Displaying 25-28 of 35 results.
January 2
Successful
Success
Success

FerriteGiant
FerriteGiant
- Referee approval report
SigmaX
SigmaX
- Committed user success report
2+ hrs first things. Organized my notes on some experiments from yesterday, and identified two research questions to guide the short term. Decided that a population-seeding transfer experiment across random mazes is still the way to go for a first result. Got started by writing a function to easily visualize the maze files that I generate. Topped it off with 20 minutes of peer review.
January 1
Successful
Success
Success

FerriteGiant
FerriteGiant
- Referee approval report
SigmaX
SigmaX
- Committed user success report
2hrs in first thing, working toward the goal of having easy-to-configure experiments that compose different pieces. Found that Hydra's pre-release version is in PyPI, so installed and started playing around. Seems it can't be used in conjunction with CLI libraries like click or argparse, which is a bummer. Setting this topic aside for now, because I have a project on my day job that will give me a chance to investigate Hydra more deeply.
December 31
Successful
Success
Success

FerriteGiant
FerriteGiant
- Referee approval report
SigmaX
SigmaX
- Committed user success report
2hr even though it's a holiday (I smell a habit!). Working toward having nice, one-stop commands for different kinds of experiments (i.e. to "evolve" a controller, and to "run" a controller for testing, and to mix-and-match different evolutionary algorithms, environments, controller representstions, etc.). Looked into Facebook AI's new hydra package for configuration. Started to implement it for my experiments, but found that Python 3.9 support isn't in the stable version yet: https://github.com/facebookresearch/hydra/issues/1062
December 30
Successful
Success
Success

FerriteGiant
FerriteGiant
- Referee approval report
SigmaX
SigmaX
- Committed user success report
Woke up with just 1.5 hours to my first meeting today, so almost failed! Rescheduling the meeting saved the day and got me the full 2 hours. Today's advance was to refactor my simulation to make it conform better to the OpenAI Gym "Env" interface. This makes it easier to select an environment by name via a command-line argument, and to test my algorithms on standard environments like CartPole. Topped it off with 15 minutes of peer review reading.