Neal McBurnett nealmcb

Keybase proof

I hereby claim:

I am nealmcb on github.
I am nealmcb (https://keybase.io/nealmcb) on keybase.
I have a public key whose fingerprint is EDC0 36EB 2397 A000 D7CD 417B B94C 992F 3314 8156

To claim this, I am signing this object:

This gist documents my OpenAI evaluations at

https://gym.openai.com/evaluations/eval_bW731pNgS2i1dZB0JvojA [failed to solve]
https://gym.openai.com/evaluations/eval_FrOsJ2oiRb2EB3oh5l5Lg [solved in 23 episodes]

I attempted to reproduce the quickest documented CartPole-v0 solution to-date, reporting 29 episodes to solve,by iaroslav-ai, documented at https://gym.openai.com/evaluations/eval_yCJkgBGRl2Nfn3TKbvkkg

I don't know why I had a different result the first time, or indeed why it was faster the second time.

I also saw it fail with a traceback:

xkcd has some great t-shirts in the xkcd store. One of them, which I got in late 2014, is the uber-geeky "greek" shirt, though I don't see it in the official store right now.

For background, see Greek letters used in mathematics, science, and engineering (Wikipedia), though that doesn't have all the uses listed by Randall.

Here are the letters and uses presented on the shirt, row-by-row:

δ  delta: dirac & kronecker deltas; isotope ratio
ε  epsilon: a small number; strain; permittivity
ψ  psi: quantum wavefunction, the planet Neptune

η eta: impedance, viscosity, efficiency

	sudo apt-get update && sudo apt-get install mycroft-core
	Hit http://mirrordirector.raspbian.org jessie InRelease
	Hit http://archive.raspberrypi.org jessie InRelease
	Hit http://mirrordirector.raspbian.org jessie/main armhf Packages
	Hit http://archive.raspberrypi.org jessie/main armhf Packages
	Hit http://repo.mycroft.ai debian InRelease
	Hit http://mirrordirector.raspbian.org jessie/contrib armhf Packages
	Hit http://mirrordirector.raspbian.org jessie/non-free armhf Packages
	Hit http://archive.raspberrypi.org jessie/ui armhf Packages
	Hit http://mirrordirector.raspbian.org jessie/rpi armhf Packages

	"""pg-pong.py: Trains an agent with (stochastic) Policy Gradients on Pong.
	Uses OpenAI Gym.
	Saves model every 100 episodes. Resume by setting resume = True
	Set render = True to watch the action.

	Modified from https://gist.github.com/karpathy/a4166c7fe253700972fcbc77e4ea32c5
	to print timestamped self-contained progress rows in TSV format (filter
	for just lines containing 'episode').

	For background, see http://karpathy.github.io/2016/05/31/rl/