Krikamol's research focus is primarily on machine learning methods. "In doing so, I face three major challenges. First, learning algorithms need to get better and be flexible to deal with changes in the observed data or the data acquisition process. Second, large amounts of experimental data are needed to make machine learning as effective as possible. Collecting these is expensive, time-consuming, and sometimes even unethical. As a result, many models are still trained on non-experimental data, making them less effective. This needs to change. And thirdly, the ubiquitous presence of AI systems could lead to people interacting with them in a completely different way in the future and exploiting their mechanisms in a targeted manner. This changes the entire database, threatening the effectiveness of the model. How to account for this when training the model remains an open problem."
At CISPA, Krikamol hopes to find the answer to some pressing research questions through interactions with his new colleagues. "One research direction that has long fascinated me is machine learning on heterogeneous data." Krikamol says this refers to data that comes from multiple sources. "For example, it could be medical data that comes from patients at different hospitals," the researcher continues. "If we want to train an ML model with a heterogeneous data set, we need to be able to make meaningful comparisons between the environments from which the data come. This is only possible if information can be exchanged securely between these environments all the time. I hope my new colleagues at CISPA will help me address this challenge," says Krikamol, a native of Thailand. "I look forward to helping shape the future of research at CISPA."
