Weekly Progress

Week One, March 30 
The group was formed and a day was picked to get together every week outside of the lab meeting.

Week Two, April 6
Set up the blog. Decided that the given data was going to be used for the project. The Chaplin movie clips and Nature clips will be used. Also, a first draft of the project proposal was written.

Week Three, April 13
 Wrote final draft of proposal. Researched statistical analysis and downloaded software and data. We decided to try and clean up our signal by concentrating on certain parts of the brain that are specific to vision and image treatment. the blog was updated to include an "about us" section.

Week Four, April 20
 Researched lobes of the brain and their functions. Also researched MEG and EEG scans and started coding the basics. In lab we learned that we can not concentrate on just the occipital lobe, we need to follow the pathway of the brain processing the video clip. A thumb drive was dedicated to the storage of the raw data.

Week Five, April 27
 Revised and added to code. Downloaded code and program to computers of all group members. A shared file was posted on the blackboard learn website that summarized the code files. Program could not find the function pmtk3, so Group 5 had to search the internet and downloaded it from Google.

Week Six, May 4
Edited code to reduce number of errors. Downloaded pmtk3 MATLAB toolbox for program. This week, Group 5 began work on the correct prompt. Running the ReadMind program written by the ICANN winning team yielded mostly path errors.The group started to run the program with different options to see which of those gave us the better time and result, creating a table of these results, as shown in Table 1.

Table 1: Week six testing, percentage of accuracy with the training and mystery data.




Week Seven, May 11
Continued to change variables in the program like the n-folds and recorded the results, as seen in Table 2. Since 1,3,4 has the best results, different n-folds were tested to test how much the accuracy would improve. Wrote the draft for final report based off of the results retrieved thus far.


Table 2: Week seven testing, percentage of accuracy with the training and mystery data. Different n-folds were tested for the generated feature 1,3,4. 


Week Eight, May 18
Continued to change variables in the program like the weight and recorded the results. The default alpha of 0.8 has the best results, so that remained unchanged. The combine classes was changed to only test for 3 and 5, 1,2, and 4 were removed. The most successful results are highlighted in green, as shown in Table 3. With 1,4,7 doing the best, 1,4,7.4-7.9 were tested and demonstrated no change, so 1,4,7 remained best. Table 4 shows the results at different n-folds for 1,4,7. With 6 folds being the best and 1,4,7 having the best results, different weights were tested and 4.8 was the best, as shown in Table 5.

Table 3: Week eight testing, percentage of accuracy with the training and mystery data. 

Table 4: Different n-folds tested at the generated features 1,4,7. Their results of percentage of accuracy for the training and mystery data.

Table 5: At 6 n-folds and 1,4,7, a weight of 4.8*50/(677+4.8*50) shows the best percent accuracy for the training and mystery data.

Week Nine, May 25
Group 5 double checked their results for the code. The final report was edited and updated with new results. Group 5 started writing up the presentation for the project.

Week Ten, June 1
Group five edited and added to their final drafts. Tables, figures and flowcharts were added to the paper. The PowerPoint presentation was made and finished. Group 05 designated who would read and present what slides. Final presentation was on Friday, June 5.