Yeah, I sort of thought it would be possible to just jump in and use Overleaf to edit a LaTeX template. I’m going to end up going back and looking at a few tutorials on YouTube to understand the finer points of what is happening within the document. It was easy enough to save and load the template. Making a copy was pretty routine and renaming the original was highly intuitive. I was able to edit the title, author information, and a few of the elements in the source file did not really make sense to me. That is why I’m going to watch a couple of tutorial videos to really get a better understanding of what is going on within the document. At this point, I’m pretty sure this will be something that I can manage to help produce papers on a more regular basis from my work. That is where things are at right now.
My current backlog of produced podcasts stands at 2 recorded and loaded episodes. One is ready to go out on July 15 and the other is ready for July 22. That leaves us with the draft of week 79 that is generally complete, but not very compelling. I had moved on and written a pretty decent missive for week 80 that is much longer. The outline for week 81 is clear enough, but it needs more work to bring it up to the standard necessary for recording. I knew that the content from week 81 to week 87 was going to be difficult to generate. Writing out an 8 part syllabus for how I would introduce machine learning is an interesting intellectual challenge. My goal of course is to allow anybody reading the material to come up to speed with a general understanding. The respondent would really have to read the materials and dig into them deeply to walk away with next level skills. That is really the hard part of putting this content together. It needs to be approachable to help provide the breath necessary to introduce machine learning. At the same time, the content contained in the syllabus has to provide enough depth for those respondents who are consuming it to gain knowledge beyond a basic introduction.
I may very well for fun take the 8 part introduction to machine learning syllabus and convert it into a LaTeX document in Overleaf at the end of the process. That would take something that I know is going to be completed and give me an opportunity to really mess around with the typesetting. It might even give me a chance to help figure out the integration between Overleaf and Github which seems to exist, but I have not had the opportunity to explore. That will probably be a good use of my time. The other way to go about getting some practice with Overleaf and LaTeX would be to take a few of my talks over the last few years and convert them over to paper format. Most of those talks have a transcript and a PowerPoint which could be easily converted over to a LaTeX document. Honestly, that content was probably a better fit for dissemination by recorded video and the follow up transcripts. Most of the content people consume is just text in a browser from a webpage, news source, or some type of application. A much smaller percentage of the population in general consumes all their content from PDFs containing academic papers.
I absolutely read a ton of articles and jump in and out of consuming content generally available and content packaged up as academic articles or research notes. Those of you who have read my work for a longer period of time will know that I enjoy a bit of research trajectory mixed into my papers. Knowing the bigger picture and where things are going is an important part of how I consume knowledge. I want to know where it fits into the broader spectrum of the academy and how the author intended it to either move things forward or cement something that needed to be shored up with additional research. That is an important part of the equation that is missing from a lot of machine learning papers that I end up reading. The authors get very focused on the mechanism of the mouse trap and how it functions. They don’t really share the importance of the mouse trap in the broader context of the research within the field. It’s possible that maybe a few papers on the research trajectory of machine learning are necessary. My thesis that has been advanced is that overcrowding is causing a problematic scenario where more content than can possibly be consumed is being created and the noise outpaces the signal by an order of magnitude.