Week 9 & 10
General Experience
Well, these are my last two weeks of this amazing research experience. During Week 9, I was refining the results, trying out different methods to improve our model, testing different layers with the set transformer, and implmenting different experiments. In the process of writing my technical report, I have read additional papers discussing similar algorithms and approaches.
Readings
During Week 10, I have read: 1 - A survey of unsupervised deep domain adaptation to learn more about deep domain adaptation and the process of constructing algorithms that are robust to different domains and different dataset distributions 2 - Dataset Shift in Machine Learning. I have also consulted this resource to understand distribution shifts and the different kinds of algorithms that are implemented to deal with distribution shifts. During the process of reading and summarizing previous papers I have read, I was able to establish a clear connection between our work and previous works that are in the domain of causal inference.
Experiments and Results
These weeks were spent in summarizing the results. The major findings were a pattern of behavior that we discovered when it comes to the properties of test and source domains. How does the test behave if certain conditions of the source domain (training groups) were met. More on that in the Final Report
Closing
This summer has been challenging, rewarding, frustrating, and exciting. I have learned loads of information, techniques, skills, and tools. I didn’t expect to learn that much when I first started. I was able to create a neural network, adjust a loss function, install different packages, solidify my python coding skills, reform my software development skills, refine my explanation and technical writing skills, and investigate open problems.