Research Statement

Introduction

Research is a complex system that includes a variety of elements, not all of which are immediately apparent. The tangible aspects, such as methodologies, findings, results, and accolades, are attractive and entice many to the field. Yet, the intangible qualities like integrity, passion, determination, and patience are equally crucial. Groundbreaking research, in my view, is characterized by the ability to integrate these tangible and intangible elements throughout the journey of becoming a researcher. In 2014, I embarked on a journey with unwavering determination to establish the right foundation for becoming a proficient researcher, guided by this ultimate perspective. In the subsequent sections, I will first present my previous and ongoing research contributions, and then outline my future research plans and visions.

Previous Research Contributions

Artificial Intelligence (AI) is now indispensable and highly effective in addressing a wide range of challenges in various domains. This field have advanced significantly over time due to substantial research and contributions. At the beginning of my Ph.D. in Management Information Systems (MIS), I made a commitment to focus on mastering AI. While this may seem broad, this commitment marked the initial milestone of my research aspirations and journey. Consequently, I have embarked on an extensive exploration of the subject, delving into a variety of materials with a particular emphasis on grasping its mathematical underpinnings. In its early stages, AI, then referred to as ML, concentrated on addressing prediction issues using the mathematical structure of the classical linear algebra equation Ax = b. My initial research endeavor involved comprehending the operational procedure for resolving the problem grounded in this mathematical representation. Throughout this period, I authored several reports summarizing my discoveries, aiming to both share knowledge and deepen my own understanding. These summaries delineated a clear distinction, categorizing prediction problems into three types: regression, classification, or clustering. Regression and classification involve predicting a specific label, whereas clustering deals with scenarios where no particular label is predicted, typically known in the field as supervised and unsupervised learning, respectively. These reports gained local popularity among scholars and postgraduates from my department and others. This recognition provided a significant boost to my self-confidence and commitment, motivating me to work harder and continue learning.

As a Ph.D. candidate, I honed my research to explore the expanding realm of Artificial Intelligence (AI), particularly deep learning. My goal was to discern the nuances and progressions in contemporary models via a mathematical framework. Armed with an extensive grasp of the operations from both theoretical and mathematical standpoints, I dedicated myself to executing an exhaustive research survey. The survey aimed to provide a comprehensive report on abstract concepts, mathematics, and the diverse tools and frameworks related to AI. Designed for publication on a local blog, it ensures public accessibility. The IT department of ISMEK (İstanbul Büyükşehir Belediyesi, ¨ 1996), an educational entity managed by the Istanbul Metropolitan Municipality in Türkiye, has ¨ recognized this effort. Consequently, I was presented with a project opportunity for my Ph.D. thesis, which entails creating a machine learning algorithm to detect lifelong learners at risk of discontinuing their courses. This marked my initial professional venture into conducting comprehensive research based on a robust grasp of AI. The project fundamentally dealt with a classification issue, particularly a binary classification, which provided the foundation for me to explore in depth and make an innovative contribution in the ensuing years.

Ongoing Research Contributions

Upon completing my project and Ph.D., I chose to update my previous summary guide on AI. This choice stemmed from the prevalent confusion and misconceptions about AI that I have witnessed in both academic and industrial realms throughout the course of my project and thesis work. Such misunderstandings often arise from a fundamental lack of knowledge in the mathematics underpinning AI. For example, it is a common misconception that probability and statistics belong to the same category of research, whereas I have delineated that probability is merely an operational branch within the AI, while statistics is a distinct discipline aimed at developing descriptive models. This marked my second effort to make a professional contribution; I chose to present the AI summary guide, but in a novel way, by initiating an initiative that encompasses projects and publications, among other elements. Specifically, I have focused on designing, authoring, and creating educational digital content titled” The Big Bang of Data Science.” (Dahman, 2023a) This content encompasses five key elements: research, analysis, prediction, coding, and AI-embedded systems, covering the entire process from beginning to end. In this revised edition of the summary, I have enhanced the methodical presentation of each section. For instance, ’The Map of Prediction’ now clearly delineates each algorithm utilized by the AI, along with the numerous frameworks and tools that implement them, presented in a lucid flowchart and organized under specific titles. As a result, the audience can now comprehend the subject matter and the practical application of AI more effectively.

Meanwhile, as a self-employed professional, I have developed a multitude of projects and undertaken diverse research in various fields, examples of which are showcased in my resume. My work has consistently focused on employing AI problem-solving approaches to address different challenges, through the optimization and selection of the most suitable algorithm for each problem. This approach has markedly benefited my career trajectory as a researcher. The deep comprehension of mathematics that underpins AI, coupled with its application in diverse fields, has led me to focus on two particular research areas in my current studies. The first is the classification problem. I view this issue as a legitimate choice for researchers aiming to develop a solution, even if the problem tends toward regression. There are many established methods and algorithms for tackling classification problems. Yet, by utilizing my solid understanding of linear and non-linear algebra, as well as various mathematical theories, I have created an innovative algorithm for solving classification issues, which I call the ”BireyselValue” algorithm (Dahman, 2023b). This algorithm is based on the idea that the unique attributes of a class can determine an observation’s classification through similarity metrics. Experimental results from six multiclass datasets in different sectors indicate that this method is not just effective in solving classification problems but may also be a strong candidate for feature reduction. The second area of focus is data privacy and AI security. While privacy is a distinct research field from security, delving into one often leads to the other. My initial research aimed to understand how AI models can be attacked. A solid grasp of AI’s mathematical underpinnings enabled me to explore various vulnerabilities, particularly during the continuous training and learning stages of the algorithm. My attention has been particularly drawn to a type of attack known as poisoning attacks. Consequently, I have developed a new method called ’norm culture.’ (Dahman, 2024a) This method is designed to protect AI learning models from data and label poisoning attacks. It posits that each category in an image classification task possesses an inherent structure that serves as a natural defense against such attacks, potentially corrupting new training and testing samples during the parameter update phase of an AI predictive model in a conventional deep learning environment. Experimental results on a binary class image classification dataset from the healthcare sector indicate that the method effectively identifies compromised training and testing sample images from both attack types. Furthermore, there is potential for enhancement within the mathematical functions of the AI framework.

Following this work, I have advanced the research on data, privacy, and AI. Specifically, I have authored a comprehensive guide titled ”Review of Data Privacy Techniques: Concepts, Scenarios and Architectures, Simulations, Challenges, and Future Directions.” (Dahman, 2024b) The review seeks to provide a comprehensive perspective on the subject. It proposes a systematic approach that links three key elements: data, AI, and privacy. The review explores each aspect in detail, both theoretically and practically, showcasing the most recent methods for tackling data privacy issues through various lab simulations. Additionally, it suggests tools and resources for advanced exploration. In conclusion, the work summarizes the core theme by highlighting the challenges and potential avenues for future research

Future Research Plans and Visions

My previous and ongoing work and research have ignited my passion for my future research trajectory. I am keen to continue addressing the optimization challenges that arise in the selection of the appropriate algorithms, tools, or methods by AI professionals. In particular, I plan to refine my guide further, integrating both current and future improvements. This will aid professionals in choosing the most fitting algorithms, tools, and methods. Additionally, I aim to delve deeper into classification issues, especially by developing the second edition of my proposed algorithm, ”The BireyselValue.” I am also seeking opportunities to apply and enhance this method to more efficiently solve classification problems and possibly address feature reduction issues. Lastly, I intend to expand my exploration of privacy and security in AI implementation. Specifically, I plan to develop the second edition of my” Norm Culture” method, where I will incorporate the method’s mathematical term into the AI’s mathematical functions, making it a fundamental part of the AI black box’s mathematical operations

References

• Dahman, D. (2023a). The big bang of data science. GitHub Repository. Retrieved from https://github.com/dahmansphi/big bang of data science project
• Dahman, D. (2023b). The bireyselvalue algorithm. GitHub Repository. Retrieved from https://github.com/dahmansphi/bireyselvalue v1
• Dahman, D. (2024a). ”the norm culture” advocates for the introduction of a security layer in continuously learning ai models to protect against data and label poisoning attacks. ScienceOpen. Retrieved from https://www.scienceopen.com/hosted-document?doi=10.14293/PR2199.000907.v1 doi: 10.14293/PR2199.000907.v1
• Dahman, D. (2024b). Review of data privacy techniques: Concepts, scenarios and architectures, simulations, challenges, and future directions. Data Science and Machine Learning. Retrieved from https://www.authorea.com/doi/full/10.22541/essoar.171995147.78801330/v1 doi:10.22541/essoar.171995147.78801330/v1
• İstanbul Büyükşehir Belediyesi. (1996). ¨ ˙Ismek. In Sanat ve meslek egitimi kursları. ˘ Retrieved from https://enstitu.ibb.istanbul/portal/egitim dallari.aspx?dalId=15