More from San Diego

It's been a very busy couple of days, but I'm having a good enough time in San Diego. There's something to be said for a place where I can go right across the street from my hotel and eat Vietnamese, Japanese, or Denny's. I wasn't able to get out much tonight, but I can't complain much.

First batch of pictures from San Diego

I just posted up some pictures from here in San Diego. All I have is my Sprint vision for data access, which is slow and spotty, so posting much has been a pain. More to come later.

Leaving Tomorrow

Tomorrow morning, I leave for San Diego. To keep me entertained in the airports, I bought Bob Woodard's new book, as well as Paul O'Neill's book, which should both be good reading. I'll be posting trip pictures here periodically, so stay tuned if you're interested in that sort of thing.

How not to find an editor

Looking for a publication editor? Try using the word "it's" properly. I dropped off an email to the address listed on the ad because they obviously need an editor for job descriptions. Even I could see this mistake, and I'm only a grammar master, not a grammar god. I'm pretty sure I could handle it, though. ;) Here's to my new night-job!

(Via eclecticism.)

Like an escape tunnel in prison I started from scratch

I just realized that tomorrow, LAB-Y turns one year old. And it still hasn't made me a dollar. ;) I started this site mostly as a place to house old college class content and a project to drive me to learn PHP and Smarty. I posted my first entry on LAB-Y one year ago tomorrow. I've since ditched Smarty and wound up with a content-management system written from scratch in PHP, designed to run on Freeshell, powering the least read blog ever.

Unconceivable, unbelievable grammar like a hammer

How! grammatically you, "is being"? Which, to taking the quiz, might find us out '.Score me' Master only. :(

Via Eclecticism.

The blogosphere as a neural network?

Yesterday Simon posted an insightful essay about blogs functioning as a news filter. Blogs as filters—what an idea. By adding sites to his blogroll that he knows pull data he is interested in, Simon filters his news intake quite effectively. It got me to thinking... If Blog rolls function this way perhaps the blogosphere itself does much more.

Sure, Simon's blogroll is a filter from his perspective, and a good filter it is. But what analogy can be drawn for the whole 'sphere?

From a larger perspective, a group of bloggers doing what they do, posting, reading, blogrolling, linking, and tracking-back (or trackbacking?) functions like a large, complicated neural network.

The model

Every description of a neural network starts with a model, and this one is no different. Starting from the idea of a perceptron, each blogger is a neuron in the network. Each neuron at different points in time fires, based upon some function of its input. In the blog-network model, this happens when a blogger posts an entry. The blogger has received enough input to cause them to compose an entry. Each blogger also has a set of incoming connections in the form of a blogroll. Like the neuron whose synapses connect to other neurons to receive their output as its input, the blogger maintains a blogroll to receive input from other bloggers. There you have it—that's the basic model.

Let's get into the details.

Incoming stimulation

A blogger neuron receives input from the output of other bloggers. Not every other blogger out there, just a set few. That few are his or her blogroll or regular reading list. Other external input comes from sites that are not regular blog-like sources, which in this model I consider outside the network. These sources are analogous to external input fed into a standard neural network such as a training set or input upon which to operate. We can measure (to some degree) what incoming connections are being simulated by looking at what the blogger is linking to. Harder to measure are the strength of the links (weight) connecting the blogger neuron to its inputs.

Recurrence

Like a recurrent network, the 'sphere records state through trackbacks, referrer logs, comments, and Technorati, feeding stimulation in the opposite direction of reading and linking. Who is paying attention to my output? These reverse stimulation mechanisms provide the answer. A network that uses this structure is called recurrent and can predict sequences, or like a Hopfield network, perhaps simulate "robust content-addressable memory".

Training?

One hole in my analogy here is the lack of usual training that is involved with generating neural networks. How can you propagate error backwards on the blogosphere and correct error? Blogger's are individualistic about their tastes, making it hard to give even a guess as to what their activation function must be. Forget sigmoidal, it seems random at times from an outsider's perspective.

Without a differentiable activation function, error minimization is hard. If you don't know what direction to move in to decrease the error, you can hardly minimize it.

Here, the blogosphere resembles more a Kohonen map (or Self Organizing Map) than a neural network. By moving similar units closer together (blogrolling and other forms of association), the network organizes itself without supervised training. There doesn't have to be a "training set" with defined correct answers.

How could we use it?

The problem with this model is that any sane researcher training neural networks defines a way of drawing conclusions from the network. This might be a set of output neurons, one of which firing indications a particular classification of the input, for example. How can someone do that using the blogosphere?

I don't have any specific answers to that, but one thing does come to mind. We have the perfect platform upon which to build an application that does just that. Google. Because as Simon points out, it isn't just a search engine; it's "the world's largest and most scalable platform for developing huge web-based applications".

So one day, someone will develop the gigantic planet eating robot, with an ultra-sophisticated blogonetwork web app as a brain, drawing on the power of Google and the blogosphere to do its evil work. I for one am glad to be a part of it.

This car I'm pushin

What a weekend. I hit 70k in my Escort not too long ago and being the responsible me, I took the POS in for the regular maintenance. $Cha-ching$ TCO goes up another $230. In addition to all of the standard stuff they do at 70k, I needed the engine decarboned, a new drive belt, and my brakes looked at.

Current total: $1515

And not 10 minutes after driving off the lot, the ghost of Henry Ford popped up again to torment me. The damn engine light came on AGAIN. I'm 99% sure it is the %$@#% IAC valve that I've had replaced several times acting up again. Funny thing is, looking through referrer logs for LAB-Y, I get a lot of hits from people for info on Escorts with IAC valve problems. What the hell was Ford thinking?

So it's another trip to the dealership soon. At least I can hold out, as it won't cause me noticeable problems. Hopefully, it will still be under warranty, like last time. Replacing it then didn't cost me a dime (not even for the diagnosis).

Operation: San Diego

I just found out Friday that I'm heading to San Diego for about a month on business. How crazy is that?

I haven't been to San Diego (or anywhere in California for that matter) since I was a child, which I have little to no recollection of. So when I volunteered was drafted for this assignment, I was pretty hyped. Cali has my name written all over it. ;) I put my name in for the mission to hopefully catch a show with one of the Beat Junkies or one of the former Skratch Piklz.

To prepare, I'm buying a digital camera, probably tomorrow. With any luck, I'll be posting pics over a Sprint PCS connection periodically. w00t! Hopefully, my girlfriend will be able to fly out and join me.

Anyone have any hints for what I should see in San Diego? Anyone familiar with the hip-hop scene there? Or should I just bite the bullet and drive up to LA?