Notebook Entry

  • Purchasing

    • [x] Check on computer and docking stations.
    • [x] Check on Newegg order.
  • Lab Website

    • [x] Enable MathJax
  • BMD 2013

    • Whipple ID Paper

      • [x] Generated R^2 values table for different models versus different data sets.
      • [x] Copy table from thesis that shows the mean VAF for simulating the models
      • [x] Make figure of instrumented bicycle with labels
      • [x] Reduce the size of the instrumented bicycle image
      • [x] Show identified parameters of best model compared to whipple
      • [x] Make plot of L-P root locus or eig components vs speed and bode plots
      • [x] Center all figures/tables
      • [x] Decrease the size of the figures
      • [x] Submit paper
    • Steer torque paper

      • [x] Compute R^2 or RMS for all torque measurements
      • [x] Submit paper
    • [x] Send in copyright forms


I basically worked non-stop this week and last to get my two papers done for BMD2013. Now I have a headache, but I think things turned out pretty good. The papers and source are available here: It is pretty close to having full computational reproducibility.

In the Whipple ID paper I realized I was computing the VAF is a silly way in my dissertation. So I got all the code running again and redid those calcs. Of course the results came out a bit different and I'm still not sure how to interpret it all. I also can't figure out why dtk.process.fit_goodness() gives different answers than my new dtk.process.coefficient_of_determination() function. I wrote some extensive tests that pass, but there is still some bug somewhere.

I've also been trying to upload my dissertation data onto figshare for easier dissemination and citing. But Figshare's UI is so damn confusing and it also has size limits which are a major limitation. Here is what I wrote to the help desk there:

I have some data sets from some experimental studies that I'd like to upload to figshare for public sharing but I'm having trouble uploading it an efficient and organized way.

Here are basic descriptions of the datasets:

  • 762 compressed data files organized in subdirectories totaling 830 mb
  • 1 compressed HDF5 file totaling ~ 300 mb
  • 525 mp4 video files of experiments totaling 2.5 Gb
  • 1,202 data and text files in a particular subdirectory structure totaling 120 mb

I'd like to be able to upload these five sets as single compressed tar balls so that a user simply has to download the tar ball, uncompress it, and run the software that works with the data. Right now, I can't upload the files that are larger than 250 mb due to your limit. I've also tried uploading the files individually but the interface makes this unwieldy and it seems impossible to keep the directory structures I want. I could break up all of the datasets into multiple tar balls (except the HDF5 file) and upload them in that fashion but it is not preferable. Do you have recommendations on how to share this data with your service?

The response I got was that Figshare doesn't allow big files and they will have a subscription that you can pay for to remove the file size limits. That's lame but there is little reason to think any web service is gonna let you upload a ton GB's of data and they front the cost of distributing that data. It'd be cool if Universities fronted the cost by paying Figshare. Then all researchers at that Uni could upload there stuff. Many universities roll their own institutional repositories, but Figshare could potentially be a better option than that.

My time tracker says I've spent 48 hours on the Whipple ID paper, 20.2 hours on the steer torque paper, and a total of 73 hours for BMD related things since the beginning of July. I need to focus on walking work now.

Time Assessment

Since I've been here this is how I've spent my time:

Since July 8th I've worked 504.4 hours. There have been 60 weekdays and 1 holiday = 59 workdays. So that's averaging 8.55 hours a day of actual work time (this number doesn't count internet surfing, breaks, general chit chatting, or anything that isn't actually doing work. I only log real work time.). I even took a week off in July (but worked about 17 hours that week.) and still averaged a lot of hours a week.

The approximate break down of the type of work I've done is: 290 hours have been doing stuff related to the HMC lab. 35 has been for CSU related stuff. And 175 hours has been doing my research interests. So the percentage is:

HMC 58%
CSU 7%
Personal Research 35%

But I don't feel like I've gotten hardly anything done in the HMC side with respect to the research project. I've spent 85 hours of the 290 on actually doing the walking system id. The rest has been website admin, purchasing, training, lab meetings, mentoring, and my lab notebook. So I've only spent 16% of my "work" time doing the research project Ton has tasked me with. Which is why I've made more progress on my research stuff than the walking stuff. I'm not sure if this counts as being productive for my job...

But I think this will change now. I've gotten over the large portion of the BMD conference hump. GSoC is over. I've gotten over the being new at CSU hump. We have more help in the lab for random things. I'm going to contract out the web work that still remains. So from now on, I think I'll be significantly more productive with the walking research projects. But I am going to take on a senior design group, which will probably take a similar amount of time as GSoC. I'd also really like to teach a class too. It's funny how time gets so much more valuable as you get older.