Apologies for the lack of mid-week post this week, I was concentrating on producing some output from the De Luca & Termini Fuzzy Entropy algorithm 1.
So besides just plodding away poking and prodding the Entropy code, what have I done?
Really started to use Taiga better
Adjusted the trapezium values to more sensible ones
Installed m2html plug in to MATLAB
Fix an 'imread' error
Get some output from the De-Luca Fuzzy Entropy
Taiga
This really is a powerful tool, tracking my progress through the week and being able to track Issues simply and easily. I finally got around to putting in Story points for each User Story (bad Laura not doing that before!)
I had a go at estimating how many sprints I will be doing over the course of the Major Project - however I'm likely to have counted wrong, so this will probably change! As you can see I'm well on target with my current amount of story points, however these are still in flux as I get used to using them.
Just a screenshot of my closed Sprints so far. As you can see I'm still tweaking my estimation on how many story points to bring into each Sprint, but I now think 15-20 is a good level each week.
Issues is really handy in Taiga, with many different options you can choose from.
\n\nAs you can see, you can add a lot of detail, even including attachments to your issue! The Issue can be categorised using the following labels:
- Type
- Bug
- Question
- Enhancement
- Severity
- Wishlist
- Minor
- Normal
- Important
- Critical
- Priority
- Low
- Normal
- High
All of these labels can be further tweaked and new ones can be added in the Admin panel, making sure that your Taiga workspace really fits the project you''re working on.
MUST KEEP USING THIS!
m2html
One hurdle I'm constantly facing is using this pre-existing code by Learned-Miller. The sheer number of functions in his demo whilst being useful, are overwhelming, and at times it's not clear what they do, and where they link. Last week you saw I had started to map out the links between the functions, but the amount of 'spare' functions not linked to anything perturbed me, had I missed something?
This led me to finding 'm2html' - a handy MATLAB .m file to HTML converter. Pair this with GraphViz, and you have a fully fledged HTML and graphical representation of your code.
As you can see, no styling has been applied to this, and it's currently not hosted anywhere online (the html files are in my Git repo). Also the dependency graphs only look at dependencies within that directory, so in the Development example above, it doesn't indicate that binaryCongeal in fact calls the entropy lookup function, as this is in the 'UTILITY' sub-directory.
To-do: Create dependency graphs spanning sub-directories
De-Luca Fuzzy Entropy output
So the big hurrah of this week is finally implementing the De-Luca & Termini Fuzzy Entropy algorithm.
BUT I''m not sure it's outputting the correct output, and it's extremely inefficient (taking 10mins to run 10 iterations on MNIST data). BUT it's improvement!! If I cannot determine the usefulness of the output by the end of the weekend, I will take my laptop in to see Neil on Monday and ask his opinion.
Output after 10 iterations. What do figure 1 and 2 really mean? Hopefully Neil will shed some light for me on Monday!
The runtime statistics - doesn't make for pretty reading.
So next week? Well it's my birthday on Tuesday (yay!) and the CS38220 Poster assignment is due in on Wednesday (this is a scientific poster on the progress of my Major Project) - but hopefully progress won't be too impeded. I may take slightly less story points into next week, and truly focus on getting a useful output out of De-Luca and Termini's algorithm (in time for Mid-Project demo!).
Next week:
- Truly finish implementing De Luca & Termini's algorithm
- Keep working on understanding the existing code & what it outputs
- Finish and hand-in CS38220 assignment
- Start the write-up for the algorithm!