You're looking at Less Wrong's discussion board. This includes all posts, including those that haven't been promoted to the front page yet. For more information, see About Less Wrong.

Zian comments on Group Rationality Diary, July 16-31 - Less Wrong Discussion

1 Post author: therufs 15 July 2013 01:39PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (34)

You are viewing a single comment's thread.

Comment author: Zian 16 July 2013 03:38:19AM 4 points [-]

I'm currently working on a website application to figure out what I can't eat. It will use Bayes Theorem. This could make a huge difference in my life.

I also found that I'm really good at asking lots of questions when I'm stuck but not good at realizing that I need to stop working on something until more resources arrive.

Comment author: Daniel_Burfoot 19 July 2013 09:46:02PM 4 points [-]

Do you really need to build a whole web application for this? It seems like overkill. Why not just use a spreadsheet?

Comment author: Zian 04 August 2013 09:57:02PM 0 points [-]

Actually, the web application already exists to a great degree.

The big tasks left are:

  • Set up something to generate metadata like the jump from storing a value that says "ate at restaurant" and having a bit go to True that corresponds to "ate in Los Angeles County".
  • Do the actual Bayes classification or data analysis

Why not use a big XLSX?

It's really hard to enter data into an XLSX on a smartphone. Also, I guess you can say that a MySQL database table has a lot of similarities with a spreadsheet so, in that way, I am.

Comment author: Dorikka 19 July 2013 02:24:31AM 2 points [-]

I'd appreciate it I'd you could share the former when you are done with it. A more general use of this setup might be tracking the effects of frequent actions. There would be some hangups here, but they are a bit inconvenient to discuss while typing oh my tablet.

Comment author: mare-of-night 27 July 2013 02:19:14AM 0 points [-]

I would also appreciate it if you shared the web app. (I've got food problems all over the place, and I probably haven't figured out all of them yet.)

Comment author: Dorikka 27 July 2013 02:37:20AM 1 point [-]

(Just FYI, I don't think Zian gets a notification unless you reply to his comment.)

Comment author: Zian 04 August 2013 10:09:26PM 0 points [-]

You're correct and thanks for posting that.

Comment author: Zian 04 August 2013 10:06:53PM *  0 points [-]

As I mention in another comment, the data crunching is completely separate and likely to generate lots of specific datapoints that are only relevant to me. This is where things get really hard to generalize/make easy to use widely.

If this doesn't scare you away, then send me a message.

Likely pre-requisites are:

  • Programming experience
  • Experience with Java
  • Willingness to set up a local database server (I use MySQL)
  • Familiarity with the various ways to use Bayes's theorem (I recommend Bayesian Artificial Intelligence and of course, EY's article) and how things can go wrong
Comment author: Zian 04 August 2013 09:59:56PM *  0 points [-]

It's actually online already but I am not publicizing it right now for a few reasons:

  • It assumes only 1 user right now so there's no encryption or data separation
  • No reporting features (this is also a security feature; can't have an XSS with no displayed values!) except for downloading the database and connecting with Excel/ODBC/Your favorite data crunching tool
  • Assumes the Pacific Time Zone
  • HIPAA/the mess around dealing with personally identifiable information

Ideas for mitigating the issues are welcome as deploying the site is very easy:

  1. Generate a random string of letters and numbers.
  2. Make a folder on your LAMP server named after #1.
  3. Copy and paste the application to the folder.
  4. Change 1 value inside a config file to point at #1.
  5. Run a script to create the databases.
  6. Fill in your database login information inside a config file.

(This assumes that you're wise enough to avoid leaving any public pointers to the site.)