Jumping on Entrepreneurship

Parkour, Startups, and Travel


Read Next

Well I've had my Phil...ly

Artem, Charlie and I travelled down to Philadelphia this past weekend for a jam at Drexel University. It was quite awesome. Pretty good turnout, had people from NYC/NJ, Rochester, Buffalo, Philly, and even Baltimore, MD.

We got to Philly at about 2AM on Saturday morning, and promptly went to sleep. We woke up at 8:30 to meet the rest of the tracuers coming in and we started jamming! We began at 30th Street Station doing some precisions on Jersey Barriers until the K9 unit showed up... After that we met up with M1l3s from NYPK - he was running 10 miles for the Step Up to Fight Diabetes charity event. Then it was a full day of walking, from Drexel's campus, to Temple, UPenn, Central City/City Hall, Philadelphia Museum of Art (where the stair scene from Rocky was filmed), and more. Saturday night, people headed back to Drexel to go to sleep, except for Jessie, Pyro, Blaarg and I. We decided to try and train all night long. We may not have ended up training all night, but we were out in Philly all night, and it was really awesome. We did SO much walking and we were exhausted, but spending the entire night in a foreign city is definitely a cool experience.

Sunday was spent doing a little bit more training at the Museum, where we had three or four groups of people come up to us and ask us what we were doing. We demonstrated, and taught, some pretty basic precisions, and then we got our pictures taken. After that, we were all very tired, and we started the 6 hour drive back home.

Thanks to everyone for making this an awesome Jam!

In other news, I'm in the middle of reorganizing all of my pictures and videos. Here's a new link to a (probably) temporary Picasa page with two albums (one for pictures, one for videos). Most are fairly new (a lot are from Philly infact, so if you want to see some of the pictures and videos from there, check it out!). Hopefully more will be added soon!

Q Learning


I am going to try and implement some kind of Q learning as my next project. I'm going to develop an AI for a game. The game is Tic Tac Toe.

I'm pretty psyched about it.

This can be used for a lot of things it seems, anything that's controlled by frames. You can determine an action once per frame with as much evaluation as you like, and then you could even nest those into each other, like a neural net or something.

I've heard of some too, where you can force a decision at anytime. So I could run for 50 milliseconds exactly and then make a decision and then over time, the AI controlling that AI would be able to produce more efficient results.

I've even thought about the possibility of synthesizing sexprs with Qlearning. It's easy to sandbox the code with Racket, then I just use the test facility to check output and train it from there to do what I want. Eventually I could even hook it up to the profiler and optimize performance and disk usage.

Rendering New Theme...