UNIVERSITY OF HERTFORDSHIRE COMPUTER SCIENCE RESEARCH COLLOQUIUM presents "A Game Theoretic Perspective of Time Series Classification Model Construction" Dr. Arijit Ukil (Embedded Devices and Intelligent Systems, TCS Research, Tata Consultancy Services, India) 27 October 2023 (FRIDAY) 13:00 -14:00 Room 1A161 Everyone is Welcome to Attend Abstract: Time series sensor data classification tasks often suffer from training data scarcity issue due to the expenses associated with the expert-intervened annotation efforts. For example, Electrocardiogram (ECG) data classification for cardio-vascular disease (CVD) detection requires expensive labeling procedures with the help of cardiologists. Current state-of-the-art algorithms like deep learning models have shown outstanding performance under the general requirement of availability of large set of training examples, while it does not perform well under training data scarcity challenges. In this talk, I shall discuss on Shapley Attributed Ablation with Augmented Learning: ShapAAL, which demonstrates that deep learning algorithm with suitably selected subset of the seen examples or ablating the unimportant ones from the given limited training dataset can ensure consistently better classification performance under augmented training [1]. In ShapAAL, additive perturbed training augments the input space to compensate the scarcity in training examples using Residual Network (ResNet) architecture through perturbation-induced inputs, while Shapley attribution seeks the subset from the augmented training space for better learnability with the goal of better general predictive performance, thanks to the “efficiency” and “null player” axioms of transferable utility games upon which Shapley value game is formulated. ShapAAL is a novel push-pull deep architecture where the subset selection through Shapley value attribution pushes the model to suitable lower dimension while augmented training augments the learning capability of the model over unseen data. We perform ablation study to provide the empirical evidence of our claim and we show that proposed ShapAAL method consistently outperforms the current baselines and state-of-the-art algorithms for time series sensor data classification tasks from publicly available UCR time series archive that includes different practical important problems like detection of CVDs from ECG data. Secondly, I discuss Priv-Aug-Shap-ECGResNet that extends ShapAAL to enable a novel data privacy preservation feature through differential privacy technique to provide measured obfuscation to render ZeroR classification equivalent knowledge gain to the adversary [2]. [1] Ukil, Arijit, Leandro Marin, and Antonio J. Jara. "When less is more powerful: Shapley value attributed ablation with augmented learning for practical time series sensor data classification." Plos one 17.11, 2022. [2] Ukil, Arijit, Leandro Marin, and Antonio J. Jara. "Priv-Aug-Shap-ECGResNet: Privacy Preserving Shapley-Value Attributed Augmented Resnet for Practical Single-Lead Electrocardiogram Classification." IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2023. Biography: Dr. Arijit Ukil is having more than 20 years of research experience in different capacities. From 2003 to 2007, he worked as Scientist in Defense Research and Development Organization (DRDO), Government of India in the area of Radar signal Processing, particularly for 3D-Radar systems. Currently, he is working as Senior Scientist in TCS Research, Tata Consultancy Services, India with expertise in the areas of healthcare analytics, deep learning and explainable AI. At present, he is leading the research and development activities in the area of sensor data analytics for the development of scientifically interpretable deep learning model generation, particularly for augmenting the clinical utility of AI-driven algorithmic screening models. He is the chief architect of Xsense, a sensor-agnostic platform for performing different analytics tasks including prediction, classification on diverse sensor signals including Electrocardiogram, accelerometer, smart energy meter, phonocardiogram (heart sound), etc. He is currently building AI-assisted heart monitoring solution to transform the patient care, particularly for cardio-vascular disease diagnosis and automated clinical intervention. Further, he is heading the development of deep learning model optimization using state of the art model optimization techniques like knowledge distillation, lottery ticket hypothesis for developing remote, affordable and portable intelligent Electrocardiogram sensing at edge devices like smart sensors, wearables, smartphones for algorithmic screening purpose. He has published more than 60 research papers in distinguished conferences and journals including ACM CIKM, IEEE ICASSP, IJCAI, IEEE ICC, IEEE IJCNN, IEEE Infocom, NeurIPS, ICML, IJCAI, IEEE Sensors Journal, PLOS-ONE, etc. (https://scholar.google.co.in/citations?user=wXFj5skAAAAJ&hl=en). He has authored 4 book chapters. He has filed more than 50 patents with more than 40 granted patents in different geographies worldwide including India, US, Europe, China, Japan, Singapore. He holds Master’s in Engineering from Jadavpur University, Kolkata, India and PhD (Cum Laude) in Computer Science from University of Murcia, Spain. He is a Senior Member, IEEE. He is in editorial board of number international journals and acts as reviewer in different top-tier conferences and journals. He is in the Board of Study of different renowned institutes. He has delivered lecture on “Single Lead ECG Classification on Wearable and Implantable Devices” in tinyML forum to illustrate the present state of the direct and future research scope in Electrocardiogram-based algorithmic screening through wearables to detect different cardio-vascular diseases as part of smart healthcare ecosystem. He served as the panel moderator of ACM CIKM workshop 2020. He was one of the keynote speakers in the 30th IEEE International Conference on Advanced Information Networking and Applications (IEEE AINA). He was the steering committee member of HealthyIoT, 2016, 2017 and General Chair of KDAH workshop in ACM CIKM in 2018, 2019, 2020, and 2021. --------------------------------------------------- Hertfordshire Computer Science Research Colloquium http://cs-colloq.cs.herts.ac.uk