Day :
- WORKSHOP
Session Introduction
Manoj Mishra
Union Insurance, UAE
Title: Emerging trends in machine learning
Time : 11:25 - 12:10
Biography:
A distinguished Technology leader driving transformation through innovation in Data Engineering, Data Management and Data Science. Holds a “Bachelor of Engineering” in Computer Science and a certification in “Data Science” from “Johns Hopkins University” Transformational leader, with a track record of developing solutions for future, creating benchmark of performance and success. More than two decade of experience spreading across multiple geographies (US, Europe, India and Middle East) working with Organizations like Adobe Systems, Dell, Perot Systems, CEB-Gartner, Rolta and Tata Group. Moved to Dubai in June 2017 to take a leadership role as “Chief Manager–Business Intelligence & Data” with Union Insurance and currently leading their Data Strategy and Technology Transformations through Data, Analytics, research & various AI initiatives.
Abstract:
In this time of immense machine learning skills shortage, Data and analytics leaders face a dilemma. Without data scientists, venturing into machine learning and data science is difficult. Without any successful pilots, convincing the business to hire data scientists is equally challenging. Auto ML is a concept started by H2o which has been recently gone to the next level by Google's Automated Machine Learning which lets you train custom machine learning models without having to code. My workshop will be focusing on these trends in a little more detail and also a demo on how these trends can be leveraged.
Break: Lunch Break 12:10-13:10
- Machine Learning | Artificial intelligence applications | Deep Learning | Deep Learning Frameworks | Data analytics in cloud | Pattern Recognition | Facial Expression and Emotion Detection | Artificial Neural Networks (ANN) & Chainer | Augmented Reality (AR) and Virtual Reality (VR)
Chair
Shabir Momin
ZengaTV, Singapore
Co-Chair
Santosh Godbole
SSN Solutions Limited, India
Session Introduction
Harshavardhana Kikkeri
Kaaya Tech Inc, USA
Title: Heralding the era of machine teaching – where you clothes become your virtual AI trainer
Time : 13:10 - 13:40
Biography:
Harsha Kikkeri is the CEO of Kaaya Tech Inc where he is building HoloSuit – An AI enabled full body analytics platform which acts as a virtual trainer for your body. He has over 18 years experience working on IoT, augmented/virtual reality, aerial and ground robots with expertise in drones, sensor fusion and machine learning. He did pioneering research at Microsoft Robotics in USA building robots which could learn by demonstration. He has won numerous leadership awards including Gold Star from Microsoft, Excellence Award from Infosys, Bharat Petroleum Scholarship and has won numerous chess tournaments. He has Masters in Electrical Engineering from Syracuse University, NY and BE Electronics from SJCE, Mysore, India. He holds 35+ international patents from US, Europe, China, Japan and other countries.
Abstract:
We are entering an era where a machine can learn almost anything and do it better than humans. Textiles are becoming smarter with embedded sensors which can track a large number of biomarkers. Health monitoring systems are getting connected to the cloud providing rich source of data. Virtual Reality experiences are becoming hyper-realistic enabling factory workers train virtually and perform remote diagnostics using augmented reality. Expensive defence simulators are getting replaced by virtual simulators at a fraction of the cost. Can we weave together all these trends and convert your clothing to be your personalized AI health and fitness coach? Can we move from the era of Machine Learning to the era of Machine teaching where machines become so smart they can analyze and correlate multiple data and provide you with real and virtual challenges to help you reach your true potential. Can we provide 24x7 real-time tracking and feedback to provide personalized recommendations to you based on objective data collected from similar demographics? This talk will explore the exciting possibilities that the era of Machine Teaching opens up by weaving together fashion, fitness, AI, AR/VR and IoT to create an AI trainer which can be used by both entreprise and consumer.
Santosh Godbole
SSN Solutions Limited, India
Title: Applying big data analytics and machine learning in precision marketing
Time : 13:40 - 14:10
Biography:
Santosh is the Co-Founder and Chief Product Officer at SSN Solutions Limited. At SSN, his role is to define product and technology roadmap. Prior to SSN, he was a Senior Director of Engineering at ARRIS (India). At ARRIS, he has managed a large team of engineers that was spread across multiple countries. He also held various senior level positions such as Director Product Management at Cisco Video Technologies, Vice President Product Management at NDS Services Pay-TV Technology Pvt. Ltd., Co-Founder and VP Engineering of Sensact Applications and Co-Founder and Architect at Metabyte Networks. He holds Executive General Management Program (EGMP) certificate from IIM Bangalore, MS in Computer Science from IIT, Madras and BE in Computer Science, MS from University of Baroda.
Abstract:
The idea of creating and using consumer personas is not new. Marketers have been going through painstakingly long way to understand and define consumer persona for their products. Further, they go through an intricate process of defining and executing elaborate campaigns to acquire consumer information and map the same to required personas. Even after spending big portion of their budget, marketers face various problems in reaching out to the right consumer: Data acquisition is an expensive task, many times data is not authentic or recent; all this starts to affect the conversion rate of the business making the ROI a far-fetched dream. Typical approach used in data acquisition and persona creation suffers from multiple problems: Most personas built today are static. Yes, the practice of updating consumer profile periodically is helpful but not ideal. Second, there are just too many factors (attributes) involved in the consumer’s decision making process. Marketer’s approach of confining consumer to few personas is quite limiting and inaccurate. The answer to these complex problems is to build a multidimensional consumer profile that is always up-to-date. This is possible by engaging the consumers at various stages during their day, be it online venues such as social network, reviews, blogs, opinions, surveys or offline venues such as surveys, transactions, logs and so on. Developing a multidimensional profile that is up-to-date is not a simple task. It is the kind of problem where tools such as big data, data analytics and machine learning can be used most effectively.
Tanya Dixit
Qualcomm, India
Title: The role of artificial intelligence in expanding sociological imagination and integrating more individuals into the societal framework thus completely eradicating destructive behavior, illiteracy and illness in developing nations
Time : 14:10 - 14:40
Biography:
Tanya Dixit has completed her BTech from BITS Pilani in Electrical and Electronics Engineering. She is working as an Engineer at Qualcomm, a pioneer in wireless and IOT chips. She is a machine learning and artificial intelligence enthusiast and is a deep thinker. She has a firm grasp of psychology, cognitive science, embedded systems, machine learning and hopes to build products that make the life of the common man better.
Abstract:
Sociological imagination is the awareness of the relationship between personal experience and the society. This talk is about methods to apply artificial intelligence in ways that can help a human being more connected to the society that he/she is living in. There are three focal points of this talk: Education, health and social interactions. Artificial intelligence can be applied to education and help more students get educated. Wearable’s and virtual reality is the technology to be used in making education accessible and make it application based. Similarly, for medicine, existing IOT products to be linked and generate an online assessment report on each person, and let a text analyzer process that for red flags. Emile Durkheim, in his classic book called suicide, mentions the main cause of suicide as the dissociation of a person from the society. AI products can help solve this dissociation by eliminating boundaries between human body and computers. Humans would be able to access information at the speed of thought and that will help a person get connected to more people like himself, thus eliminating suicide. Also, video camera analyzers to analyze footage and assess the body language of humans in real time thus predicting rapes, assaults and general nuances, thus leading to their timely resolution.
Ahmed AlMaqabi
Almaqabi, Kingdom of Bahrain
Title: Could machine learning and other technologies disrupt audit industry ?
Time : 14:40 - 15:10
Biography:
Ahmed has completed his CPA exams at the age of 22 years from California Board of Accountancy. He is the Founder and CEO of Almaqabi, a startup leading blockchain in audit & banking industry in Middle East. He has more than 10 years Big 4 experience in USA and Middle East.
Abstract:
Almaqabi is exploring the use of blockchain, machine learning, artificial intelligence, automation and Internet of Things (IOT) in audit. Department of eAuditchain focus on 4th industrial revolution technologies solutions to banks and financial instituions. Blockchain Audit Council was annouced recently to help top management of companies and bank understand the impact of 4th industrial revolution technologies on how it will change the way we live, work and interact with each other. Almaqabi won and recognized in regional competitions including AIM startup roadshow competition organized by Dubai Ministry of Economy.
Break: Coffee Break 15:10-15:30
Kai Khalid Miethig
Tariq Faqeeh Engineering, Bahrain
Title: W-AI-STE and architecture waste management in buildings with the support of machine learning and artifical intelligence
Time : 15:30 - 16:00
Biography:
Kai Khalid Miethig has completed his architectural study in 2004 at the University of Siegen, previous known as University of Applied Sciences of Siegen and gained further experience in waste management at Lobbe Environmental Consultancy. He has more than a decade experience in project management and lecturing. He is the Managing Director of Tariq Faqeeh Engineering in Bahrain, dedicated to enhance lifestyle and quality of living, offering various unique services for developers, government entities and individuals. He initiated the environmental awareness campaign “A Wave of Change” in cooperation with the Supreme Council of Environment Bahrain, German Embassy Bahrain and CleanUp Bahrain in 2017 and is providing/lectures on environmental awareness as well on the context and connection of architecture and waste management automation.
Abstract:
W-AI-STE - The implementation of artificial intelligence and machine learning in the processes of waste management, segregation and collection in architectural developments. It is a new and unique form of combining the existing technologies with the latest technologies of artificial intelligence and machine learning for an ancient issue the humanity is facing ever since mankind exists: Waste management. Due to the fact that the population is increasing continuously and therewith the production of waste as well, why not using the available technologies and features for supporting that basic task, which no one wants to deal with and is a topic which is now becoming a worldwide problem. Anywhere in the world where humans are present and were present as well even in areas humans never have been waste can be found as a remains of human existence. As we all are aware of smart home systems, e.g. fingerprints for entrance control and access or speech control featuring various functions in the house, this AI and ML technologies can be as well implemented for our daily legacies: WASTE- W-AI-STE. All humans have a certain pattern of whatever they do or their habits, these pattern and habits are as well detectable in their waste. Most waste comes from food packing and packing in general, so people have their favorite food, ingredients and other recurrent favorites which can be recorded, tracked and analyzed with the help of artificial intelligence and machine learning. So in an early stage of developments for residential and commercial purpose the pattern can be used to develop the concept of waste management systems.
Abed Benaichouche
Inception Institute of Artificial Intelligence, Abu Dhabi, UAE
Title: Overview of recent advance of deep learning application in computer vision
Time : 16:00 - 16:30
Biography:
Abed Benaichouche has completed his PhD from Ecole Nationale Superieur des Mines de Paris and Postdoctoral studies from French Geological Survey. He is a Senior Research Engineer and Team Leader at Inception Institute of Artificial Intelligence, (IIAI) the UAE’s national research organization, which aims for breakthroughs in fundamental and applied AI research. He has published more than 10 papers in reputed journals and international conferences. Before joining IIAI, he has worked for 2 years in BRGM (France) as Research Engineer. He designed and developed the Suricate-Nat platform; interactive web platform to collect information and analyzing it using NLP techniques on natural disasters emitted by citizens using social networks Twitter. He was also the Project Manager of the development of international interoperable platform for multi-risk analysis.
Abstract:
In recent years, deep learning (DL) has won numerous contests in computer vision and machine learning. In this presentation, we will present real world applications of Conventional Neuronal Network (CNN), Recurrent Network (RNN) and Generative Adversarial Network (GAN) in computer vision area. In the presentation, we will show a selection of recent research that Inception Institute of Artificial Intelligence (IIAI) is leading in the field of computer vision and artificial intelligence. For the CNN, we will present its application for face detection and annotation, demo for object detection and pose camera estimation. For the GANs, we will show its use for image colorization and art style transfer. And finally we present a new approach for face detection and super-resolution using both CNN and GAN models. For each demo we present the designed network, its limitations and the give perspectives for possible improvement.
- POSTER PRESENTATION 16:30-17:30
Session Introduction
Balasubramanyam Pisupati
Robert Bosch Engineering and Business Solutions Limited, India
Title: Ensemble of time series forecasting in complex structure
Biography:
Balasubramanyam Pisupati is currently working with Robert Bosch Engineering Solution and Business Solution as a Senior Manager in Data Analytics team. He has accomplished senior statistical professional with rich experience of more than 10 years in software industry related to product development, testing and data mining.
Abstract:
Forecasting is necessary for better business understanding and decision making. When it is done manually, it requires a lot of effort and time from multiple departments like logistics, sales, finance, etc. It also involves lot of gut feeling from experienced people and sometimes it might lead to error prone prediction if person is inexperienced or is not aware of past behavior. It is even very challenging for any data scientist to find a forecast model that performs best for all scenarios and in all forecast horizons. In this paper an approach for forecasting using ensemble model is discussed. Ensembling is done using symmetric mean absolute percentage error and mean absolute percentage error calculated from rolling forecast approach. For validation of forecast model, M3 competition data is used. This approach has resulted in better performance on out of sample prediction.
Beesung Kam & Byung Kwan Choi
Pusan National University, South Korea
Title: Predicting future rank and score of graduation by using student’s status and temporal behavior
Biography:
Beesung Kam has his degree in Computer Science and another in Medical Anatomy. He has completed his second PhD from Pusan National University School of Medicine.He is the Director of Maritime Mobile Health Research, a premier bio-soft service organization. He is researching in the area of medical education since 2005 and has digitalized student assessments of different grades.
Abstract:
Although there is a rich literature predicting student performance based on courses and lectures, it is much less studied in predicting degree completion upon affection of student’s personality. Student’s low score of graduation is a critical issue in medical school because students who result in lower scores mostly fail from Korean Medical Licensing Examination (KMLE). This paper introduces a method of supervised learning to predict future rank and score of students by using information provided prior of entrance to the school. Data sets of 256 number of graduated students from Pusan National University School of Medicine school year of 2016 and 2017 were distributed to machine learning by 2500 times. By using method of ordinary least squares regression three groups of students were established depending on their achieved final graduation to low, mid and high scores. These groups were used as a guideline to pivot personality data of freshmen for further comparison. Prediction analyzed by student’s status, based on their student’s temporal behavior such as age, sex, blood type, school of graduation, district, major, hobby, religious, drinking habit, parent’s status, tuition method of payment and registration to an application can influence on their future earned rank and score. Predicating future student rank and score helps to monitor, review and re-establish student’s road map to enhance learning progress. This prediction not only helps student to know how they do but also encourages them as a feedback to reinforce their current method of learning for further improvement. Although increasing pilot data for this study can enhance the study achievement and improving student personality can be challenging, method of determining low score of graduation by machine learning can sparkle new era to progress learning.
Semakula Abdumajidhu
Makerere University, Uganda
Title: Sooty mold effect on cassava yields using convolutional neural networks
Biography:
Semakula Abdumajidhu is currently pursuing his Master of Science degree in Computer Science at the College of Computing and Information Sciences (CoCIS), Makerere University. His major is Computer Vision and Image Processing with his current research on crop diseases.
Abstract:
Whiteflies have been observed to cause two forms of damage to the host cassava plants that is feeding damage and sooty damage. The secondary damage, referred to as sooty mold damage, is a result of honeydew dropped by the feeding whiteflies on lower leaves. It is characterized with a sooty darkening of the lower leaves that affect chlorophyll content levels, because chlorophyll is an indicator of plants nutritional stress, photosynthetic capacity and the healthy status of plants. Plants need much possible levels of chlorophyll content to absorb enough light that is to be used during photosynthesis for them to make their own food. Sooty mold damage with its black-darkening-effect abstract the leaves from getting direct light rays during the photosynthesis process which leads to low cassava yields. The use of machine learning for surveilling the health of crops has been looked at in a number of related settings, including the segmentation of diseased leaves and disease-related dis-coloration in citrus fruit. The study will be aimed at understanding the effect of sooty mold on chlorophyll content in cassava plants and see how we can improve and raise on the cassava yields. Five cassava varieties that are narocass 1, nase 14, mkumba, njule and bamunanika will be considered during the research. We shall determine the level of infestation by sooty mold on these cassava varieties and tell which variety is most affected. We shall analyze and relate spectral reading data with chlorophyll content to measure how much of chlorophyll is affected by sooty mold damage in cassava images. We shall develop a Convolutional Neural Network (CNN) model to estimate pigment (chlorophyll) content based on spectrometer readings. Using our model, we shall determine the percentage of the plant being affected and the extent covered by sooty mold. We shall use field spectrometer to take spectral field measurements and also carry out remote sensing analysis of the data. This data will be related to the chlorophyll content to determine the levels of sooty mold effect. The research will lead to increase in cassava yields as a result to a more accurate detection of sooty molds. This will improve the response time for whiteflies and sooty mold infected cassava care. This will no longer be need for handcrafting/manually extracting features.
Rohit Agarwal & Gaurav Pawar
Mobisy Technologies Pvt Ltd, India
Title: An overview of deep learning based object detection techniques in retail domain
Biography:
Rohit Agarwal is working as a Data Scientist in Mobisy Technologies Pvt. Ltd., Bangalore where he leads a team of data scientists and software engineers, focusing on sales force automation by applying state of the art ML and deep learning techniques. He has 12 years of industry experience with 11 years in GE where he worked on conceptualizing, designing, prototyping a number of software and data solutions using cutting-edge technologies for solving large industrial problems. He has completed his asters in IT from IIIT, Bangalore and Bachelors in Computer Science from IET, Lucknow.
Gaurav Pawar has completed his BE degree in Electronics and Telecommunication with more than 3 years of practical hands-on experience in computer vision and machine learning. He is specialized in building quick prototypes using python environment by leveraging GPU platform. He has expertise in using popular deep learning libraries such as TensorFlow, PyTorch and Keras. Currently, he is interested in solving data science problems in Indian retail industry using images and other sources of data.
Abstract:
We have used image recognition and machine learning technology to automate some of the time consuming and error prone auditing use cases pertinent for SME retail stores (aka mom-n-pop stores) like which all SKU type are present in other sourcesparticular store. Check if the store has put out advertising of the brand as was agreed upon. Current process of store auditing is conducted through feet on street sales force/external auditing agencies which is manual, biased, time consuming, costly affair and is not scalable. Leveraging the state of art deep learning image recognition technology, our platform helps in automating store auditing process by acting as eyes for tracking all types of in-store visibility executions like window displays, POS material and outdoor advertising/banners with high degree of precision (>90%) which is much better than classical approached like SVM (~80%). The platform can analyze millions of retail store images to generate actionable insights for brands/company. Role of the sales force is limited to take pictures of the stores and upload them to our platform. Most of the current image analytics platform works with high quality images of organized environment like supermarkets, this makes our platform different as it has been specifically designed for mom-n-pop store setup which symbolizes unorganized shelves and cluttered environment. The images obtained are also low quality as they are typically shot by sales force using low quality mobile camera. Some of the common issues with these images are low light, partial visibility, occlusion, glare, incorrect angle, etc. In this, we intend to give a technical overview of the platform, highlight its capability to analyze images of varying nature, showcase few use cases in SME domain that can be implemented using this platform.
Noor Alasadi
Damascus University, Iraq
Title: Lamassu, Detecting cyber extremism via automatic lexical feature discovery and social analysis - An application on Jihad of demand networks
Biography:
Noor Alasadi is a Senior Data Scientist at Creditinfo Group, a leading service provider of credit information and risk management solutions worldwide. He is a Graduate Instructor and a Researcher at Damascus University, Department of Artificial Intelligence and Natural Language Processing. He is also a Member of the Scientific Committee of the ACM-ICPC in which he was a Judge, Problem Setter, Coach and Organizer in the Arab Collegiate Programming Contest (2012-2017). He was also involved in multiple private-public partnership projects with companies and authorities in the Middle East to build intelligent security systems.
Abstract:
Social networking websites have enjoyed a great success in recent years, apart from the numerous new opportunities that they are providing, extremist groups and terrorist organizations are using them to promote their ideology to facilitate internal communications and to evoke a planned psychological reaction in their enemies. Many web resources contain information about extremism, but a relatively small proportion comes from terrorist groups themselves and since manually monitoring and analyzing all their content separately during warfare is unattainable, solutions using automated methods are sought. This study applies machine learning techniques to perform automated extremist language detection. In this project, we proposed an approach for detecting extremist content and identifying potential extremist users in social media. The study’s methodology explores features in users’ histories to predict extremism via statistical topic model on an Arabic corpus which detects extremist posts with automatically generated features and a graded structure in which, whether extremism applies to a given person is a matter of degree related to multiple factors. To demonstrate our work, we created a dataset containing over 360,000 web forum posts. Experiments on a sampled data set show precision of 96.20% and recall of 94.90%.
Aisha Al Owais
Sharjah Center for Astronomy and Space Sciences, UAE
Title: Deep learning - An application of machine learning to classify images
Biography:
Aisha Al Owais has completed her BSc in Computer Science from the College of Engineering at the American University of Sharjah. She is working as a Research Assistant in the Meteorites Center at the Sharjah Center for Astronomy and Space Sciences.
Abstract:
Living in the 21st century, mankind’s most powerful weapon is technology. The field of technology we are interested in is computer science, specifically Artificial Intelligence (AI). As the name suggests, AI is about turning devices into intelligent agents that take actions based on the environment they perceive. They are also flexible in terms of changing their goal- what they are meant to do as well as adjusting their actions depending on its changing environment. What makes AI agents peculiar is their ability to learn and remember from their mistakes. Furthermore, Machine Learning (ML) is one of AI’s applications that enable systems to learn automatically, improve through experience and adjust its actions without human intervention. This takes us to Deep Learning (DL), a new subfield of ML concerned with algorithms inspired by the structure and function of the human’s brain called artificial neural networks. It has networks which are capable of learning data obtained from instructed or unlabeled data; therefore it is also known Deep Neural Network (DNN). All those terms lead us to what we are mostly interested in, Convolutional Neural Networks (CNNs), which is a deep neural network that is particularly wall-adapted to classify images, in our case to classify images of meteorites.
Nabil Belgasmi
Banque de Tunisie, Tunisia
Title: Multiobjective deep reinforcement learning approach for ATM cash replenishment planning
Biography:
Nabil Belgasmi holds a PhD and an Engineering degree in Computer Science from Manouba University. He is a full stack Data Scientist at Banque de Tunisie, Tunisia. He is involved in three main activities: (1) Applied R&D, (2) Data Analytics Technology Watch and (3) Data Science consulting. He achieved many successful Data Science POCs and Quick-Wins: Credit Scoring, Forecasting, Cash Planning, Anomaly/Fraud Detection, Customers Profiling, Intelligent Transactions Scoring & Monitoring, etc. He is a Member of the Industrial Editorial Board of the Engineering Applications of Artificial Intelligence journal (EAAI).
Abstract:
The current framework of reinforcement learning is based on a single objective performance optimization that is maximizing the expected returns based on scalar rewards that come from either univariate environment response to the agent actions or from a weighted aggregation of a multivariate response. But in many real world situations, tradeoffs must be made among multiple conflicting objectives that have the different order of magnitude, measurement units and business specific contexts related to the problem being solved (i.e. costs, lead time, quality of service, profits, etc.). The aggregation of such sub-rewards to get a scalar reward assumes a perfect knowledge about the decision maker preferences and the way she perceives the importance of each objective. In this study, we consider the problem of learning the best ATM cash replenishment policies in an uncertain multi-objective context given an arbitrary history of cash withdrawals that may be non-stationary and may contain outliers. We propose a model-free Multi-objective Deep Reinforcement Learning approach that allows us to compete against the human decision maker and to find the best policy per ATM that outperforms the current human policy. The idea is to disaggregate the performance of a replenishment policy to form a vector of objective functions. The performance of the human policy is then a multi-dimensional reference point (Rh). The task of the deep reinforcement learning algorithm is to find a policy that generates a set of performance points which Pareto-dominate the current human reference point (Rh).
Break: Panel Discussion , Awards and Closing Ceremony
- PRODUCT LAUNCH BY SHABIR MOMIN & SANTOSH GODBOLE (11:10 - 11:25)
Session Introduction
SHABIR MOMIN & SANTOSH GODBOLE
SSN Solutions, London
Title: Product Launch
Time : 11:10 - 11:25
Biography:
Abstract:
- WORKSHOP
Session Introduction
Tanya Dixit and Sriharsha Allenki
Qualcomm, India
Title: Deep Learning – A beginner’s guide to start coding neural networks and apply in areas of computer vision and natural language processing
Time : 11:25 - 12:10
Biography:
Tanya Dixit has completed her BTech from BITS Pilani in Electrical and Electronics Engineering. She is working as an Engineer at Qualcomm, a pioneer in wireless and IOT chips. She is a machine learning and artificial intelligence enthusiast and is a deep thinker. She has a firm grasp of psychology, cognitive science, embedded systems, machine learning and hopes to build products that make the life of the common man better.
Abstract:
Deep learning is an emerging area in machine learning. It is based entirely on the understanding of neural networks and their different architectures to achieve specific goals. Convolutional neural networks are used in computer vision to help machines identify faces, animals, cars, etc. Different convolutional neural network architectures have been developed in recent times to make computer vision tasks easier. Transfer learning can be applied to use the weights and architecture from an already trained network like VGG16 or VGG19 and apply it to relevant tasks. Recurrent neural networks are another class of neural networks that allows context to be preserved and hence is used for language processing. There are various types such as LSTMs, GRUs. Each has their advantages. This study discusses the basics of neural networks, mathematics of back-propagations, intricacies of CNNs, RNNs and then advance architectures like residual networks and ladder networks.
Break: Lunch Break 12:10-13:10
- Workshop
Chair
Tanya Dixit & Sriharsha Allenki
Qualcomm, India
- Computer Science and Technology | Machine Learning | Big Data, Data Science and Data Mining | Big Data Analytics | Business Intelligence | The role of AI & Machine Learning in Medical Science | Deep Learning | Object Detection with Digits | Computer Vision and Image Processing | Pattern Recognition | Robotic Process Automation (RPA) | Natural Language Processing (NLP) and Speech Recognition | Deep Learning Frameworks
Chair
Niladri Shekhar Dutta
Ericsson, UAE
Co-Chair
Abbas M Al-Bakry
University of Information Technology and Communications, Iraq
Biography:
Abstract:
Jayatu Sen Chaudhury
American Express, India
Title: Machine learning applications in credit card domain
Time : 13:10 - 13:40
Biography:
Jayatu Sen Chaudhury is the Vice President, Global Commercial and Merchant Data Science and Head of Enterprise Digital & Analytics India for American Express, India. Prior to this role, he was the Head of Global Information Management, Big Data Labs & Advanced Risk Capabilities. He has been a part of American Express since 2001, working in the various decision science functions for both US and international markets. He has earned his PhD in Financial Economics from IGIDR, Economic Research Institute funded by the Central Bank of the Country (Reserve Bank of India). Prior to joining American Express, he has worked in decision science for two years each in GE Capital and ICICI Bank.
Abstract:
Given the huge volumes of data available (both structured and un-structured) for American Express card members, American Express has adopted machine learning in all its core business processes of credit and fraud risk management, marketing analytics and operations. Work entailed building in-house data warehouses with right level of privacy controls and then using state of art machine learning algorithms from open sources to solve unique business problems across various business verticals. Adoption of machine learning has ensured building of robust economic models leveraging the best possible information, delivering the highest predictive power with utmost accuracy. The models are updated at the highest possible frequency ensuring the models incorporate the most recent information. This has led to significant improvement in the controls for fraud risk and also improved the targeting of appropriate segments with far higher accuracy in marketing. As a part of the presentation, 3-4 actual use cases of core American Express processes and how machine learning has completely changed the game will be discussed. Discussion will also include the new areas where company is thinking of doing research and bringing the best value for its card members.
Niladri Shekhar Dutta
Ericsson, UAE
Title: The advent of big data analytics in the world of ICT and digital
Time : 13:40 - 14:10
Biography:
Niladri Shekhar Dutta is a seasoned professional with more than 13 years of global consulting experience. He is a Consulting Practitioner with focus on operations and transformation consulting for top-tier telecommunication operators globally. He has worked in more than 30+ consulting engagements in varied culture and markets across Western Europe, Middle East, North and Central Africa, India and New Zealand. His expertise is primarily around C-level advisory, digital transformation, digital enterprise architecture, business and operational process management, IT operational strategy, digital risk and revenue management consulting. He is responsible for driving consulting business across the MEA region for Ericsson and is involved in both sales and delivery. He has his MBA in Marketing and Finance from Symbiosis, Pune, India and is an Engineering graduate in Electronics from University of Nagpur, India. He is presently also undergoing a specialized PG program from MIT Sloan and Columbia facilitated by Emeritus on Digital Business.
Abstract:
With the advent of technology transformation in the fast changing and ever evolving world of Information, Communications and Technology (ICT), the importance of data is supreme. This is often being referred to as big data and is perhaps the single most entity which forms the backbone of any major transformation within any large global corporation across industries. Data is no longer being looked and used as a tactical medium for storage or operations; on the contrary it becomes extremely strategic in nature. In fact the 3 main pillars of today’s disruptive world of digital are driven by big data, IoT and cloud. Out of which big data is the nucleus of transformation. In the world of digital, this is very well centered on three main life cycle entities. They are the customer, the product and the revenue. Each of these, i.e. customer life cycle, product life cycle and revenue life cycles behave very differently from one another. The practical emphases of data in each of these entities are also very different and unique. The concepts of big data within these 3 life cycles are core to the change we witness in the world of digital. Each data entity centered on these life cycles is instrumental in C-level decision making and major change management that happens within the organization. The data element acts as a central aspect to strategic decisions whether it comes to newproduct/service development or behavior of customer or user data, appreciation or acknowledgement of revenue. All use cases around big data will be largely centered on these and any specific case would be a secondary derivation of the above. With big data being so strategic in nature a large part of the focus has now shifted to data extraction and normalization to ensure meaningful information is extracted and utilized for business benefits by customers. Like the traditional mindset used to be, focus was largely around data operations and reporting. We will soon see a world where we cannot live without any form of data and in truest sense the phrase big data would essentially be big and super imposed in all aspects of our lives, right from our behavior, buying and consumption of products and services to distribution of our resources. The extraction and transformation of data for key benefits will be very much a business as usual thing, without which survival will become questionable within ICT industry, especially whilst looking at the concept of digital disruption. This article largely focuses on the key aspects of the same within the world of ICT and how a corporation is heavily dependent on such aspects for generation of its sales and management of its operations.
Manoj Mishra
Union Insurance, UAE
Title: Data virtualization - Using data virtualization for an integrated analytics platform
Time : 14:10 - 14:40
Biography:
Manoj Mishra has completed his Bachelor of Engineering in Computer Science and a Certification in Data Science from Johns Hopkins University. He has more than two decade of experience spreading across multiple geographies (US, Europe, India and Middle East) working with organizations like Adobe Systems, Dell, Perot Systems, CEB-Gartner, Rolta and Tata Group. He is currently a Chief Manager of Business Intelligence and Data with Union Insurance and currently leading their data strategy and technology transformations through data analytics, research and various AI initiatives.
Abstract:
In order to have the competitive advantage, organizations worldwide are driving the need for better analytics (historical, realtime, predictive and cognitive) of data across various domains including customers, products, services and operations. Due to this, the data available for such analytics is exploding in size, technology and complexity. For many years companies have invested in technologies like data warehouses, data marts, OLAP tools, Big Data/Hadoop systems and streaming real-time analytics platforms to take advantage of these opportunities. Total value preposition to the business is maximized only when these are combined into an integrated analytics platform. However, traditional tools cannot integrate streaming data and dataat-rest especially when the data is spread on-premises, cloud, websites and documents everywhere. Data virtualization can be used to provide cross platform logical views of data and analytic insights across the enterprise to provide an integrated analytics platform. By utilizing native integration with in-memory data grids for data processing, data virtualization can deliver a unified and centralized data services fabric with security and real-time integration across multiple traditional and big data sources, including Hadoop, NoSQL, cloud and software-as-a-service (SaaS). Hence data virtualization is becoming a need to address the unique challenges of data explosion in today’s changing business climate.
Eman AbuKhousa & Najati Ali-Hasan
UAE University, UAE
Title: Predictive big data analytics and healthcare fraud - From detection to prevention
Time : 14:40 - 15:10
Biography:
Eman Abu Khousa is a Researcher-Instructor (Big Data Applications) at the College of Information Technology, UAE.
Najati Ali Hasan is an experienced health information technology (IT) professional with 25-year experience in the field. Najati is an expert in advising GCC clients on strategies for selections & implementations of health IT with focus on achieving demonstrable clinical, operational and financial benefits. Najati is well versed in the revenue-cycle-management (RCM) field with knowledge of the various nuances and requirements of GCC countries. Najati’s other areas of expertise include smart use of health IT for enhanced patient experience, EDI, data analytics and applications of Artificial Intelligence/Machine Learning (AI/ML) in healthcare. Najati has co-authored three articles for conferences and journals – one having received a best-paper award. Najati’s work experience spans top USA medical centers to world class suppliers of health IT.
Abstract:
The losses from healthcare fraud, over-prescribing and improperly coded insurance claims leading to claim-denials are estimated in the billions of dollars annually. The costs associated with fraud and acts of abuse are increasing insurance premiums for patients and cuts into the profitability of healthcare service providers and payers. The continuing adoption of Electronic Health Records (EHRs) and the advances of machine learning and big data analytics enable more efficient and automated methods for detecting and effectively mitigating the risk of fraudulent activities and illegitimate claims. This paper provides an overview of the new systems and methods to reduce medial claims fraud and a review of open issues and challenges. This paper also proposes a predictive analytics approach to detect potential fraudulent patterns using a set of supervised and unsupervised learning techniques. The proposed approach incorporates both historical and real-time data to identify illegal claims and prevent payouts to fraudsters early in the claims management process lifecycle.
Tilila El Moujahid
Microsoft Corporation, UAE
Title: Using genomics cloud platform and machine learning for genome variant analysis
Time : 15:10 - 15:40
Biography:
Tilila is currently a Data Scientists and a Technical Evangelist for data and AI working at Microsoft. She accompanies partners in architecting and building their
cloud based, AI powered solutions. She was previously a Technology Strategist for enterprise accounts including Education and healthcare industry. She’s also a Fulbright scholar who earned, in 2012, a master in Computer Science and Business from San Francisco State University where her focus was on Bioinformatics including Genomics and Biomedical Image Analysis.
Abstract:
Next Generation Sequencing (NGS) allows performing massively parallelled DNA sequencing and is currently revolutionizing biological studies. Instead of sequencing a specific set of genes solely, NGS allows to sequence a wider portion of the genome (even a whole genome), which opens the door for a wider analysis of biological pathways within an individual. Researchers have never before accessed such a wealth of genomic data which holds the promise of unvealing the secrets of the most daunting deseases of the century such as cancer. It also comes with its own set of challenges for the management and analysis of Big Data to extract meaningful and actionable insights. Cloud computing and machine learning do have the capacity to solve this challenge. In this talk we will show how Cloud-based Genomics platform allows to manage petabytes of genomic data as well as foster fast and agile Secondary Analysis. We’ll also also use genomics specific machine learning packages to perform Tertiary Analysis on gene variants data and visualization tools to expose and share the results with the scientific community. We will begin the talk with the introduction to the genomics field and the commonly used genomic analysis process and will present practical applications of the above services and analysis.
Break: Networking and Refreshments Break 15:40-16:00
Abbas M Al-Bakry
University of Information Technology and Communications, Iraq
Title: Computer aided diagnosis In cloud environment based on multi agents system
Time : 16:00 - 16:30
Biography:
Abbas M Al Bakry has completed his PhD in Computer Science (Artificial Intelligence) from the University of Technology, Baghdad, Iraq. Currently, he is the President of the University of IT and Communications in Baghdad, General Chair of the NTICT annual conference-Baghdad, Editor in Chief of IJCI Iraqi Journal for Computers and Informatics, Editor-in-Chief, Intelligent Computing in the IJNC International Journal of Network Computing and Advanced Information Management and Editor in JCIT, AISS, JINT, IJACT, IJIIP international journals.
Abstract:
In this speech we address solutions for the problems of the low accurate decision; low availability especially in maintains procedures and the scalability in online Computer Aided Diagnosis (CADs). Most CADs became available online and provide a high importance medical services which develop the health of the human beings. CADs are to increase the detection of disease by reducing the false negative rate due to observational oversights. The online CADs face three major problems: (1) The CADs cannot diagnose some diseases because the symptoms of these diseases are not available in the knowledge bases of this systems, (2) problem is the availability of CADs is depends on the web server which hosted them. Web server may possible to stop for maintenance that will implies to stop the CADs systems. (3) The problem is scalability related with the cost if their admins want to expand them to cover more medical problems. In this lecture we proposed a new framework to solve the above problems. The framework is composed of multi agents system to work on the environment of the cloud computing. The framework consists from three Sections: SaaS components, PaaS components and IaaS components. Each section has its own algorithms and procedures. To evaluate the resulted framework we make a survey in for 150 persons from medical health sector, students, specialists, physicians and other. The results pointed to good ratio of acceptance from the users.
Rohit Agarwal & Gaurav Pawar
Mobisy Technologies Pvt Ltd, India
Title: An overview of deep learning based object detection techniques in retail domain
Time : 16:30 - 17:00
Biography:
Rohit Agarwal is working as a Data Scientist in Mobisy Technologies Pvt. Ltd., Bangalore where he leads a team of data scientists and software engineers, focusing on sales force automation by applying state of the art ML and deep learning techniques. He has 12 years of industry experience with 11 years in GE where he worked on conceptualizing, designing, prototyping a number of software and data solutions using cutting edge technologies for solving large industrial problems. He has completed his Masters in IT from IIIT, Bangalore and Bachelors in Computer Science from IET, Lucknow.
Gaurav Pawar has completed his BE degree in Electronics and Telecommunication with more than 3 years of practical hands-on experience in computer vision and machine learning. He is specialized in building quick prototypes using python environment by leveraging GPU platform. He has expertise in using popular deep learning libraries such as TensorFlow, PyTorch and Keras. Currently, he is interested in solving data science problems in Indian retail industry using images and other source of data.
Abstract:
We have used image recognition and machine learning technology to automate some of the time consuming and error prone auditing use cases pertinent for SME retail stores (aka mom-n-pop stores) like which all SKU type are present in particular store. Check if the store has put out advertising of the brand as was agreed upon. Current process of store auditing is conducted through feet on street sales force/external auditing agencies which is manual, biased, time consuming, costly affair and is not scalable. Leveraging the state of art deep learning image recognition technology, our platform helps in automating store auditing process by acting as eyes for tracking all types of in-store visibility executions like window displays, POS material and outdoor advertising/banners with high degree of precision (>90%) which is much better than classical approached like SVM (~80%). The platform can analyze millions of retail store images to generate actionable insights for brands/company. Role of the sales force is limited to take pictures of the stores and upload them to our platform. Most of the current image analytics platform works with high quality images of organized environment like supermarkets, this makes our platform different as it has been specifically designed for mom-n-pop store setup which symbolizes unorganized shelves and cluttered environment. The images obtained are also low quality as they are typically shot by sales force using low quality mobile camera. Some of the common issues with these images are low light, partial visibility, occlusion, glare, incorrect angle, etc. In this, we intend to give a technical overview of the platform, highlight its capability to analyze images of varying nature, showcase few use cases in SME domain that can be implemented using this platform.
Sharan Kumar Santhanakrishnan & Rakhsanth Rammohan
St.Josephs’s College of Engineering, India
Title: Efficient way for detecting and tracking of object irrespective of speed of the motion
Time : 17:00 - 17:30
Biography:
S Sharan Kumar is pursuing final year engineering at St Joseph’s College of Engineering, Chennai, India. He has participated in world’s biggest hackathon and was one among top 10 teams in the respective node.
R Rakhsanth is pursuing final year engineering at St Joseph’s College of Engineering, Chennai, India. He has participated in world’s biggest hackathon and was one among top 10 teams in the respective node.
Abstract:
Multiple object detection and tracking is still facing some difficulties like efficiency, reliability and creation of datasets. In this paper we are going to present about the efficient way of detecting and tracking of multiple objects regardless of the speed of its motion. We use tensor-flow in combination with OpenCV in which both are open source. We use tensor-flow for training the data sets for object identification and Kalman filtering algorithm for motion tracking and embed them into OpenCV for real-time image/video manipulations. We use SSD (Single Shot multi-box Detector) architecture of convolutional neural network for training the dataset for minimizing the GPU requirement than RCNN (Region Convolution Neural Network). We have also implemented a special algorithm that will boost the SSD model’s accuracy to some extent. The datasets for any training model needs to be well suited and labeled which is somewhat difficult to create that dataset. To reduce the difficulty of that we have proposed another technic by which we can just send the dataset with noise into the OpenCV only with labels. Then system will automatically crop those image from the datasets according to the requirement of the architecture (SSD) used. This will help us in creating datasets in a simple and cost efficient way.
Aisha Al Owais
Sharjah Center for Astronomy and Space Sciences, UAE
Title: Deep learning - An application of machine learning to classify images
Time : 17:30 - 18:00
Biography:
Aisha Al Owais has completed her BSc in Computer Science from the College of Engineering at the American University of Sharjah. She is working as a Research
Assistant in the Meteorites Center at the Sharjah Center for Astronomy and Space Sciences.
Abstract:
Living in the 21st century, mankind’s most powerful weapon is technology. The field of technology we are interested in is computer science, specifically Artificial Intelligence (AI). As the name suggests, AI is about turning devices into intelligent agents that take actions based on the environment they perceive. They are also flexible in terms of changing their goal- what they are meant to do as well as adjusting their actions depending on its changing environment. What makes AI agents peculiar is their ability to learn and remember from their mistakes. Furthermore, Machine Learning (ML) is one of AI’s applications that enable systems to learn automatically, improve through experience and adjust its actions without human intervention. This takes us to Deep Learning (DL), a new subfield of ML concerned with algorithms inspired by the structure and function of the human’s brain called artificial neural networks. It has networks which are capable of learning data obtained from instructed or unlabeled data; therefore it is also known Deep Neural Network (DNN). All those terms lead us to what we are mostly interested in, Convolutional Neural Networks (CNNs), which is a deep neural network that is particularly wall-adapted to classify images, in our case to classify images of meteorites.
Break: Panel Discussion
Session Introduction
Jayatu Sen Chaudhury
American Express, India
Title: Machine learning applications in credit card domain
Biography:
Jayatu Sen Chaudhury is the Vice President, Global Commercial and Merchant Data Science and Head of Enterprise Digital & Analytics India for American Express, India. Prior to this role, he was the Head of Global Information Management, Big Data Labs & Advanced Risk Capabilities. He has been a part of American Express since 2001, working in the various decision science functions for both US and international markets. He has earned his PhD in Financial Economics from IGIDR, Economic Research Institute funded by the Central Bank of the Country (Reserve Bank of India). Prior to joining American Express, he has worked in decision science for two years each in GE Capital and ICICI Bank.
Abstract:
Given the huge volumes of data available (both structured and un-structured) for American Express card members, American Express has adopted machine learning in all its core business processes of credit and fraud risk management, marketing analytics and operations. Work entailed building in-house data warehouses with right level of privacy controls and then using state of art machine learning algorithms from open sources to solve unique business problems across various business verticals. Adoption of machine learning has ensured building of robust economic models leveraging the best possible information, delivering the highest predictive power with utmost accuracy. The models are updated at the highest possible frequency ensuring the models incorporate the most recent information. This has led to significant improvement in the controls for fraud risk and also improved the targeting of appropriate segments with far higher accuracy in marketing. As a part of the presentation, 3-4 actual use cases of core American Express processes and how machine learning has completely changed the game will be discussed. Discussion will also include the new areas where company is thinking of doing research and bringing the best value for its card members.
Niladri Shekhar Dutta
Ericsson, UAE
Title: The advent of big data analytics in the world of ICT and digital
Biography:
Niladri Shekhar Dutta is a seasoned professional with more than 13 years of global consulting experience. He is a Consulting Practitioner with focus on operations and transformation consulting for top-tier telecommunication operators globally. He has worked in more than 30+ consulting engagements in varied culture and markets across Western Europe, Middle East, North and Central Africa, India and New Zealand. His expertise is primarily around C-level advisory, digital transformation, digital enterprise architecture, business and operational process management, IT operational strategy, digital risk and revenue management consulting. He is responsible for driving consulting business across the MEA region for Ericsson and is involved in both sales and delivery. He has his MBA in Marketing and Finance from Symbiosis, Pune, India and is an Engineering graduate in Electronics from University of Nagpur, India.
Abstract:
With the advent of technology transformation in the fast changing and ever evolving world of Information, Communications and Technology (ICT), the importance of data is supreme. This is often being referred to as big data and is perhaps the single most entity which forms the backbone of any major transformation within any large global corporation across industries. Data is no longer being looked and used as a tactical medium for storage or operations; on the contrary it becomes extremely strategic in nature. In fact the 3 main pillars of today’s disruptive world of digital are driven by big data, IoT and cloud. Out of which big data is the nucleus of transformation. In the world of digital, this is very well centered on three main life cycle entities. They are the customer, the product and the revenue. Each of these, i.e. customer life cycle, product life cycle and revenue life cycles behave very differently from one another. The practical emphases of data in each of these entities are also very different and unique. The concepts of big data within these 3 life cycles are core to the change we witness in the world of digital. Each data entity centered on these life cycles is instrumental in C-level decision making and major change management that happens within the organization. The data element acts as a central aspect to strategic decisions whether it comes to newproduct/service development or behavior of customer or user data, appreciation or acknowledgement of revenue. All use cases around big data will be largely centered on these and any specific case would be a secondary derivation of the above. With big data being so strategic in nature a large part of the focus has now shifted to data extraction and normalization to ensure meaningful information is extracted and utilized for business benefits by customers. Like the traditional mindset used to be, focus was largely around data operations and reporting. We will soon see a world where we cannot live without any form of data and in truest sense the phrase big data would essentially be big and super imposed in all aspects of our lives, right from our behavior, buying and consumption of products and services to distribution of our resources. The extraction and transformation of data for key benefits will be very much a business as usual thing, without which survival will become questionable within ICT industry, especially whilst looking at the concept of digital disruption. This article largely focuses on the key aspects of the same within the world of ICT and how a corporation is heavily dependent on such aspects for generation of its sales and management of its operations.
Chair
Abbas M Al-Bakry
University of Information Technology and Communications, Iraq