primaryobjects

Looks neat. Why did you bother using <PAD> words to have sentences be the same length, when you're using a bag-of-words (document-term matrix) model anyway?

Each sentence vector ends up being the length of the vocabulary, so they're already the same length. You can probably drop step #3 in this case.

This looks a lot like STRIPS planning http://www.primaryobjects.com/2015/11/06/artificial-intellig...

The Ceptre tutorial even has the classic example of moving blocks on a table. It's a different language than STRIPS (using PDDL), but seems a similar idea.

I feel compelled to also mention "Using Artificial Intelligence to Write Self-Modifying/Improving Programs" http://www.primaryobjects.com/CMS/Article149

The link above describes a project where an AI wrote programs for Hello World, addition, subtraction, multiplication, Fibonacci, bottles of beer on the wall, and a bunch more.

I was recently using js-combinatorics https://github.com/dankogai/js-combinatorics

Worth taking a look, if anything, to gauge various functionality.

Nice editor. I also recommend taking a look at inline editors, like Aloha http://www.alohaeditor.org/demo/aloha-ui/

I had previously written a node.js module that uses it for embedding little CMS content blocks on a web site http://contentblocks.herokuapp.com

I was concerned about this when recently releasing a meteor.js site, whether Google would index the pages and content.

Meteor serves pages as a single-page-application (same as AngularJs and other ajax frameworks), so if you simply view the source of the home page, you'll see mostly empty HTML. I didn't add any extra modules for SEO and simply published the site.

It looks like Google has indeed indexed the pages and text https://www.google.com/webhp?sourceid=chrome-instant&ion=1&e...

That's from Neuromancer, when the hologram (like a chat bot) is having a conversation with the main character.

The Associated Press | Cranbury, NJ | Junior Front-End Web Developer | Full Time, Onsite http://www.ap.org

The Associated Press has an opening for a junior web developer with experience in C# .NET MVC 5, AngularJs, jQuery, Javascript, CSS, MongoDb, and REST web services. Experience with Node.js and Github is a plus!

The AP is one of the largest and most trusted sources of independent newsgathering, supplying a steady stream of news to its members, international subscribers and commercial customers. Founded in 1846, AP has covered all the major news events of the past 165 years, providing high-quality, informed reporting of everything from wars and elections to championship games and royal weddings.

http://chc.tbe.taleo.net/chc04/ats/careers/requisition.jsp?o...

The Associated Press | Cranbury, NJ | Full-Stack Web Developer | Full Time, Onsite http://www.ap.org

The Associated Press has an opening for a junior web developer with experience in C# ASP .NET MVC 5, AngularJs, jQuery, Javascript, CSS, MongoDb, and REST web services. Experience with Visual Studio 2013, TFS, AWS, and Github is a plus!

The AP is one of the largest and most trusted sources of independent newsgathering, supplying a steady stream of news to its members, international subscribers and commercial customers. Founded in 1846, AP has covered all the major news events of the past 165 years, providing high-quality, informed reporting of everything from wars and elections to championship games and royal weddings.

http://chc.tbe.taleo.net/chc04/ats/careers/requisition.jsp?o...

There was another project like this posted a few weeks ago https://news.ycombinator.com/item?id=9491978, and I have the same question for this one. :)

Why did you decide to use a supervised learning approach for this? You're hard-coding the categories in this manner.

You could try using clustering instead (K-means) to group the posts together into their own categories. This is similar to a "trending topics" approach, where the categories could change each day.

On a side note, the domain name is terribly confusing.

Only because you asked, this has been my pet project :)

Using Artificial Intelligence to Write Self-Modifying/Improving Programs

http://www.primaryobjects.com/CMS/Article149

Just curious, why did you decide to use a supervised learning approach for this? You're hard-coding the categories in this manner.

You could try using clustering instead (K-means) to group the posts together into their own categories. This is similar to a "trending topics" approach, where the categories could change each day.

Here's a program that solves this type of logic problem. It works on both "Cheryl's Birthday" and "Teacher's Birthday".

Just replace the initialization dates with the ones from the problem.

Cheryl's Birthday http://www.r-fiddle.org/#/fiddle?id=8zUu1GdT&version=4

Teacher's Birthday http://www.r-fiddle.org/#/fiddle?id=8zUu1GdT&version=5

The Associated Press http://www.ap.org - Cranbury, NJ - Web Developer, Full Stack

The Associated Press has a short-term contract role open for a software developer with strong experience in C# ASP .NET MVC 5, AngularJs, jQuery, Javascript, CSS, MongoDb, and REST web services. Experience with Visual Studio 2013, TFS, AWS, and Github is a plus! Must be corp to corp.

The AP is one of the largest and most trusted sources of independent newsgathering, supplying a steady stream of news to its members, international subscribers and commercial customers. Founded in 1846, AP has covered all the major news events of the past 165 years, providing high-quality, informed reporting of everything from wars and elections to championship games and royal weddings.

For those interested, I created a node.js LDA library that's pretty easy to use.

https://github.com/primaryobjects/lda

Update: I was able to train the network to sort sets of two 3-digit numbers. I used a neural network with 2 hidden layers of 25 nodes. The training/test accuracy after 10 minutes is 78%/74%. Not bad. https://github.com/primaryobjects/nnsorting

I started playing with this. I take each digit and normalize it by dividing by 9. Then use each normalized digit as an input:

Example sorting 987654 and 123456

  Input: 1, .9, .8, .7, .6, .5, .2, .3, .4, .5, .6, .7
  Expected output: .2, .3, .4, .5, .6, .7, 1, .9, .8, .7, .6, .5

You can then encode/decode the inputs and outputs accordingly. if (value <= 1) digit = 9; if (value <= 0.9) digit = 8; ... if (value <= 0.2) digit = 1; if (value <= 0.1) digit = 0; etc.

I'm able to get 100% accuracy on a limited training set with 2 hidden layers of 10 nodes. 33% accuracy on the test set (but likely need a lot more data to train with).

I installed via the Linux Mint package manager and found no issues of malware.

Thanks, I see it. The translation code is inside ir.py. So you can call ir_to_c(bf_to_ir("+++.")) to do the conversion.

Here is the code to convert BF to C, in case anyone else is interested https://gist.github.com/primaryobjects/0ba06f1f6f0d3e984d89

Here it is running the converted C code https://ideone.com/z16KPJ

Very cool.

Is there an implementation of this that converts any BF code into C/C# code? Might make a handy web-based tool.

It would be interesting to apply this to AI generated programs (http://www.primaryobjects.com/CMS/Article163)

Great work. I also really like the json files that it produces https://github.com/gaganpreet/hn-hiring-mapped/blob/gh-pages...

Makes it easier to play with data analysis on the hiring threads in R. http://pastebin.com/TpLfSKjP

Mazes are certainly fun. Here's a javascript maze solver I wrote a while back, including a REST url for building your own maze. http://www.primaryobjects.com/maze

Artificial intelligence, currently popular by machine learning. Lots of potential and undiscovered applications. Everyone will have their own opinion on this, however.

Some people frown upon training neural networks with genetic algorithms, just because gen algs are so random and hard to dissect. I think it's silly to discard anything that is a potential solution to a problem.

I go by a rule of thumb like this:

- If you are able to collect lots of training data (inputs and valid outputs) then use back-propagation. It's faster and you might get better results.

- If you don't know the outputs for inputs or if there are simply too many possible combinations (such as in the case of a game), then use a genetic algorithm. It's effectively a search engine that finds the best solution within the problem space (the solution being the optimal weights for the neural network).

Using Neural Networks and Genetic Algorithms http://primaryobjects.com/CMS/Article105.aspx

Neat. You could also probably do this with a genetic algorithm. Thinking out loud here:

- Have a neural network with 6 inputs for:

   - player x
   - player y
   - acceleration speed
   - acceleration direction
   - next pipe mid-point x
   - next pipe mid-point y

- Two outputs of 0 or 1 for click or no-click

The fitness score would be how close the player is to the pipe mid-point. Hopefully, this would cause the bird to stay as close as possible to fly between the pipes. The genetic algorithm would select the best neural network that knows when to flap based on the current input state.

Just today, I finished a deep-learning example in C# with Accord .NET. It solves a simple XOR function, as well as classifying ascii digits.

If you're looking for a basic example in .NET, I think the XOR one is as simple as it gets.

Deep-Learning XOR: https://github.com/primaryobjects/deep-learning/blob/XOR/Dee...

Deep-Learning Digits: https://github.com/primaryobjects/deep-learning/blob/master/...

A fellow tinkerer in program synthesis? Just today, I was tweaking my AI-Programmer project http://primaryobjects.com/CMS/Article149.aspx, which uses genetic algorithms to generate programs.

Glad to see I'm not the only one who thinks programmers can/should become obsolete. :)

It works on an explicit set, but not on initialization of the object. So:

var x = [{"user":"dude"}]; This won't trigger, and this is what the script include tag executes via the response.

x.user = "wow"; This will trigger, however.

That's interesting. Apparently, the __defineSetter__ call is still valid in Chrome.

Example http://jsfiddle.net/V53BL

Yes, all very good points. I didn't use a neutral category, as it's difficult to gauge whether a tweet should be neutral. I've thought of detecting them based upon news-speakish tweets, as you've pointed out. That would require its own machine learning run, just to sort out neutral from containing-sentiment.

Also keep in mind, different topics work better. The term "election" has a lot of news headlines, which many are probably neutral, skewing the results. More consumer-ish topics yield better results. But yes, tweets are difficult to analyze. I've done another recent experiment with tweet analysis, if you like this kind of stuff http://primaryobjects.com/CMS/Article158.aspx