Final project reports from 2012 Stanford Machine Learning class | Hacker News Reader

Final project reports from 2012 Stanford Machine Learning class

cs229.stanford.edu

99 pointsadmp14 comments

Maybe it's just me, but I absolute loathe the format and language style of "proper" papers. The column view heavy on text and weak on media feels highly outdated. They come across as a wall of words that almost purposefully seem difficult to read. It can't possibly be the best way to convey complicated information. I'll take a slightly less formal, media content rich blog post every day of the week.

Coming from a research background I can totally understand your feeling. Once you have read enough of those papers, you start to kind of skim through them more easily.

Unfortunately, quality of papers and their readability highly depend on the skill of the author. There are some papers which are easy and entertaining to read because of the writing skill and effort put into it. Many though are not nice at all.

Here is an entertaining but serious one... http://www.cl.cam.ac.uk/~rja14/Papers/cocaine.pdf

I particularly thought the "Analyzing CS 229 Project Topics" was clever.

http://cs229.stanford.edu/proj2012/ChangSaeta-AnalyzingCS229...

For anyone interested, a few weeks ago, someone posted a paper on the popular data mining / machine learning algorithms that gives a brief overview of some common algorithms [1]. Someone also posted a few presentations on them if you just want a bullet point summary of the gist of each algorithm [2].

I just finished looking through both. They're both great if you're hoping to get some traction when it comes to learning ML.

[1] - https://news.ycombinator.com/item?id=4938162

[2] - http://www.cis.hut.fi/Opinnot/T-61.6020/2008/

"Machine learning application on detecting nudity in images" Trying to make it awkward with the professor.

I bet the part they most enjoyed was creating the initial training set.

(if it's supervised learning they use, I haven't read the paper)

All reports as a zip archive: http://ul.to/6wix6k9y

I absolutely love looking at other people's ideas -- it's so motivating and inspirational. The ideas of others so often spark my own desire to get cracking on a few new projects.

Based on the nature of the course, I assume these reports are all generated using Markov chains or similar techniques?

Some of the projects are genuinely interesting and if funded may have bright future. Note to investors. It's fascinating to see how many ideas there are, which are still waiting to be explored and conquered.

Thanks for this. I've recently been focused on machine learning myself and this is a gold mine. The titles alone made me open a Google doc to list potential project ideas of my own.

great find! Wish they would go in more detail or contained a bit of source. However, I can understand why they would not open source etc to prevent plagiarism, bit of a shame though.

There was a 5 page limit :) Also, there was nothing to prevent anyone from releasing source code, I'd imagine that a lot of projects are out there if you look carefully on GitHub

Fantastic list

Reply on news.ycombinator.com