Thesis Week 6 (Spring)

Despite the lack of blog posts, I have been working this whole time, and sending updates to Dr. Yampolskiy. Sending him weekly updates was redundant with making these updates, but now I’m feeling very serious about doing this so making both feel more appropriate.

I’m also committing to a 4-day thesis work week - Monday-Tuesday, Thursday-Friday. On those days, I’ll drink caffeine (which I usually don’t like) with L-Theanine for a few weeks top see if that regiment will help me work. Having fixed work hours will help to. 10-12am + 1-3pm is a lot of time, but as long as I’m taking this seriously should be doable. I should have done this sooner.

Week 13 (April 9th) is the unofficial due date of my thesis. That means I have 7 full weeks to do this, while staying motivated and productive. I think before now I was imagining this thing being 10-15 pages and just taking it easy. Not anymore, I’ve surveyed the thesis submitted by UofL graduate students and found they’re 50-80 pages. And my topic is much “softer” than most so the expectation is higher.

Now I wish I was doing this on ISC. I could totally write a thesis about that. But alas, indecision and regret are both fatal so instead I’ll dream about that as being my PhD thesis someday. I can do 150 pages worth of research about indefinitely scalable computing, especially with the hardware to scale it up, an electrical engineer to make hardware work, and an undergraduate assistant to pair program with. But alas, that’s a far off dream and right now it’s just a distraction from the stress of my current, real thesis.

Anyway, isn’t this what I dreamed about a few years ago? Making a useful work on AI safety in a practical way? I’m exactly where I want to be.

Work log:

An explainable sequence-based something rather

It’s clear how they get so many pages now. The text is VERY spaced out, and they have a huge gap between paragraphs. In this whole paper there are only 5 pages of novel content, which would fit into 2 pages of IEEE format minus the pictures.

Automatic IQ estimation using stylometry methods

(yikes, that sounds like a really bad idea, for classism reasons)

Anyway, they have double spaced text (maybe that’s the norm? I can’t say I’m a fan), some code for fluff, and a few tables. It’s really more like 20 pages worth of content. BUT. It’s published. Dang, and I was getting stressed about not being able to write as much. I should do a word count on all these papers.

Clustering heterogeneous autism spectrum disorder data

Oh wow, in the table of contents everything is covered in red boxes. Did they not know how to turn off draft mode? Or maybe it’s a setting with the hyperlink package. Anyway, it seems strange to leave a mistake like that in the paper.

It’s also double spaced. I guess it is going to be annotated so that makes sense, but still it doubles the pages count.

This paper, being machine learning, has a lot of background that every single other ML paper has along with borrowed low-res images from educational material.

Formulas, diagrams, bulleted lists. The amount of content is greatly exaggerated by the page count.

Here are the newline, word, and byte counts for all the theses I downloaded:

  1133  19148 112613 An explainable sequence-based deep learning predictor with applic.txt
   350   8009  55157 Automatic IQ estimation using stylometry methods..txt
  1227  13818  81274 Clustering heterogeneous autism spectrum disorder data..txt
   941  13181  79794 End-to-end learning framework for circular RNA classication from.txt
   758  15462  84098 Horse racing prediction using graph-based features..txt
   980  17198 102293 Landmine detection using semi-supervised learning..txt
   973  11561  65256 Machine learning for omics data analysis..txt
  1713  13366  92154 Maintainability analysis of mining trucks with data analytics..txt
  1584  13520  80094 Modeling and counteracting exposure bias in recommender systems..txt
  1881  15139  86039 Robust fuzzy clustering for multiple instance regression..txt

Word (middle column) being what I’m interested in. If I do the same to my pdf (or my tex) then I get about 3,000 words. So apparently I’m anywhere from 3/8ths of the amount of content I need to 1/5th or 3/20.

That feels about right. I’m ready to get to work now.

Outline

I guess I can write an outline now. Turning on double spacing, I’ve got 9 pages already. Adding on the standard 5 pages of opening material, That means I have 14 pages already done! Well maybe not really but I feel much better about this than I did yesterday.

Now for the outline: