My AI lecturer remarked that AI was a great term for hard problems, and that once a field or area was effectively solved or developed, it no longer got called AI.
777 karma · joined April 25, 2010
EDIT: Ah I found the problem, following internal links is conflicting with jQuery Mobile. Thanks for the report.
This is fixed now.
I'll add offline to my list of features to work on, it hopefully shouldn't be difficult, although MathJax lazily loads files it needs from the CDN, which could be tricky.
EDIT: I just pushed a patch related to fieldSelection in Firefox 4.
* Edit: this interacts a bit weirdly with changing between narrowscreen/widescreen. I'll work on making the textarea width more flexible.
The source is available on github here: https://github.com/fivesixty/notepages
Related, my Third Year project is a graph database that loads its data lazily from configurable back ends (databases, caches, APIs, written in Clojure). Now I have more evidence for my dissertation that this kind of tech is useful.
I wonder what their language for queries is going to look like, the only main graph language I've found is Gremlin, which used to be XQuery derived, but recently switched to chained objects.
I recently released a web editor for Markdown/LaTeX mix (http://notepag.es/introduction) using Showdown/Mathjax. Have found it very useful for quick technical docs (assignments at university primarily). Though it still needs some smoothing between markdown and tex.
I did work on my own take, using some of the ideas from SproutCore, though didn't get very far in polish (http://github.com/erudites/eventful)
I've only had a brief look, but the dependency tracking looks much cleaner than other solutions I've seen for partial template redrawing.
Do Backbone and Knockout interact well, or are there conflicts in getting them to play nicely together?
A clarification here is that these rely upon the existence of one way functions, of which discrete logarithms and factoring are conjectured to be in. A proof of the existence of one way functions is actually a slightly stronger result than P!=NP (as it relates to languages with only a single accepting state rather than many. The name of the class escapes me and Complexity Zoo is unresponsive at the moment)
I haven't checked whether the content is the same, however.
But generally the CSS spec seems to be increasing faster than vendors implement it. Or rather every vendor is creating new extensions and pushing for them to go into the standard so their effort isn't just for fun, leading to the bloat.
In a 3 layer network you have an input layer, a hidden layer, and an output layer. Between nodes in each layer you have a weighting, and a nodes output is the sum of how strongly its activated by nodes in the previous layer. For calculation, you set the activations in the input layer to your input. These then cause hidden nodes to be activated, given the weights, which then activate output nodes, which is what you see as the outcome.
It can be shown that any function can be approximately arbitrarily closely by a single hidden layer by increasing nodes.
Backpropagation uses a set of test data, and expected outputs, to adjust weights. The test data is put as an input to the network, and how much each node is activated is recorded. The algorithm then works backwards through the layers, adjusting weights so that the output will be the/closer to the desired output. Backpropagation itself is about a page worth of differentiation derivation, as you need to work out how much each weight affects each node.
There are other details such as the sigmoid function (increases 'contrast' of whether a node is activated or not, to give better choice boundaries) and the issue of over-training, where your network becomes too trained for your specific test data, and actually gives worse results than a less trained set.
"Operations on integers (and floats) include addiction.." Just a rather amusing typo in one of the first sections.