Last month was so busy I couldn't even do proper input... This month I'm gradually increasing time for data analysis (by forcibly reducing time spent on other work), so I'm getting better at input.
Since I'm mostly implementing routine analysis automation on weekends, it's frustrating that I can't find time for new studies like single cell analysis and transformer models... I want to get the automation done this month somehow.
An article by suin-san about how connecting via SSH every time takes time, so let's skip authentication once a connection succeeds. I tried the settings and it's visibly faster. About 500ms faster for connecting to AWS EC2.
A video about resetting when you get this error in VSCode because DNS server resolution is stuck. I referenced it as it was an error I encountered for the first time, but it didn't solve it.
Actually, the root cause was that Azure services went down on 7/19...
Lecture materials by Professor Kota Matsui of Nagoya University. I just skimmed through and wrote out the formulas for parts that interested me, but the transformation showing that the posterior distribution becomes normal again when using normal distribution as prior on p.32-33 was quite interesting.
Bayesian optimization is fascinating.
They also published a book last year. Since Bayesian optimization can also be used for deep learning hyperparameter search, I need to study it properly somewhere.
A list of information collection sources by Ogawa-san of Matsuo Research Institute. It looks great but most links are broken. Seems to be due to service provider issues.
A crazy study where they artificially infected healthy donors with SARS-CoV2, but susceptibility to infection correlated with HLA-DQA2 expression providing infection protection.
So HLA Class 2 really is involved in infection protection...
LabCode-san's HP explanations are very high quality and I often reference them, but I didn't know such a skilled chemoinformatics person was editing...
I read this for the first time and it was very thought-provoking, a useful post.
A wonderful encouragement post from an MIT graduate to a high school student near graduation who had been getting straight A's but got one B near graduation and didn't do well with university admission, lamenting "I wasn't as smart as I thought I was."
Just reading it gives you courage, and conversely when facing difficulties, it makes you feel that if you find a solution (new way), it will work.
I sometimes write bash code blocks in regular markdown to keep logs. Copying and pasting these to execute is extremely tedious, but using the runme extension in VSCode makes markdown have the same interface as Jupyter Notebook, and you can execute on the spot (amazing).
That's what I thought when I installed and used it, but in the end, cases where I want to check the file tree on terminal are rare, and basically as long as I'm working in VSCode, it doesn't get much use.
If you often type ls or pwd 3+ times for a single work purpose, Yazi will improve your work efficiency.
The Vim key-bindings are really nice.
##Books Read
###Alzheimer's Disease Research, Structure of Failure - Karl Herrup, Ayumi Kajiyama (Translation)
A book I had been wanting to read for a long time.
The author, who researched from the position that amyloid plaques are not the cause of Alzheimer's disease during the era when Alzheimer's disease research was all about amyloid β research, explains how Alzheimer's disease research progressed through history to today.
First, just having an explanatory book by a researcher who understands Alzheimer's disease science to this extent is appreciated. It broadly explains the story of Alzheimer's disease research that should be noted, including how APOE gene is a risk factor for onset, and how inflammation in the brain was a promising hypothesis but failed in clinical trials. Of course, including the recent Lecanemab.
However, to put it bluntly, the author is an extreme anti-amyloid β hypothesis theorist, so throughout there's tremendous critical perspective toward the opposing research group. This stems from experiences like being told by grant reviewers "if it's not amyloid research, it's not Alzheimer's disease research," so I can understand becoming anti... but still, I found it too persistent.
Also, being a translated book, it inevitably has the persistent repetition characteristic of Western books and somewhat awkward translation, making it not the easiest read.
That said, reading it while thinking there's probably considerable bias makes it a good book for understanding the history of Alzheimer's disease research. Essential if you want to know about Alzheimer's disease.
###The Best Study Methods Based on Scientific Evidence
There are mountains of study method books, but this one was displayed prominently at bookstores so I was curious.
Right when I was reading this book, I was frustrated about not being able to do input as I wanted, so I read it along with the next book "How to Read Technical Books" seeking input methods.
This book is a published version of a video by U.S. internal medicine specialist Kosuke Yasukawa that has 3.55 million views on YouTube (as of August 2024). As the title suggests, it presents multiple pieces of research evidence and logically explains in an easy-to-understand manner how classic study methods like "copying the textbook verbatim" are inefficient.
What this book particularly emphasizes is repeated learning and active recall. It's well known that learning is retained through regular review, but using research examples, it introduces methods like "first repetition after X weeks, second time after..." Active recall is a method of trying hard to remember with minimal hints, like doing your best to remember without looking at anything on a crowded train. Apparently, stressing the brain helps memory retention. Seems like it might be good for Alzheimer's prevention too.
If you need to do a substantial amount of studying for university entrance exams or some certification, I think you should definitely read this book. It's very light reading so anyone should find it accessible. Highly recommended.
###"How to Read Technical Books"
This book had also been a topic of discussion for a while and I was curious. I accidentally bought it both on Kindle and as a physical book because I had too many unread books. As mentioned below, it has many diagrams and photos and is a graphical book, so the physical version is recommended.
While the content is specialized for "technical books," I personally think it's a masterpiece that explains reading methods that apply to all books. I hope people who aren't good at reading will also read it. You'll fly.
What was particularly impressive about this book was "you don't have to try to read the whole book." With technical books, you often have specific information you want before reading, and it's totally fine to skim and read only those specific parts for information collection purposes. Knowing this really lowers the barrier to reading.