this is for holding javascript data
Denes Csala edited Future work.md
over 8 years ago
Commit id: a0eea0caa08a27679acdfd875998096ece66b883
deletions | additions
diff --git a/Future work.md b/Future work.md
index a557e92..0974f12 100644
--- a/Future work.md
+++ b/Future work.md
...
future #Future work
MTD has a great potential for usability in large online platforms and thus it is worthwhile for further development. The keyword here is speed - hinting its usability at online merchandise applications or even more crucial, medical applications. For example, based on the cookie stored on a visitors computer, an online merchant website, such as Amazon can choose what type of comments to show to the visitor, and what types of products recommend just after the purchase. With AMC, purportedly all of this can be done real-time, within the refresh time of the webpage. A search engine run on ads, such as Google can prioritize top results by also by the cookie, and show the ones with the highest click-through chance first, leading to a potential buy and thus a Google dividend with the greatest probability. In a very black world, a police enforcement agency could mine the social media accounts of its citizens and identify the topics that they are commenting/talking about real-time and isolate any potential threats to public security.
From the technical viewpoint, I think it would be interesting to explore further the Bayesian mixture idea. Instead of having a must-link and a cannot-link binary decision space, we could have a global word-by-word co-occurence distribution matrix for each topic. The computational costs would be staggering of course, but this approach would actually provide us with potentially the first, fully machine-driven language model!
