Väitös (tietojenkäsittelytiede): D.Sc. Jatin Kumar Chaudhary
D.Sc. Jatin Kumar Chaudhary esittää väitöskirjansa ”Algorithmic Foundations for Generalizable Artificial Intelligence Models: A Multi Domain Study” julkisesti tarkastettavaksi Turun yliopistossa torstaina 14.8.2025 klo 12.00 (Turun yliopisto, Agora, XXI-luentosali, Turku).
Vastaväittäjänä toimii professori Pekka Toivanen (Itä-Suomen yliopisto) ja kustoksena tohtori Rajeev Kanth (Turun yliopisto). Tilaisuus on englanninkielinen. Väitöksen alana on tietojenkäsittelytiede.
***
Tiivistelmä väitöstutkimuksesta:
This dissertation explores how smart computer algorithms can be used to solve real-world problems in healthcare, clean energy, and artificial intelligence. The work combines three areas: designing better solar panels, improving early cancer diagnosis using medical images, and making machine learning algorithms smarter and more reliable.
1. What are the key findings of your dissertation research?
One major finding is that machine learning can be used to speed up the discovery of new materials for solar panels. Normally, testing whether a material works well in a solar cell takes time and complex lab experiments. This research shows that computer models can predict material performance accurately just by looking at basic physical properties. This was demonstrated on a group of materials called perovskites, which are promising for next-generation solar cells.
In the healthcare part of the work, the research shows that AI can help detect cancer from MRI images. The study used a technique called radiomics, which turns images into numerical data that machines can analyze. A foundation model was built using large datasets of prostate MRI scans. The model was tested across different hospitals and machines and proved to be both reliable and adaptable. It can even be fine-tuned to work well in new hospitals with only a small number of patients.
Finally, the dissertation introduces a new type of algorithm that improves how machine learning models learn over time. Traditional training methods can be unstable or get stuck. This work proposes an approach using exponential decay and dynamic learning rates to keep training smooth and stable. It also introduces a way to measure model confidence and identify when a model might be overfitting or misled by noisy data.
2. What is the impact of your research on the surrounding world?
This research has practical benefits in three important areas:
- Clean energy: By helping researchers discover better solar cell materials faster and more cheaply, the work supports the global move toward renewable energy.
- Medical diagnosis: The AI model developed here can help doctors detect cancer earlier, using standard MRI images. This could reduce delays in diagnosis and ensure more people get the treatment they need, especially in hospitals with fewer resources.
- Smarter AI: The new training method for AI models helps make them more reliable, especially in difficult tasks or when working with limited data. This is important as AI is being used more widely in critical areas like medicine, finance, and transportation.
Vastaväittäjänä toimii professori Pekka Toivanen (Itä-Suomen yliopisto) ja kustoksena tohtori Rajeev Kanth (Turun yliopisto). Tilaisuus on englanninkielinen. Väitöksen alana on tietojenkäsittelytiede.
***
Tiivistelmä väitöstutkimuksesta:
This dissertation explores how smart computer algorithms can be used to solve real-world problems in healthcare, clean energy, and artificial intelligence. The work combines three areas: designing better solar panels, improving early cancer diagnosis using medical images, and making machine learning algorithms smarter and more reliable.
1. What are the key findings of your dissertation research?
One major finding is that machine learning can be used to speed up the discovery of new materials for solar panels. Normally, testing whether a material works well in a solar cell takes time and complex lab experiments. This research shows that computer models can predict material performance accurately just by looking at basic physical properties. This was demonstrated on a group of materials called perovskites, which are promising for next-generation solar cells.
In the healthcare part of the work, the research shows that AI can help detect cancer from MRI images. The study used a technique called radiomics, which turns images into numerical data that machines can analyze. A foundation model was built using large datasets of prostate MRI scans. The model was tested across different hospitals and machines and proved to be both reliable and adaptable. It can even be fine-tuned to work well in new hospitals with only a small number of patients.
Finally, the dissertation introduces a new type of algorithm that improves how machine learning models learn over time. Traditional training methods can be unstable or get stuck. This work proposes an approach using exponential decay and dynamic learning rates to keep training smooth and stable. It also introduces a way to measure model confidence and identify when a model might be overfitting or misled by noisy data.
2. What is the impact of your research on the surrounding world?
This research has practical benefits in three important areas:
- Clean energy: By helping researchers discover better solar cell materials faster and more cheaply, the work supports the global move toward renewable energy.
- Medical diagnosis: The AI model developed here can help doctors detect cancer earlier, using standard MRI images. This could reduce delays in diagnosis and ensure more people get the treatment they need, especially in hospitals with fewer resources.
- Smarter AI: The new training method for AI models helps make them more reliable, especially in difficult tasks or when working with limited data. This is important as AI is being used more widely in critical areas like medicine, finance, and transportation.
Viestintä