At CyberAgent AI lab, researchers have the chance to create seminars in which members would gather to carry out activities related to their job and improve their skills. After I entered the company, I decided to start a new seminar: the “research skill-up” seminar. The goal of this seminar is to polish our paper writing, research productivity and other management skills. Each session, each member would gather information about a topic of their interest and present it to the other members. In this blog post, I talk about one of the main topics discussed in the seminar: How to write scientific papers.
You May Also Enjoy
less than 1 minute read
A big majority of the machine learning systems existing in society assume that the data used for training and the data used after deployment follow the same distribution. However, in reality, these systems suffer a drop in performance due to the differences between training and test data. Multiple methods have been proposed in order to transfer the knowledge learned from a set of data to another set belonging to a different domain, but it is hard to achieve good generalization depending on the ``gap’’ between these domains. This blog post introduces an overview of domain adaptation (DA) techniques, why they are necessary, and different types and scenarios they can be applied to. I hope that, once readers understand these points, this blog post may serve as a starting-guide for surveying more specific methods (multimodal DA, etc.) that adapt to their needs.