Machine Learning and Statistical Inference

Welcome to the webpage of the Machine Learning and Statistical Inference Lab (MLSI-LAB), led by Dr Reshma Rastogi in SAU’s Department of Computer Science, which is a cross-disciplinary Machine Learning and Statistics lab.

Current Research Domains:

Quantum Machine Learning (QML):

Quantum computing is a rapidly-emerging technology that harnesses the laws of quantum mechanics to solve problems too complex for classical computers. These are equivalent to large classical computers, often with thousands of classical CPU and GPU cores. Basically, quantum computers can do billions of years’ worth of computing over the course of a weekend and untangle some of the world’s most complex problems in the process.

Quantum computing is preferred when the goal is to find the best decision out of many possible decisions such as optimization problems. Quantum Machine Learning (QML) is a growing field that brings together quantum computing and machine learning. The main premise of QML is to use the inherent advantages of quantum computing (including superposition, entanglement, and quantum parallelism) to improve the performance of classical machine learning algorithms. For example Quantum embedding kernels, helps quantum computers to encode data in ways that are difficult for classical machine learning methods.

The Field of Quantum Computing and Machine learning finds tremendous scope in Financial Services Industry. The field of financial services industry deals with huge amounts of data which increase with the speed of trades, transactions and data processing. Finding the best solutions requires solvers like Quantum computing, whose biggest potential use is in simulation. It helps identify a better way to manage risk in the financial world. Quantum computers can deal with the processing time and cost for high-quality solutions that increase exponentially if a classical computer is used. Thus increased optimisation capabilities, driving new cost savings and opportunities for revenue generation makes quantum computing a preferable choice for researchers and traders.

Applying emerging quantum technology to financial problems, particularly those dealing with uncertainty and constrained optimization, should also prove hugely advantageous to traders who are able to make calculations that reveal dynamic arbitrage possibilities that competitors are unable to see. Further, compliance employing behavioral data to enhance customer engagement, and faster reaction to market volatility are some of the specific benefits we expect quantum computing to deliver.

Multi-Label Classification (MLC):
Multi-Label is a sub field in pattern classification which requires an in depth understanding of traditional classification paradigms (i.e. binary and multiclass classification) in order to develop successful multi-label learning algorithms,Though multi-label learning is one aspect of machine learning, it inspires learning most of the machine learning algorithms, techniques ranging from various binary classification problems like supervised, semi supervised etc., to multi-class classification problems to popular algorithms like Google’s Page-Rank algorithm being adapted to multi-label learning. In short, multi-label learning is emerging as an umbrella domain that brings together various learning paradigms in machine learning, which one can find beneficial applications in multi-label learning. Our motive is to develop simple techniques and classifiers that can handle a broad spectrum of problems and generalize well by learning the properties of the data, hence eliminating the need for complicated techniques that modulate the same. We observed different machine learning tools, techniques, and algorithms working in the multi-label domain. Apart from this, other emerging domains branching from multilabel learning like multi-view learning, extreme multilabel classification, online multi label learning, statistical multi label learning etc., are also gaining attention. On the whole, multi-label classification is not only challenging but also very interesting. It has a broader range of applications and further extensions, thus making it an excellent field for study. One can explore the concepts and techniques that belong to some domains in machine learning and try to discover its creative applicability for multi label learning through innovative methods.

Efficient Machine Learning Based Data Integration Approaches in Health Care Applications:

Medical data include electronic health records (EHRs), medical images (such as X-rays, CT scans, and MRI scans), genomics data, vital signs, medication records, laboratory test results, and patient demographics. These data are generated by a variety of sources, such as hospitals, clinics, pharmacies, medical devices, and wearable technology. Integration of these data is one of the important aspects as it helps to get the holistic view to the patient outcome especially in the diagnosis and treatment. These data are often complex, heterogeneous and incomplete, which makes it difficult to perform meaningful analysis and visualization.

It is always a big challenge to integrate these data therefore appropriate mathematical techniques and machine learning algorithms need to be selected in order to overcome these shortcomings. Mathematical techniques such as linear algebra and graph theory are used to represent and manipulate the data and Machine learning algorithms can be used to identify patterns and relationships between different datasets. In many cases, labelled data is scarce hence making it difficult to train supervised machine learning algorithms. Thus, learning needs to be implemented in a semi-supervised machine learning framework. Traditional machine learning algorithms fail to learn the complex relationships between features and also to deal with noise and variations involved while data integration.

** For getting more information and updates regarding our research team please visit our lab website. Link: https://rebrand.ly/ml-lab

Machine Learning Workflow

Current Lab members

PhD Scholars

Masters Students

Alumni

Graduate PhD Sholars

Graduate MSc Students

Sl. No.	Thesis Title	Students Name	Programme	Year
1	Twin Support Vector Regression	Pritam Anand and M. Alam	MSc Computer Science, SAU	May 2015
2	Machine Learning Application to Human Activity Recognition	Sweta Sharma	MSc Computer Science, SAU	May 2015
3	Machine Learning Application in Big Data Analytics	Chettan Kumar	MSc Computer Science, SAU	May 2015
4	Active Learning via Twin Support Vector Machines	Nishu Nath and Shivanghi Singhal	MSc Computer Science, SAU	May 2016
5	Privacy Preservation via SVM and TWSVM	Jay Pratap Pandey and Lagneesh Pandey	MSc Computer Science, SAU	May 2016
6	Incremental Algorithms for Big Data	Amirjan Nepal	MSc Computer Science, SAU	May 2016
Maternity Leave 2016-2017
7	Extreme Multi label Classification	Puneet Kumar and Vikash Kumar Dubey	MSc Computer Science, SAU	May 2018
8	Variants of Incremental Twim Support Vector Machines	Saubhagya Kataria	MSc Computer Science, SAU	May 2018
9	Variants of Twin Extreme Learning Machines	Amisha Bharti	MSc Computer Science, SAU	May 2018
10	Omega Based Portfolio Optimization	Niharika Tiwari	MSc Computer Science, SAU	May 2018
11	Time Efficient Classifiers	Hamid Safdari	MSc Computer Science, SAU	May 2018
12	Personalized Intelligent Assistant	Ankit Kumar	MSc Computer Science, SAU	May 2018
13	Oblique Random Forest via NHCs	Anubhav David	MSc Computer Science, SAU	May 2019
14	Transfer Learning with Generative Adversarial	Ritesh Gagnani	MSc Computer Science, SAU	May 2019
15	Multi Task Learning Models and Extension	Mustaffa Hussain	MSc Computer Science, SAU	May 2019
16	Multi Label Learning : Models and Extension	Simran Popli and Nima Dorji	MSc Computer Science, SAU	May 2019
17	Multi-label Learning using Minimax Probability Machine	Sambhav Jain	MSc Computer Science, SAU	May 2020
18	Regularized Minimax Probability Machine	Upasana Pandey	MSc Computer Science, SAU	May 2020
19	Review of Label Specific Features Algorithms for Multi-Label Learning	Reza	MSc Computer Science, SAU	May 2020
20	Diagnosis Of Alzheimer’s Disease Using Machine Learning Techniques	Saloni Kawatra and Udhay Singh Tomar	MSc Computer Science, SAU	May 2020
21	Variants of Multilabel Learning with Missing label	Sayed Mortaza	MSc Computer Science, SAU	May 2021
22	Network Traffic Classification for Intrusion Detection using Incremental Twin Support Vector Machines	Ruhullah Yusufi	MSc Computer Science, SAU	May 2021
23	Joint Multi-Label Learning and Feature Selection with Missing Label using Local Linear Embedding	Nadira	MSc Computer Science, SAU	May 2021
24	Efficient variants of Random Forest based classifiers	Ashish Kunmar	MSc Computer Science, SAU	May 2021
25	Multi-Label Learning Using Variants of Extreme Learning Machine	Divyanshi Gupta	MSc Computer Science, SAU	May 2021
26	Addressing Class-Imbalance using Generative Adversarial Networks	Aatif Nisar	MSc Computer Science, SAU	May 2022
27	Entropy Value at Risk Based Portfolio Optimization and Diversification	Ajeet Yadav	MSc Computer Science, SAU	May 2022
28	Analysis of Risk Measures, Portfolio Optimization and Technical Indicator for Stock Market Prediction	Phuntsho Dema	MSc Computer Science, SAU	May 2022
29	Variants of Twin Support Vector Machine for Pattern Classification	Shuvo Saha Roy	MSc Computer Science, SAU	May 2022
30	Twin Support Vector Machine for Functional Data Classification	Muskan Singhal	MSc Computer Science, SAU	May

Projects in Collaboration with IIT, Delhi

Projects jointly supervised with Dr. Suresh Chandra (IIT Delhi) and Dr. Aparna Mehra (IIT Delhi)

2014: Semi Supervised Twin Support Vector Machines (M.Tech. Project)
2013: Twin Support Vector Regression (Mini M. Tech Project)

South Asian University

A University established by SAARC Nations