Today is Monday, and the summary of this week is also coming late. Let's review the objectives from last week:
Objectives from last week#
- Start reading and thinking about the Xclip and VOLm papers, and consider how to integrate the video modality into VOLm or BEiT.
- Communicate with Eugenio about the projects in the lab and decide whether to accept them.
- Complete the article on Memory, review the process of memory formation, and explore the most suitable learning methods.
Review#
Objectives 1 and 2 from last week have been completed. This week, while translating the paper, I found a paper that is very suitable for the ProSieben project. This paper is based on Clip, it is from CVPR 2023, and the code is open source. It is very suitable as a baseline for the ProSieben project. While working, I have also gradually summarized a method for training a Video Modality starting from the MOME architecture. I hope to conduct some experiments on the new architecture next week, play around with it, and improve my pipeline based on his training method to see if there is anything I haven't considered or worth considering.
The second thing is that I have finalized the graduation thesis this week. I have also started looking for a house in Freiburg. By the way, I thought Freiburg was a small city and it would be easier to rent a house there. I didn't expect that Freiburg is the second most difficult place to rent a house in Germany, just after Munich. I took a look at the prices this week, and it seems that the idea of saving on housing prices in a small city has been shattered. Although the prices are still lower than in Munich, the difference is not much.
This week, I also found a paper on the direction of my thesis. I received a confirmation email from Eugenio on Friday and started looking for papers based on this topic. But unfortunately, I haven't read these papers carefully yet. Eugenio and I plan to complete a proposal within two weeks. I have many ideas now, and I may want to do some Causality Integration, which may be challenging. Another task is to choose a promising backbone, and I have found some good options. I need to preliminarily assess the feasibility of this new backbone based on the content of the PaLM-E paper. So this week, I hope to read at least 3 papers on the thesis direction. It would be great if I can produce some output as well.
Another time-consuming task this week is the HSA assignment. Yesterday (Sunday of the second week), Yi Yi finally helped me solve the problem. For some unknown reason, with the usbcore.autosuspend=2 kernel parameter set as default, my laptop still automatically suspends the connection between the board after a few seconds. However, the same parameter does not exhibit this behavior on desktop computers or other people's laptops. I believe this investigation is quite interesting, but time is limited, so I can only admire Yi Yi's understanding of the Linux system.
Other Interesting points#
- I started having deep conversations with my loved one on a weekly basis, and this is the first week. It feels great, and I hope we can continue this in the future.
- The rainy season in Munich should be over, so I can start jumping rope again. It's important to maintain basic exercise every day.
- I solved the domain name resolution issue for lumeny.io. It turned out to be a small problem with Cloudflare's settings.
- I started using the baserow service, which allows me to manage my finances and contacts through a database. It has better scalability than previous methods. It serves as a supplement to my plain text notes and personal management.
New Desires#
- Continue jumping rope every morning.
- Continue practicing meditation.
- Want to invest more time in reading papers.
Minor things
- Cancel the gym membership, at least by July.
- Contact Vivian about moving out. I really like this place and my roommates, but I still have to leave.
- Merge the baserow databases into a personal one to provide more connections for the important aspects of my life.
Goals for 2023-W18#
- Jump rope daily & continue daily meditation.
- Thesis: Explore three important papers.
- Complete the HSA assignment, at least get a basic understanding of embedded systems -> Complete T1 report.