Dear Admission,
I am Peter. I grew up in a family with a strong football atmosphere, where my dad and I watched football games every weekend during football season. I have witnessed the immense growth of technology involved in the game. At that point, while most of the fans only followed the results, I became fascinated with how statistical analysis can analyse player and team performance on the pitch. Advanced metrics incorporating probability theorems and big data to evaluate player performance have been valuable as a more robust performance indicator. These advancements have fuelled my interest in the power of computer science in computing real-life problems. I was eager to spend most of my time in crucial subjects in math and computer science courses. I enjoyed the problem-solving process in providing solutions for needs encountered in various scenarios. I particularly enjoyed the experience of building an inventory and client management software for a Foreign Trading Company using the knowledge and practice skills I learned in class. The process includes multiple challenges, especially in the initial design stage, but overcoming these hurdles has been exceptionally fulfilling. This experience is a significant confidence boost for me in leveraging what I learn and solving real-life needs. Witnessing the wide use of wearable technology and machine learning techniques to analyse player physicality and performance has given me a strong interest in research about the intersection of health care and deep learning. Combining with reading on the impact of computer vision techniques on medical imaging, it gave me a strong interest in this interdisciplinary field. I discussed this with my current supervisor, Professor. xxx at the time, and my interests were strongly supported by him. It is also a confidence boost for me to conduct impactful research in computer science. These experiences kept me motivated in my endeavour in computer science. I firmly believe that a solid foundation in computer science, coupled with practical experience, equips me with the critical thinking and problem-solving skills necessary to contribute solutions, positively influence the healthcare domain, and many more.
As mentioned, my undergraduate thesis is on the intersection of computer science and digital pathology. It centred on scaling Graph Neural Networks for expansive graphs, particularly concerning their application in digital pathology. Utilising graph representation, we could capitalise on the cellular and tissue information inherent in histopathology images. My thesis proposed a hierarchical graph where nodes symbolised cells and tissues while edges represent their spatial relationship. With the intricate Whole Slide Images, the sheer number of entities leads to a substantial graph size for each slide being generated. Thus, efficiently processing these extensive graphs while retaining crucial information for superior captioning is vital. As such, I examined the effects of various coarsening methods and their enhancement in crafting detailed captions for histopathology images. With this coarsening method, the model can extract vital information from pathology images and produce high-quality captions.
Additionally, the pruned highly informative nodes can be incorporated into a visual explainer, which could provide a visual explanation for captions in the pathology image for pathologists. The thesis is a year-long work for my honour year in computer science, and my thesis work was completed under the supervision of Professor xxx and Dr xxx. The thesis project is divided into three terms. The first term is devoted to a detailed thesis topic proposal, a compelling abstract, and an exhaustive literature report, where I secured a score of 92/100 from my supervisors.
An initial deep dive into the digital pathology field gave me a robust understanding and passion to pursue it as my master's thesis project. Thus, for the master's thesis, I intend to concentrate on a more expressive graph representation approach for histopathology images. Earlier methods used entities (e.g. cells, tissues) from histopathology images and applied segmentation algorithms to determine entities' regions. They then employed pre-trained CNN-based methods to extract features. Further work can be done in extracting additional regions of interest other than cells and tissues to form a more robust representation of pathology images. The process involves detection, localisation, and selection of Region of Interest (ROI) in pathology images. With more effective ROI representation, it could be combined with my thesis research in Graph representation and propagation for a broader range of downstream tasks. I believe in the potential of having strong ROI representation and graph representation in medical imaging in general, in that a more robust representation method could be extended to other types of medical imaging such as CT scans and MRI. Thus, it is a project that I would be excited to work on for my master's thesis. Similar to how computer science in wearable technology gave me the inspiration and passion to work in the intersection of deep learning with health industry tasks, the incorporation of VAR (Video Assistance Referee) to act as an automatic referee raised my interest in computer vision, specifically video semantic understanding. I researched in sign language recognition, incorporating 2D Contrastive Predictive Coding as an unsupervised feature extractor to achieve SOTA performance. I want to extend into more general video and multimodal tasks in my master's thesis. I aim to grow on the current unsupervised method and extend to a more robust spatial-temporal feature extraction method that could be effective in a wider range of video understanding tasks. An effective video feature extraction method can serve wide benefits, including video editing, deepfake, and misinformation detection. I am passionate about both hypothetical research and thrilled to contribute to the academic environment at TU Delft through my work on this platform.
Due to my interest mentioned above, I am most profoundly captivated by the Artificial Intelligence Technology Track among the specialisations offered by the program. The profound impact I believe Artificial Intelligence can have on society fuels my strong desire to research AI's potential to provide valuable solutions. Throughout my bachelor's and honours degree, I proactively pursued multiple elective courses in machine learning, focusing on building a robust foundation in mathematics, traditional machine learning, and advanced design of deep learning models." Working with researchers at UNSW exposed me to more practical aspects of researching deep learning methods and providing solutions to modern problems in various fields. These experiences and interactions have built my passion in pursuing this track of computer science. I am excited to build on my knowledge and research experience in AI Technology. There are intriguing core courses in the Artificial Intelligence Technology Track on areas that I would be excited to explore and gain insights into, such as Information Retrieval and Algorithms for Intelligence Decision Making, as well as topics that I am familiar with but would hope to dive further into, such as Deep Learning and Artificial Intelligence Techniques. The Artificial Intelligence Technology Track could provide me with more knowledge and practical experience in AI to assist me with the technical skills to be at the forefront of AI's development. Given my research background, I am confident that my interests and experience will make me an asset to research projects within the AI domain of the MSc in Computer Science.
Building on my experiences and enthusiasm for cutting-edge research and exceptional education has naturally led me to consider institutions that align with my aspirations. TU Delft's reputation for outstanding education and cutting-edge research, especially in computer science, has long captured my interest. Its ethos of imparting core knowledge while fostering student innovation aligns with my values, making it my preferred choice. I recognise that TU Delft stands at the forefront of academic research, and I yearn to be part of its research-centric ecosystem where foundational understanding, skills, and ingenuity are cherished. Furthermore, TU Delft's robust affiliations with esteemed research institutions across Europe and globally ensure a platform for expansive networking with global researchers and academics. In essence, pursuing my master's at TU Delft will empower me with in-depth knowledge, rich research experiences, and a valuable global network, setting the stage for a promising academic research career.
In conclusion, I am enthusiastic about the chance to apply to the MSc in Computer Science at TU Delft. With my extensive research experience and a robust theoretical foundation, I can be an excellent match and contribute meaningfully to the outstanding research environment at TU Delft.
I am Peter. I grew up in a family with a strong football atmosphere, where my dad and I watched football games every weekend during football season. I have witnessed the immense growth of technology involved in the game. At that point, while most of the fans only followed the results, I became fascinated with how statistical analysis can analyse player and team performance on the pitch. Advanced metrics incorporating probability theorems and big data to evaluate player performance have been valuable as a more robust performance indicator. These advancements have fuelled my interest in the power of computer science in computing real-life problems. I was eager to spend most of my time in crucial subjects in math and computer science courses. I enjoyed the problem-solving process in providing solutions for needs encountered in various scenarios. I particularly enjoyed the experience of building an inventory and client management software for a Foreign Trading Company using the knowledge and practice skills I learned in class. The process includes multiple challenges, especially in the initial design stage, but overcoming these hurdles has been exceptionally fulfilling. This experience is a significant confidence boost for me in leveraging what I learn and solving real-life needs. Witnessing the wide use of wearable technology and machine learning techniques to analyse player physicality and performance has given me a strong interest in research about the intersection of health care and deep learning. Combining with reading on the impact of computer vision techniques on medical imaging, it gave me a strong interest in this interdisciplinary field. I discussed this with my current supervisor, Professor. xxx at the time, and my interests were strongly supported by him. It is also a confidence boost for me to conduct impactful research in computer science. These experiences kept me motivated in my endeavour in computer science. I firmly believe that a solid foundation in computer science, coupled with practical experience, equips me with the critical thinking and problem-solving skills necessary to contribute solutions, positively influence the healthcare domain, and many more.
As mentioned, my undergraduate thesis is on the intersection of computer science and digital pathology. It centred on scaling Graph Neural Networks for expansive graphs, particularly concerning their application in digital pathology. Utilising graph representation, we could capitalise on the cellular and tissue information inherent in histopathology images. My thesis proposed a hierarchical graph where nodes symbolised cells and tissues while edges represent their spatial relationship. With the intricate Whole Slide Images, the sheer number of entities leads to a substantial graph size for each slide being generated. Thus, efficiently processing these extensive graphs while retaining crucial information for superior captioning is vital. As such, I examined the effects of various coarsening methods and their enhancement in crafting detailed captions for histopathology images. With this coarsening method, the model can extract vital information from pathology images and produce high-quality captions.
Additionally, the pruned highly informative nodes can be incorporated into a visual explainer, which could provide a visual explanation for captions in the pathology image for pathologists. The thesis is a year-long work for my honour year in computer science, and my thesis work was completed under the supervision of Professor xxx and Dr xxx. The thesis project is divided into three terms. The first term is devoted to a detailed thesis topic proposal, a compelling abstract, and an exhaustive literature report, where I secured a score of 92/100 from my supervisors.
An initial deep dive into the digital pathology field gave me a robust understanding and passion to pursue it as my master's thesis project. Thus, for the master's thesis, I intend to concentrate on a more expressive graph representation approach for histopathology images. Earlier methods used entities (e.g. cells, tissues) from histopathology images and applied segmentation algorithms to determine entities' regions. They then employed pre-trained CNN-based methods to extract features. Further work can be done in extracting additional regions of interest other than cells and tissues to form a more robust representation of pathology images. The process involves detection, localisation, and selection of Region of Interest (ROI) in pathology images. With more effective ROI representation, it could be combined with my thesis research in Graph representation and propagation for a broader range of downstream tasks. I believe in the potential of having strong ROI representation and graph representation in medical imaging in general, in that a more robust representation method could be extended to other types of medical imaging such as CT scans and MRI. Thus, it is a project that I would be excited to work on for my master's thesis. Similar to how computer science in wearable technology gave me the inspiration and passion to work in the intersection of deep learning with health industry tasks, the incorporation of VAR (Video Assistance Referee) to act as an automatic referee raised my interest in computer vision, specifically video semantic understanding. I researched in sign language recognition, incorporating 2D Contrastive Predictive Coding as an unsupervised feature extractor to achieve SOTA performance. I want to extend into more general video and multimodal tasks in my master's thesis. I aim to grow on the current unsupervised method and extend to a more robust spatial-temporal feature extraction method that could be effective in a wider range of video understanding tasks. An effective video feature extraction method can serve wide benefits, including video editing, deepfake, and misinformation detection. I am passionate about both hypothetical research and thrilled to contribute to the academic environment at TU Delft through my work on this platform.
Due to my interest mentioned above, I am most profoundly captivated by the Artificial Intelligence Technology Track among the specialisations offered by the program. The profound impact I believe Artificial Intelligence can have on society fuels my strong desire to research AI's potential to provide valuable solutions. Throughout my bachelor's and honours degree, I proactively pursued multiple elective courses in machine learning, focusing on building a robust foundation in mathematics, traditional machine learning, and advanced design of deep learning models." Working with researchers at UNSW exposed me to more practical aspects of researching deep learning methods and providing solutions to modern problems in various fields. These experiences and interactions have built my passion in pursuing this track of computer science. I am excited to build on my knowledge and research experience in AI Technology. There are intriguing core courses in the Artificial Intelligence Technology Track on areas that I would be excited to explore and gain insights into, such as Information Retrieval and Algorithms for Intelligence Decision Making, as well as topics that I am familiar with but would hope to dive further into, such as Deep Learning and Artificial Intelligence Techniques. The Artificial Intelligence Technology Track could provide me with more knowledge and practical experience in AI to assist me with the technical skills to be at the forefront of AI's development. Given my research background, I am confident that my interests and experience will make me an asset to research projects within the AI domain of the MSc in Computer Science.
Building on my experiences and enthusiasm for cutting-edge research and exceptional education has naturally led me to consider institutions that align with my aspirations. TU Delft's reputation for outstanding education and cutting-edge research, especially in computer science, has long captured my interest. Its ethos of imparting core knowledge while fostering student innovation aligns with my values, making it my preferred choice. I recognise that TU Delft stands at the forefront of academic research, and I yearn to be part of its research-centric ecosystem where foundational understanding, skills, and ingenuity are cherished. Furthermore, TU Delft's robust affiliations with esteemed research institutions across Europe and globally ensure a platform for expansive networking with global researchers and academics. In essence, pursuing my master's at TU Delft will empower me with in-depth knowledge, rich research experiences, and a valuable global network, setting the stage for a promising academic research career.
In conclusion, I am enthusiastic about the chance to apply to the MSc in Computer Science at TU Delft. With my extensive research experience and a robust theoretical foundation, I can be an excellent match and contribute meaningfully to the outstanding research environment at TU Delft.
