Google plumbing

Since one of my main interests in neural networks it seems reasonable that I use them to enhance the search results from Google as an appliance. Google may know quite a lot about how I search but it actually knows very little about my interests and goals. It is the old Bayesian problem of being 90% accurate on a terabyte of data that has been manipulated by SEO and link farms as well as advertiser preference.

By using my pythons to look around a bit on subjects of interest, it cuts out much of the link farming and dead ends as well as those things that link to pages with too many ads. I can take the links and then apply my own twist to the process by ranking pages by how many crappy flash animations are loaded before the information and reject beforehand those things that I would never sit through, even if they contained the secret of the universe. By creating my own secondary ranking, I keep my intent on the local system. Google may do a good job of understanding intent, but my motives and direction change so quickly that Google or any engine does not keep up. I suppose I can make a browser appliance that has my preferences and fabricate a wxPython browser interface with a neural net that reflects my interests as well as my idiosyncrasies as it relates to how I use the web and what is important. In the same way that a neural network learns a task, I can have my own helper.

The interface can create structured resources in the zim wiki with some relevant ranking as well as images to give clues and a local rank of how many blinking, distracting, confusing, misdirecting, and seizure inducing flash animations I would be subjected to if I were to choose that path.

Alice Infinity pointed out something very interesting and I would share it, but some things have to be discovered for themselves or they will be memorialized on the web and then disappear when abused.

This is part of the local server web pages for Alice Infinity and entry point for search initiation as well as configuring Neural Nets. It seems to me that the networks and sets and equations blur in their character and I often find myself thinking of the first derivative of a Markov matrix chain loop and considering the negative slope as an indicator of its outcome.