I am a PhD fellow in Informatics at the University of Oslo, Norway. I have previously taken a MSc in Renewable Energy at Oldenburg University, Germany. My current work focuses on applying machine learning in smart grids. I participated in many energy informatics research projects and I am a Marie-Curie felloship alumnus

My Projects

Project 4: Explainable Medical Image Segmentation via GANs and Layer-wise Relevance Propagation

I teamed up with my 7-year-old daughter, Leen, to contest in MedAI: Transparency in Medical Image Segmentation Challenge. We had a lot of fun in making the computer to learn to segment medical images and explain how it came up with predictions. We also managed to submit a short working paper to Nordic Machine Intelligence open access journal. Personally, I was amazed by how the layer-wise relevance propagation method (LRP) was able to specify which pixels in the input are more relevant to the prediction and to what extent! Especially with our GAN models which seemed to go through “chaotic” model training process and I did not expect to get those meaningful relevance maps! A main takeaway of this is that: (Model Accuracy + Model Explainability = Authentic Value). Here is our working paper and the code. Our team (called Leen) got an honorable mention too!!

Server IP: