Just released PtCut 3.0.2 with bugfixes:
- Grid sampling didn’t work under Python3. Fixed.
- Option –stl caused print to screen. Fixed.
- Toyed around with common restrictions. Hopefully nothing is broken. 🙂
Monthly Archives: June 2018
PtCut 3.0.1 and PPLpy for Python 2.7 wheels released
Jonas Weinz has produced pre-compiled wheels of PPLpy for Python 2.7 (for Linux 64-bit). Have look at Jonas’ page here.
I have mirrored the files here as well:
- gmpy2-2.1.0a1-cp27-cp27mu-linux_x86_64.whl
- pplpy-0.7-cp27-cp27mu-linux_x86_64.whl
- pplpy_dependencies-0.0.1-py2-none-any.whl
I have released a slightly updated version 3.0.1 of PtCut that works under plain Python 2.7 as well (and Python 3.x and SageMath).
AI playing video games
Here are some videos of AIs playing video games:
- DQN Breakout and DQN SPACE INVADERS by DeepMind. That’s a real AI learning to play. It only got a 15Hz view of the pixels and learned to play the game.
- NES Tetris Lua AI playing to 999999. It’s real fun to watch, but is it really AI? It makes some errors or at least questionable choices, but maybe that all the more proof of its AI-ness.
- AI Playing FPS: Limited Deathmatch and AI Playing FPS: Full Deathmatch by Devendra Chaplot. That a real AI that only got the pixel inputs and learned to play Doom. Pretty impressive! Also see their paper: Playing FPS Games with Deep Reinforcement Learning. There are some additional videos on Youtube.
- Deep Reinforcement Learning: Pong from Pixels by Andrej Karpathy. This is code that learns how to play Atari 2600 Pong just by “looking” at the pixels. It’s only 140 lines of Python (the Pong game engine is externally provided). Great article!
PtCut v3.0.0 released!
I am happy to announce the latest release of PtCut, my software to calculate Tropical Prevarieties and Tropical Equilibria.
The news for version 3.0.0 is that it now supports pplpy instead of SageMath. That makes it much smaller, faster to start, slightly faster in execution and allows Python 3.x to be used. If you work natively on Linux, you can start it easily from the commandline.
Jonas Weinz has produced a first set of pplpy wheels for 64-bit Linux and Python 3.5 & 3.6, see here.
Share and enjoy!