I believe Machine Learning is being used in this application because there was mention of training data. The system work with individuals that can control their larnyx but cannot necessarily produce audible tones. By thinking of a word and making an effort to speak it the larnyx activity is captured and interpreted into actions. Thinking of words can guide your wheelchair
This may be pushing the envelope of recent but it was in 2007. I am very interested in understanding how Google handles pages which are set up to game the system. This document describes a patent filed by Google to determine valid word phrases in indexed documents and then it determines the significance of those phrases. If a phrase is more significant it is a better predictor of the content. This is useful in determining when a person is manipulating the AdSense system. Google’s New Algorithm to Rank Pages and Detect Spam: PhraseRank?
Either only one of the objectives is adopted as the cost function or multiple objectives are aggregated to a scalar cost function in traditional Machine learning procedures. And conventional learning algorithms can (mostly) deal with only one scalar cost function. However, multiobjective learning approaches are more powerful because Machine learning is inherently multiobjective. I'm really excited about this approach in this recently published paper (May 2008) because in Electrical engineering, two unhandled issues: multiagent control and wireless channel modeling (basically finding an accurate but less complex model for the electromagnetic broadcast channel) are related. The Pareto-Optimization idea combined with Evolution Algorithm is also fascinating. http://ieeexplore.ieee.org/iel5/5326/4492353/04492360.pdf?tp=&arnumber=4492360&isnumber=4492353
For decades, people would believe that Traditional Chinese Medicine is a popular complementary and alternative medicine in Western countries in estern Asian erea at best. I, as an Asian, would probably turn to Western medicine although Chinese Medicine is said to have a tone of funky complicated theorems in it. The western medical system is largely based on experiments and clinical data while the Chinese one is kind of fussy logics. But what's interesting about this 5-page paper is that it tells us both medical systems may be generalized into one theory from a machine learning point of view. P.S. The introduction part may be a little crazy for those who haven't got an idea of Chinese medicine http://ieeexplore.ieee.org/iel5/4446227/4449533/04449631.pdf?tp=&arnumber=4449631&isnumber=4449533
Might be old news to those who stay abreast of machine learning news, but between th 22 and 24 of March 2008, the MoGo artificial intelligence (IA) engine developed by INRIA - the French National Institute for Research in Computer Science and Control- running on a Bull NovaScale supercomputer, won a 9×9 game of Go against professional Go expert (it must be big in France to have professionals) Catalin Taranu. This is significant because the tournament was the first ever officially sanctioned 'non blitz' victory of a 'machine' over a Go Master and because Go is more complex than chess and has “More Possible Combinations Than the Number of Particles in the Universe” (when a machine beat a grand master in chess it was reported that the computer simply memorized moves for certain situations). Apparently, Catalin Taranu was able to beat the machine on 19 x 19 board configuration, but the machine beat the human on a 9×9 board. See Machine wins at Go for more information.
OK, so not really, this is basically the case of personalizing medicine. With people being so different and each drug causing different side effects for different people, some researchers have posed the question of using machine learning algorithms and data mining techniques to suggest drugs to doctors based on their observations of the patient. This is interesting because of the ethical issues and how people would adopt to an unorthodox method. Do people trust a human doctor prone to make mistakes or a cold/personality-less machine that is generally more accurate? I personally think that an integration of such a system would be beneficial to society and could help those who cannot afford to have a doctor diagnose them and could be diagnosed by a machine instead. For more information see Machine Learning Blog and A Health Care Blog.
We are drowning in information but starved for knowledge.
– John Naisbitt, Megatrends: Ten New Directions Transforming Our Lives, 1982
This is more obvious in health-care. When we feel bad about some parts in our body, we might go to a doctor's. After describing a lot about how bad we were feeling, it's normal that the doctor still have difficulties to say what is exactly wrong. Most often, we need to take various extra examinations to come to a result. It's time consuming and sometimes still draw a blank. This program aims to track the health status and finally can even predict chronic diseases automatically. Based on collecting information from ordinary medical records, using machine learning techniques, the program is able to give suggestions for better health. Here the machine might take the place of a doctor, providing sufficient knowledge. However, for me, there are two main questions: (1) how can we provide sufficient knowledge (not information) to the machine? (2) how much can we trust the result from the machine? The name for workshop is Machine Learning for Healthcare Applications. For more information, please check the official page here.
“There are two ways to build a fast machine learning algorithm: Start with a slow algorithm and speed it up, or develop an intrinsically quick learning algorithm from the ground up.”
This is a project at Yahoo! Research to design a fast, scalable, useful learning algorithm. Actually it belongs to the second method mentioned above. Rather than analyzing all existing data, this algorithm starts with only a part of the data. With the small part of samples, the system can still make right decisions. When collect more samples over time, the algorithm improves and possibly can make better decisions. This research is like a self-development algorithm. The greatest contribution would lies in the learning speed. However, for me, it would be very interesting to see how the algorithm improves automatically. And how well can it make decisions at the beginning only with few samples. Still I have a question: what does the “online” here mean? Here is more information about Yahoo! Research and this project.
This is a paper/video lecture about a machine learning study on the anti-cancer activity of about 4,000 unique compounds. The ability to predict such activity is important because while the number of small molecules in databases that could potentially be used for drug therapy is rapidly growing, lab testing is limited and can be very slow. Decision trees, inductive logic programming and support vector machines were used with molecular weight, octanol water partition coefficient, and fragment counts as features. I thought this was really cool and relevant to life because this study and others like it should help speed up the improvement of cancer treatment. Who knows, we might really even cure cancer someday! To read more and/or watch the lecture, go here.
OK, I will be the first to admit that I was extremely skeptical when I came across this article. I actually extensively Googled the company as well as the news site to see if it was a hoax or something. It's about a new handwriting analysis tool called Candidate Insight that claims to be able to produce a personality report from your handwriting that employers can use to pre-screen candidates for dangerous or troublesome personality types. Has anyone ever heard of it? They supposedly were able to prevent an attack on a European campus using this tool – the suspect in question had a profile that closely matched the one of the Virginia Tech murderer. I find it hard to believe that you can learn personality traits from such tiny handwriting samples as the ones they claim to use, but then I don't know anything about handwriting analysis so maybe it checks out. Anyway, if anyone knows if this is actually fake let me know. Original article on California Newswire (is that a legit site?), article about the allegedly prevented attack, and the company homepage. Words cannot describe how skeptical I am, but if it works, that's really cool! Scares me if they go corrupt or something though, they could easily blacklist people by claiming that they're homicidal or something.