I've been reading Mahout in Action and its very good to say the least. The one thing I was annoyed at was the lack of an up to date chapter on creating a Web Service to feed Mahout data. I mean this is Java right? We have access to many different libraries, so my idea now is to create a basic recommender and use an http library to listen to requests or even create a web service layer in Python and create the recommender in Java and find a way to link them up.

One way to get python to work with Mahout may be ...

and what do I expect to do with it. Don't know what this is? Click here for my first post.

Intro

The HN News Recommender is due to my interest in recommendation systems, more or less content-based ones. I would really like to know more about what the industry is doing, but I thought I'd go figure out how simple classification works, since I'm not learning much about it in college, but college does help with understanding the theoretical parts and explaining why you believe something is reasonable.

Anyway, wher...

Read: http://www.laurentluce.com/posts/twitter-sentiment-analysis-using-python-and-nltk/

According to this, I could make some modifications to my chrome recommender extension, bundle a trained model with it, and then classify tweets on the fly. Hmm, maybe, maybe not. I've been more into Mahout lately because I've going to use it for Project Flare.

On another note, I haven't released v0.4 of the Chrome Extension due to a hiccup in Chrome. It works on stable releases, but the dev releases will screw up due to t...

Anyone ever think about why the residuals need to be independent and identically normally distributed around mu = 0 and variance = 1? Well, I thought about it for a while and figured if you look at what linear regression does then you'll connect it with the reason why we like normally distributed residuals. Linear Regression attempts to fit a line to the data while minimizing the distance to each point. Any point not on the line is considered to have an error or residual (remains), so a reasonable assumptio...

Hmm, I just read an article on Coding Horror about hashing, but my take away is to use passphrases. Now I'm thinking wouldn't pass phrases be easier to remember and longer than a password? Since passwords should be ideally be 12 characters or more, then instead of making up passwords that are crypto as heck, I could get a user to create a pass phrase with the account. I mean obviously, they'd need to be able to see the pass phrase before they hit login, but I think that's where I'm going towards. I mean my ...

Ever wanted to install Hadoop or Mahout? Let's go. Keep in mind, I'm using Fedora 16.

Apache Hadoop

1. sudo yum install maven java hadoop eclipse --> you should install eclipse, but you don't have to. I believe when you install eclipse it drags in all the java dev stuff you'd need.

2. vi /usr/etc/hadoop/hadoop-env.sh and edit the export JAVA_HOME line reflect where java is installed. On F16 is installed in /usr/lib/jvm/java

3. type '/usr/etc/hadoop/hadoop-env.sh' there should be no output

4. I'm assuming single-...

Looks like I'm going to v0.4 soon. Updated the extension to use IndexedDB on chrome even though webkit's IndexedDB is is a little messy right now. Versioning between webkit and mozilla is different, but someday they'll match up and this extension is ready for that. Oh and if you update to v0.4, your current results will be transferred over to indexedDB and used there.

and if you've no idea what this is: visit my first post on this

Anyway, that's about it for v0.4.

In future updates, I will be rebranding the ex...

Looks like I've been interested in Test Driven Development for some time and I will attempt this using Jasmine to start testing my HN News Chrome extension v0.4. I hope it is worth the time as it is a small extension. Definitely worth it if I can find creative ways to break the thing.

Google Chrome Extension: Get it here. Updated Git source here as well: https://github.com/normano/ycChrome

What is it? Visit here.

What's new?

• In the options page, there are two new options. Color and reading frequency.
• Increased accuracy of recommendations.
• For more details look below!