mayoatcensi.blogspot.com
MaYo at Censi: mei 2009
http://mayoatcensi.blogspot.com/2009_05_01_archive.html
Vrijdag 15 mei 2009. 3 algorithms tested on single task data. I have tested three algorithms on my single task data, so the data of the subjects when they were asked to perform one single task at the time. I have trained these algorithms on the data of all subject minus 1. The algorithm is tested on the data of the left out subject. I have repeated this for all 11 subjects. So I use "leave one subject out" cross-validation. SMO 94,54 %. Naive Bayes 92,72 %. Abonneren op: Berichten (Atom).
mayoatcensi.blogspot.com
MaYo at Censi: oktober 2008
http://mayoatcensi.blogspot.com/2008_10_01_archive.html
Donderdag 30 oktober 2008. Okay it has been a long time since my first and only post. sorry. I will do my best to post regular updates from now on. In my last post I wrote I was orienting on the kind of behaviour I wanted to recognize. I had already set my mind on people with a mental disability and I had to narrow down which kind of behaviour from this group I wanted to recognize, so which problems I wanted to solve. Because the other group from CENSI " structure in the day. A less invasive system is ne...
mayoatcensi.blogspot.com
MaYo at Censi: data analyse
http://mayoatcensi.blogspot.com/2009/04/data-analyse.html
Dinsdag 7 april 2009. I have performed the tests with 11 test persons. The test persons were given 10 different tasks and in the first round they were instructed to perform these tasks one at a time. In the second round they were given 2,3 or 4 tasks at the same time and they were allowed to multitask if possible. I'm now playing with the data I have gathered from these experiments. My first preliminary results give a correct classification of 72,7273 %. This is not very good, but also not very bad.
mayoatcensi.blogspot.com
MaYo at Censi: april 2009
http://mayoatcensi.blogspot.com/2009_04_01_archive.html
Vrijdag 17 april 2009. I have tested the "support vector machine" algorithm implemented by WEKA. This algorithm is actually a variant on the SVM algorithm and is called "Sequential Minimal Optimization" (SMO). I train the system on all the data minus 1 person. The data of this person I use as test database. With this approach I get a recognition rate of 100% on 7 persons, 90% on 3 persons and 70% on 1 person. This seems like a very good result. Dinsdag 7 april 2009. I have performed the tests with 11 tes...
mayoatcensi.blogspot.com
MaYo at Censi: Pictures of the test environment
http://mayoatcensi.blogspot.com/2009/03/pictures-of-test-environment.html
Maandag 9 maart 2009. Pictures of the test environment. Here are some pictures of my test environment:. Abonneren op: Reacties plaatsen (Atom). Pictures of the test environment. Structuur in de dag.
mayoatcensi.blogspot.com
MaYo at Censi: september 2008
http://mayoatcensi.blogspot.com/2008_09_01_archive.html
Maandag 22 september 2008. I'm Marjolein Terhaard, a student Artificial Intelligence at the Radboud University Nijmegen. I will work on my master thesis at the CENSI. With my study artificial intelligence I hope to be able to help people in their everyday life by using technical solutions. My interest goes to children with learning deficiencies, elderly people, or people with a mental or physical disability. To recognize all behaviour possibly conducted by persons would be a too big project, so my next s...
mayoatcensi.blogspot.com
MaYo at Censi: november 2008
http://mayoatcensi.blogspot.com/2008_11_01_archive.html
Dinsdag 18 november 2008. Last week I had a brainstorm session with some other CENSI people (Anne-Margot, Nienke, Nikita and Edwin) under the professional leadership of Richard Vos. Together with some previous ideas, this has resulted in the following overview of behaviours, objects and sensors that could play a role. These rules can be made by observations of the sequences of handlings or by having the system learn these sequences. One way or the other, this will take too much time. But the question is ...
mayoatcensi.blogspot.com
MaYo at Censi: maart 2009
http://mayoatcensi.blogspot.com/2009_03_01_archive.html
Maandag 9 maart 2009. Pictures of the test environment. Here are some pictures of my test environment:. Abonneren op: Berichten (Atom). Pictures of the test environment. Structuur in de dag.
mayoatcensi.blogspot.com
MaYo at Censi: juni 2009
http://mayoatcensi.blogspot.com/2009_06_01_archive.html
Dinsdag 23 juni 2009. The next step was to see how well the different classifiers were capable of classifying the multi task data. This was the data where the subjects had performed multiple tasks at the same time. The subjects were given 2,3 or 4 different tasks that they were allowed to perform at the same time. Multiple labels were attached to this data. I used 2 different ways of training and testing. First I trained on all the single data and tested the algorithms on the multidata. SMO - 82 %. The r...
