poh
Healthcare
About The Publication

Due to the chronic nature of diabetes, patient self-care factors play an important role in any treatment plan. In order to understand the behaviour of patients in response to medical advice on self-care, clinicians often conduct cross-sectional surveys. When analysing the survey data, statistical machine learning methods can potentially provide additional insight into the data either through deeper understanding of the patterns present or making information available to clinicians in an intuitive manner. In this study, we use self-organising maps (SOMs) to visualise the responses of patients who share similar responses to survey questions, with the goal of helping clinicians understand how patients are managing their treatment and where action should be taken. The principle behavioural patterns revealed through this are that: patients who take the correct dose of insulin also tend to take their injections at the correct time …

06 Dec 2016

Automatic detection of acute kidney injury episodes from primary care data

IEEE Computational Intelligence (SSCI)


HealthcareMachine Learning S Tirunagari, SC Bull, A Vehtari, C Farmer

Automatic detection of acute kidney injury episodes from primary care data

S Tirunagari, SC Bull, A Vehtari, C Farmer
HealthcareMachine Learning
About The Publication

Acute kidney injury (AKI) is characterised by a rapid deterioration in kidney function, and can be identified by examining the rate of change in a patient’s estimated glomerular filtration rate (eGFR) signal. Due to the potentially irreversible nature of the damage AKI episodes cause to renal function, their detection plays a significant role in predicting a kidney’s effectiveness. Although algorithms for the detection of AKI are available for patients under constant monitoring, e.g. inpatients, their applicability to primary care settings is less clear as the eGFR signal often contains large lapses in time between measurements. However, waiting for hospital admittance before AKI is undesirable, as detecting AKI early can help to mitigate the degradation of kidney function and the associated increase in morbidity and mortality. Traditionally, a clinician in a primary care setting would manually identify AKI episodes from direct …

09 Feb 2017

Dynamic Mode Decomposition for Univariate Time Series: Analysing Trends and Forecasting

HAL Archive


Machine Learning S Tirunagari, S Kouchaki, N Poh, M Bober

Dynamic Mode Decomposition for Univariate Time Series: Analysing Trends and Forecasting

S Tirunagari, S Kouchaki, N Poh, M Bober
Machine Learning
About The Publication

This study introduces the method of Dynamic Mode Decomposition (DMD) for analysing univariate time series by forecasting as well as extracting trends and frequencies. The key advantage of DMD is its data-driven nature which does not rely on any prior assumptions (like Singular Spectrum Analysis (SSA)) except the inherent dynamics which are captured over time. Indeed, this study will show that the DMD eigenvalues with frequencies that are closer to the origin in the complex plane capture the trends in the time series. Moreover, the temporal evolution of the DMD modes, which is preserved via the Vandermonde matrix, can be used to reconstruct the desired components and perform forecasting at the same time. The results at various noise levels on simulated data suggests that DMD is a promising approach to modelling a time series with a noisy structure. Although these properties are not new in the DMD literature, the novel contributions of this paper are in making the method of DMD work for univariate time series through a four staged pipeline. Thus, this is the first work that shows DMD can be used for modelling, predicting, and forecasting a univariate time series.

01 May 2017

Movement correction in DCE-MRI through windowed and reconstruction dynamic mode decomposition

Machine Vision and Applications


HealthcareMachine Learning S Tirunagari, N Poh, K Wells, M Bober, I Gorden

Movement correction in DCE-MRI through windowed and reconstruction dynamic mode decomposition

S Tirunagari, N Poh, K Wells, M Bober, I Gorden
HealthcareMachine Learning
About The Publication

Images of the kidneys using dynamic contrast-enhanced magnetic resonance renography (DCE-MRR) contains unwanted complex organ motion due to respiration. This gives rise to motion artefacts that hinder the clinical assessment of kidney function. However, due to the rapid change in contrast agent within the DCE-MR image sequence, commonly used intensity-based image registration techniques are likely to fail. While semi-automated approaches involving human experts are a possible alternative, they pose significant drawbacks including inter-observer variability, and the bottleneck introduced through manual inspection of the multiplicity of images produced during a DCE-MRR study. To address this issue, we present a novel automated, registration-free movement correction approach based on windowed and reconstruction variants of dynamic mode decomposition (WR-DMD). Our proposed method …

23 Aug 2017

One dimensional local binary patterns of electroencephalogram signals for detecting Alzheimer’s disease

Digital Signal Processing (DSP)


HealthcareMachine Learning S Tirunagari, S Kouchaki, D Abasolo, N Poh

One dimensional local binary patterns of electroencephalogram signals for detecting Alzheimer’s disease

S Tirunagari, S Kouchaki, D Abasolo, N Poh
HealthcareMachine Learning
About The Publication

Alzheimer’s disease (AD) is neurodegenerative, caused by the progressive death of brain cells over time. One non-invasive approach to investigate AD is to use electroen-cephalogram (EEG) signals. The data are usually non-stationary with a strong background activity and noise which makes the analysis difficult leading to low performance in many real world applications including the detection of AD. In this study, we present a method based on local texture changes of EEG signals to differentiate AD patients from the healthy ones, using one-dimensional local binary patterns (1D-LBPs) coupled with support vector machines (SVM). Our proposed method maps the EEG data into a less detailed representation which is less sensitive to noise. A 10 fold cross validation performed at both the epoch and subject level show the discriminancy power of 1D-LBP feature vectors with application to AD data.

23 Aug 2017

Marginalised stack denoising autoencoders for metagenomic data binning

IEEE CIBCB 2017 Proceedings


HealthcareMachine Learning S Kouchaki, S Tirunagari, A Tapinos, DL Robertson

Marginalised stack denoising autoencoders for metagenomic data binning

S Kouchaki, S Tirunagari, A Tapinos, DL Robertson
HealthcareMachine Learning
About The Publication

Shotgun sequencing has facilitated the analysis of complex microbial communities. Recently we have shown how local binary patterns (LBP) from image processing can be used to analyse the sequenced samples. LBP codes represent the data in a sparse high dimensional space. To improve the performance of our pipeline, marginalised stacked autoencoders are used here to learn frequent LBP codes and map the high dimensional space to a lower dimension dense space. We demonstrate its performance using both low and high complexity simulated metagenomic data and compare the performance of our method with several existing techniques including principal component analysis (PCA) in the dimension reduction step and fc-mer frequency in feature extraction step.

01 Dec 2017

Probabilistic broken-stick model: A regression algorithm for irregularly sampled data with application to eGFR

Journal of biomedical informatics


HealthcareMachine Learning N Poh, S Tirunagari, N Cole, S de Lusignan

Probabilistic broken-stick model: A regression algorithm for irregularly sampled data with application to eGFR

N Poh, S Tirunagari, N Cole, S de Lusignan
HealthcareMachine Learning
About The Publication

In order for clinicians to manage disease progression and make effective decisions about drug dosage, treatment regimens or scheduling follow up appointments, it is necessary to be able to identify both short and long-term trends in repeated biomedical measurements. However, this is complicated by the fact that these measurements are irregularly sampled and influenced by both genuine physiological changes and external factors. In their current forms, existing regression algorithms often do not fulfil all of a clinician’s requirements for identifying short-term (acute) events while still being able to identify long-term, chronic, trends in disease progression. Therefore, in order to balance both short term interpretability and long term flexibility, an extension to broken-stick regression models is proposed in order to make them more suitable for modelling clinical time series. The proposed probabilistic broken-stick model can robustly estimate both short-term and long-term trends simultaneously, while also accommodating the unequal length and irregularly sampled nature of clinical time series. Moreover, since the model is parametric and completely generative, its first derivative provides a long-term non-linear estimate of the annual rate of change in the measurements more reliably than linear regression. The benefits of the proposed model are illustrated using estimated glomerular filtration rate as a case study used to manage patients with chronic kidney disease.

12 Dec 2012

Mining Causal Relations and Concepts in Maritime Accidents Investigation Reports

International Journal of Innovative Research and Development


Machine Learning S Tirunagari, M Hänninen, K Ståhlberg, P Kujala

Mining Causal Relations and Concepts in Maritime Accidents Investigation Reports

S Tirunagari, M Hänninen, K Ståhlberg, P Kujala
Machine Learning
About The Publication

Text mining is a process of extracting information of interest, from the text. In here, we applied text mining methods to extract causal patterns from the maritime accident reports collected from the Marine Accident Investigation Branch (MAIB). These causal patterns from the accident reports provide information on various mechanisms behind accidents. These include human and organisational concepts. A careful and manual investigation of causal patterns extracted from the reports provided opportunity to collect a list of concepts present in an accident according to the investigation. In this paper we discuss the statistics of the accidents that are caused by the list of concepts that were collected in this research work and also apply Self Organising Maps for visualization.

21 Nov 2011

Effect of dimensionality reduction on different distance measures in document clustering

ICONIP 2011, Neural Information Processing


Machine Learning MS Paukkeri, I Kivimäki, S Tirunagari, E Oja, T Honkela

Effect of dimensionality reduction on different distance measures in document clustering

MS Paukkeri, I Kivimäki, S Tirunagari, E Oja, T Honkela
Machine Learning
About The Publication

In document clustering, semantically similar documents are grouped together. The dimensionality of document collections is often very large, thousands or tens of thousands of terms. Thus, it is common to reduce the original dimensionality before clustering for computational reasons. Cosine distance is widely seen as the best choice for measuring the distances between documents in k-means clustering. In this paper, we experiment three dimensionality reduction methods with a selection of distance measures and show that after dimensionality reduction into small target dimensionalities, such as 10 or below, the superiority of cosine measure does not hold anymore. Also, for small dimensionalities, PCA dimensionality reduction method performs better than SVD. We also show how l 2 normalization affects different distance measures. The experiments are run for three document sets in English and one in Hindi.

17 Nov 2018

The 1st Competition on Counter Measures to Finger Vein Spoofing Attacks

IEEE ICB


Biometrics P Tome, R Raghavendra, C Busch, S Tirunagari, N Poh

The 1st Competition on Counter Measures to Finger Vein Spoofing Attacks

P Tome, R Raghavendra, C Busch, S Tirunagari, N Poh
Biometrics
About The Publication

The vulnerability of finger vein recognition to spoofing attacks has emerged as a crucial security problem in the recent years mainly due to the high security applications where biometric technology is used. Recent works shown that finger vein biometrics is vulnerable to spoofing attacks, pointing out the importance to investigate counter-measures against this type of fraudulent actions. The goal of the 1st Competition on Counter Measures to Finger Vein Spoofing Attacks is to challenge researchers to create countermeasures that can detect printed attacks effectively. The submitted approaches are evaluated on the Spoofing-Attack Finger Vein Database and the achieved results are presented in this paper.

24 Feb 2015

Detection of Face Spoofing Using Visual Dynamics

IEEE Transactions on Information Forensics and Security


Biometrics Santosh Tirunagari ; Norman Poh ; David Windridge ; Aamo Iorliam ; Nik Suki ; Anthony T. S. Ho

Detection of Face Spoofing Using Visual Dynamics

Santosh Tirunagari ; Norman Poh ; David Windridge ; Aamo Iorliam ; Nik Suki ; Anthony T. S. Ho
Biometrics
About The Publication

Rendering a face recognition system robust is vital in order to safeguard it against spoof attacks carried out using printed pictures of a victim (also known as print attack) or a replayed video of the person (replay attack). A key property in distinguishing a live, valid access from printed media or replayed videos is by exploiting the information dynamics of the video content, such as blinking eyes, moving lips, and facial dynamics. We advance the state of the art in facial antispoofing by applying a recently developed algorithm called dynamic mode decomposition (DMD) as a general purpose, entirely data-driven approach to capture the above liveness cues. We propose a classification pipeline consisting of DMD, local binary patterns (LBPs), and support vector machines (SVMs) with a histogram intersection kernel. A unique property of DMD is its ability to conveniently represent the temporal information of the entire video as a single image with the same dimensions as those images contained in the video. The pipeline of DMD + LBP + SVM proves to be efficient, convenient to use, and effective. In fact only the spatial configuration for LBP needs to be tuned. The effectiveness of the methodology was demonstrated using three publicly available databases: (1) print-attack; (2) replay-attack; and (3) CASIA-FASD, attaining comparable results with the state of the art, following the respective published experimental protocols.

07 Jan 2013

Large-eddy Simulation of Highly Underexpanded Transient Gas Jets

Physics of Fluids


Computational Fluid Dynamics V Vuorinen, J Yu, S Tirunagari, O Kaario, M Larmi

Large-eddy Simulation of Highly Underexpanded Transient Gas Jets

V Vuorinen, J Yu, S Tirunagari, O Kaario, M Larmi
Computational Fluid Dynamics
About The Publication

Large-eddy simulations (LES) based on scale-selective implicit filtering are carried out in order to study the effect of nozzle pressure ratios on the characteristics of highly underexpanded jets. Pressure ratios ranging from 4.5 to 8.5 with Reynolds numbers of the order 75 140 are considered. The studied configuration agrees well with the classical picture of the structure of highly underexpanded jets. Similarities and differences between simulation and experiments are discussed by comparing the concentration field structures from LES and planar laser induced fluorescence data. The transient stages, leading eventually to the highly underexpanded state, are visualized and investigated in terms of a phase diagram revealing the shock speeds and duration of the transient stages. For the studied nozzle pressure ratio range, the Mach disk dimensions are found to be in good agreement with literature data and experimental observations. It is observed how the nozzle pressure ratio influences the Mach disk width, and thereby the slip line separation, which leads to co-annular jets with inner and outer shear layers at higher pressure ratios. The improved mixing with increasing pressure ratio is demonstrated by the probability density functions of the concentration. The coherent structures downstream of the Mach disk are identified using proper orthogonal decomposition (POD). The structures indicate a helical mode originating from the shear layers of the jet. Despite the relatively low energy content of the dominant POD modes, the frequencies of the POD time coefficients explain the dominant frequencies in the pressure fluctuation spectra.

21 Dec 2017

Using Benford’s Law to Detect Anomalies in Electroencephalogram: An Application to Detecting Alzheimer’s Disease

IEEE CIBCB 2017 Proceedings


Healthcare S Tirunagari, DE Abasolo, A Iorliam, A Ho, N Poh

Using Benford’s Law to Detect Anomalies in Electroencephalogram: An Application to Detecting Alzheimer’s Disease

S Tirunagari, DE Abasolo, A Iorliam, A Ho, N Poh
Healthcare
About The Publication

Alzheimer’s disease (AD) is a neurodegenerative disease caused by the progressive death of brain cells over time. It represents the most frequent cause of dementia in the western world, and affects an individual’s cognitive ability and psychological capacity. While clinical diagnoses of AD are made primarily on the basis of clinical evaluation and mental health tests, diagnostic certainty is only possible through necropsy. One non-invasive approach to investigating AD is to use electroencephalograms (EEGs), which reflect brain electrical activity and
so can be used to detect electrical abnormalities in brain signals with non-invasive cranial surface electrodes.

Generally EEGs in AD patients show a shift to lower frequencies in spectral analysis and display less complexity and contain more regular patterns compared to those of control subjects.

Here we present a method for differentiating AD patients from healthy ones based on their EEG signals using Benford’s law and support vector machines (SVMs) with a radial basis function (RBF) kernel. EEG signals from eleven AD and eleven age-matched controls were divided into artefact-free 5-sec epochs and used to train an SVM. 10 fold cross validation was performed at both the epoch and subject-level to evaluate the importance of each electrode in discriminating between AD and healthy subjects. Substantive variability was seen across the different electrodes, with electrodes O1, O2 and C4 particularly being important. Performance across the electrodes was reduced when subject-level cross validation was performed, but relative performance across the electrodes was consistent with that found using epoch-level cross validation.

 

 

.04

RESEARCH

Collaboration

University of Surrey

Department of Computer Science

Middlesex University

Department of Computer Science

Mangalore University

Department of Computer Science

RESEARCH PROJECTS

MRC CKD

Modelling and Predicting the Progression of Chronic Kidney Disease

An overarching objective of this MRC research is to revisit the problem of modelling the progression of Chronic Kidney Disease (CKD) using state-of-the-art machine learning techniques and methodologies. http://modellingckd.org/

BioVacSafe

Biomarkers for enhanced vaccines immuno safety

BioVacSafe - Biomarkers for enhanced vaccines immunosafety - is a 5 years project funded by Innovative Medicine initiative (IMI). The goal of BioVacSafe is to develop cutting edge tools to speed up and improve the testing and monitoring of vaccine safety, both before and after release to the market. By bringing together for the first time three of Europe's leading vaccine development and manufacture companies as well as top experts from academic institutions and small and medium-sized enterprises (SMEs), the project will ultimately accelerate the development and introduction of a new generation of safer, more effective vaccines. By sharing their expertise, the BioVacSafe partners have a unique opportunity to make progress in this important area. http://www.biovacsafe.eu/

CAFE Project

The Competitive Advantage by Safety

The Competitive Advantage by Safety (CAFE) project is coordinated by the Kotka Maritime Research Centre, Finland. The aim of the CAFE project is to examine can the maritime sector achieve competitive advantage by focusing on safety aspects. The major focus is on operational safety which is expected to both directly and indirectly influence the opportunities in the competitive European surface transport sector. http://www.merikotka.fi/cafe/

Volition

A machine learning decision support platform for Innovate UK

The Volition platform is a secure cloud based web application designed to help Innovate UK make better use of its existing operational data using machine learning (ML). It will be driven by a blend of ML capabilities, natural language processing (NLP), advanced information retrieval techniques and data visualisation. The Synoptica approach is unique in that we automatically find and extract additional data from the web from a variety of structured and unstructured sources and use this to enhance the ML analysis of Innovate UK’s existing data. https://www.synoptica.com/semantic-industry-classifier-sic/
.05

TEACHING

CURRENT
  • Oct 2018
    Present
    Middlesex, United Kingdom

    Guest Lecturer

    University of Middlesex

    Machine Learning for Natural Language Processing
TEACHING HISTORY
  • Jan 2014
    Dec 2017
    Surrey, United Kingdom

    teaching assistant

    UNIVERSITY OF surrey

    Image Processing Machine Learning for Business Analytics Databases and Knowlede Discovery Information Security
  • Jan 2014
    Ded 2017
    Surrey, United Kingdom

    lab demonstrator

    UNIVERSITY OF surrey

    Machine Learning Databases & Knowledge Discovery Advanced Signal Processing
.06

SKILLS

Data Science Skills
Machine Learning Theory > 8+ years of solid theoritical machine learning experience
LEVEL : ADVANCED EXPERIENCE : 8 YEARS
Support Vector Machines Auto Encoders Recurrent Neural Networks Dynamic Mode Decomposition
Machine Learning Programming > 8+ years of solid practical machine learning experience
LEVEL : ADVANCED EXPERIENCE : 8 YEARS
MATLAB Natural Language Tool Kit TensorFlow Keras Word2Vec Doc2Vec Scikit Learn
PROGRAMMING SKIILLS
Web development > I have been doing web development on and off over a period of 12 years. Most of the time, I use wordpress, joomla or html5 templates to kick start the process.
LEVEL : INTERMEDIATE EXPERIENCE : 12 YEARS
Php Python MySQL HTML CSS
Software as a Service > Flask is a great web mirco-framework, that is best utilised with event-loop concurrency
LEVEL : Expert EXPERIENCE : 4 YEARS
Flask server PyCharm Apache Solr Django
.07

WORKS

MY PORTFOLIO
Artificial Intelligence

Company Size Tool

Company Size Tool

Artificial Intelligence

Semantic Industry Classifier

Semantic Industry Classifier

Web Development

Nicsys solutions

Nicsys solutions

Web Development

IEEE ICASSP 2019

IEEE ICASSP 2019

Web Development

VarunRaj.studio

VarunRaj.studio

Web Development

Murali-Mushayira.com

Murali-Mushayira.com

Web Development

medecode solutions

medecode solutions

.08

AWARDS

HONORS AND AWARDS
  • 2018
    2018
    London, United Kingdom

    outstanding contribution to AI-Postdoc Research

    CognitionX

    The prestigious Award, is one awarded annually by the CogX to recognise best-of-the-best individuals/AI products/Companies who have made a significant contribution to their domain in the field of artificial intelligence across the world. https://www.surrey.ac.uk/news/postdoctoral-researcher-honoured-ai-research
  • Aug-16
    Portsmouth, United Kingdom

    Best Presentation Award

    IEEE

    Received an award for the best presentation at IEEE International Conference on Image, Vision and Computing.
  • May-15
    Surrey, United Kingdom

    Best Research Potential Award

    University of Surrey

    Received for presenting a talk on how to counter spoof fingervein attacks in Biometrics.
  • May-15
    Surrey, United Kingdom

    Outstanding Service Award

    University of Surrey

    Received for organising the 12th PhD computing conference.
  • Jan-13
    Espoo, Finland

    ACE Invention Award

    Aalto Center for Entrepreneurship (ACE)

    Received for introducing the concept of real estate price prediction algorithm as a web service.
  • Jan-14
    Surrey, United Kingdom

    Full Studentship Grant

    University of Surrey

    Received a full scholarship for pursuing PhD in Computer Science (£87,000 equivalent to INR 90,48,000 in 2014) from the Centre for Vision Speech and Signal Processing (CVSSP) and Department of Computer Science, University of Surrey, Guildford, UK.
  • Jan-12
    Espoo, Finland

    Travel Scholarship

    Merenkulun Saatio

    Received scholarship for travelling as well as for conducting research at Department of Energy Technology, Aalto University.
  • Mar-08
    Hyderabad, India

    First Prize Project Display

    CVR College of Engineering

    I won the first prize for inventing Malware doctor (anti-malware software) at the CIENCIA 2K8, a National Level Technical Symposium held on 14 and 15th March 2008 in India.
.09

CONTACT

Drop me a line

GET IN TOUCH

Happy to collaborate in many ways or any way. I have a good infra structure set up, with opportunities for collecting both survey and experimental data as well as for analysing it using advanced machine learning techniques. Get in touch if you interested, and we can discuss possible opportunities, including writing peer reviewed papers or starting spin offs.

  • Brainstorm sessions
  • Idea Generation
  • Artificial Intelligence
  • Research Proposals