I originally wired two of my motors in parallel in order to reduce current draw, but it turns out you can't do that with ESCs. I assumed I could use the two motors as voltage dividers and assume constant resistance, but that wasn't the case. So I contacted HobbyKing and requested a battery that allowed a higher draw, so I could wire all four motors in parallel. Unfortunately, its delivery took three weeks (oh Hobbyking...). After problem solving for about a month now, though, I got a successful four motor test! Techmasters, one of the clubs that I currently am a co-president of, updated its website! The Webmaster did such a great job that I thought it was worth sharing! Here's the link - https://techmasters.io/
Just learned that last year's PACTF (Phillips Academy Capture the Flag) was featured on Carnegie Mellon's website!! Looking forward to this year's competition. As problem master, I can guarantee that the problems will be original and unique.
I'm trying to teach a GAN to reproduce musical scales, but it is proving very difficult. I'm training on an AWS Tesla K80, but it is still taking forever. In addition to that, this graph makes very little sense to me. Usually, the discriminator would converge to fifty percent and the generator would converge to one hundred percent, but I am at at iteration 250,000 and this still isn't happening. I also have no clue what that kerfuffle was at the beginning (Maybe it's just getting over a local minimum?). I'll keep you posted on any new information.
I recently flew overseas on an eight hour flight, so I decided to get some work done on the plane. I learned the basics of Django and D3. I then coded my team's MIT Launch MVP (Minimal Viable Product). I spent fifteen hours working on that and stayed awake for a total of forty hours working and adjusting to the new time zone. Right before I went to sleep, I tried to upload to Git and accidentally force pulled. My work was overwritten, and I was devastated. Then, just as I had half-convinced myself to accept the Buddhist idea of calm acceptance, I remembered that all my work was stored in my Git history. Of course I didn't realize this until I ran TestDisk. I got all my work back! Then, I accidentally deleted all of my static files, but I rewrote those fast enough. Clearly, I need more sleep.
Although my team (Parallax 8900) was disappointed in our performance at this competition, we won the skills component, which qualifies us for the World Vex Competition.
I was also reminded about our awesome Vex videos from last year. I designed and uploaded an application onto the Vex Cortex from a Mac! Usually, you have to use a Windows computer running outdated firmware and RobotC, but I managed to get it to work with Atom on a Mac! Huge shoutout to the people at Purdue Robotics for their help. Their repository is here, though it is not entirely working yet. Getting it to work required some debugging of PySerial and the call to flush the serial. If you need help setting it up for yourself, feel free to contact me.
The other day, my tech team noticed that our school's server was running more slowly than usual and that Fail2Ban was freaking out. We were getting DDoSed. We used an open source Geo-IP service and discovered that an IP sending hundreds of requests per second was from China. Our school's server is fairly nice, so it might be a target for further attacks. We banned the Chinese IP and haven't had any problems since. One of my more security conscious friends felt vindicated because in this particular case, his security concerns were apparently spot on. |
AuthorGotta make a thing or two. Archives
April 2018
Categories |