Hello all, sorry for the recent blackout. The server crashed badly yesterday while doing a reboot and we just restored the webserver and the database. Sorry for any inconvenience!

It's been a long time since the last time I updated this site, but I'm still here. I'll be developing other projects but I have a bunch of articles I'll upload here pretty soon. Some days ago I found a very interesting website called WikiCFP. It's a wiki where you can see all current call for papers of the most important conferences around the world. Each call for paper is grouped inside a category (Natural Language Processing, Linguistics, Databases, AI and Communications are the current categories) so you can see just the ones you are interested in. Apart from that, you can contribute to the wiki with new additions, new categories, etc. The best part is that they have a feed that you can use to subscribe to one of the categories and keep yourself updated. I've updated the links section with this resource and a couple more.

I read some pretty old articles recently and I was shocked to see how much truth they hold and how few people actually follow them. It's incredible how we keep tripping with the same problems our fathers and older collegues had already tripped with. This is true for both, AI and security fields. So I'm uploading two articles, the first one is titled "Artificial Intelligence Meets Natural Stupidity" by Drew McDermott. The article is from 1976, but it's a priceless jewel. It makes a great reading for all Academics and PhDs out there. The second article has to do with security, more exactly with password schemes. It's the historical paper from Robert Morris and Ken Thompson from Bell Labs titled "Password Security: A case history". It's the case study for the current password security schemes we find in any Unix. Worth reading, specially if you are into Security Engineering fields.

Large scale simulations of neural and AI models are more and more common every day. It's clearer everyday that intelligent behavior is a byproduct of neural activity at a macroscopic scale. Nevertheless, being able to simulate large models is quite expensive in terms of raw computing power and not many people have the resources to do it. That's why the latest paper from Dr. Dharmendra Modha titled "Anatomy of a Cortical Simulator" is so interesting. In it you'll find several solutions and hacks to the problems you can encounter when simulating very large cortical systems. You can also read Dr. Modha's FAQ about the paper here. Beware that the simulation was just to test the technical issues of simulating such a huge model and not about the actual network dynamics or cortical circuit connections.

Finally some news. I have been a little bit sloppy with the site lately and I'm sorry about it. But I'm back with some new articles. First of all, the site has experienced some growth in terms of hits per month and there are quite some RSS subscriptors (thank you all!). I'm changing the feed url to this one at feedburner so I can monitor the RSS better. Please if you are an RSS subscriptor and you could re-subscribe to the new url it will be great. Ok, now lets get on with the news. I've recently started a personal blog where I talk about broader topics, if you are interested, you can take a look here. The latest post talks about Powerset's semantic search engine technology, so those of you interested in real commercial technologies might be interested in reading it.

Continuing with our NLP saga, I am uploading a very interesting article from Silviu Cucerzan and Eric Brill titled "Spelling correction as an iterative process that exploits the collective knowledge of web users". It gives some very good insights into how automatic spelling correction and how non-trivial the task is. As a side note, Norvig's post on spelling correction is also worth reading.

After summer we come back with some interesting articles. I have just uploaded an article of a current model of short term memory by Roudi and Latham from Gatsby Computational Neuroscience Unit at UCL. The article answers two important unanswered questions: how does a neural network exhibits irregular firings? and how many memories can be stored in the network under realistic biological conditions?. Really interesting article, although pretty mathematical. "

Here comes another paper linked to Dr. Robert Hecht-Nielsen about the Cocktail Party Problem. With the advent of new Natural Language Processing systems, and smart AI technologies, it is a matter of time until you face the Cocktail Party Problem. It is basically the ability to focus on one listeners stream among a group of simultaneous conversations. It is been a widely known problem and has been tackled in many ways. Nevertheless, the model presented here is based on biological data and proves to be very robust in real world scenarios. What I like more about the model is the use of predictions (expectations) of what words come next while listening to someone. If you think about it, humans don't grasp every single word in a conversation in a noisy environment. We normally fill in the gaps with expectations. This is the same mechanism proposed in this model. Worth reading!

The Perceptron is a type of Artificial Neural Network invented in 1957 by Rosenblatt. It is the most well known ANN algorithm since the creation of the field. Nevertheless, it's hard to find good papers (and free) about it (except the original ones). I must thank Robert Hecht-Nielsen for posting this wonderful monograph about it, I find kind of amusing the appendix where he talks about the history of the Perceptron and a fact which I didn't know, that was that Rosenblatt used Minsky's hardware schemes for his Perceptron machine. This could have been, as pointed by Dr. Hectt-Nielsen, one of the reasons behind the attacks to the Perceptron by Minsky. Read more in the monograph.

First of all, sorry if someone has experienced problems with the atom feed during the weekend. I had some troubles with an upgrade. Now it should work without a glitch. After posting my master thesis I have decided to upload the articles from Jorg Lucke in which it's based. While thinking about the thesis I came through Lucke's works and I was very shocked to find that he had been working on minicolumn dynamics for quite some time. I really recommend his papers for anyone interested in minicolumn/macrocolumn neural network dynamics. Some of the papers are also coauthored by Von der Malsburg but I really recommend all of them, as they give a really good insight in to how to model inhibition between populations of minicolumns."