ChefsHatPlayersClub 2.2.6

A collection of ready-to-use agents for the Chef's Hat Gym Environment

Chef's Hat Players Club

The Chef's Hat Player club provides ready-to-use agents for the ChefsHatGym simulator.

Instalation

The instalation can be done via pip:

pip install ChefsHatPlayersClub

##Available Agents

Agents are divided into types, currently these are the types of available agents. You can check th examples folder for more examples!

Classic Agents

The classic agents are the ones available in our initial investigations with the Chef's Hat:

• Barros, P., Tanevska, A., & Sciutti, A. (2020). Learning from Learners: Adapting Reinforcement Learning Agents to be Competitive in a Card Game. arXiv preprint arXiv:2004.04000.

Each type of the agent refers on how the agent was trained. Refer to the above paper for more information.

Agent	Types
DQL	"Scratch", "vsRandom", "vsSelf", "vsEveryone"
PPO	"Scratch", "vsRandom", "vsSelf", "vsEveryone"

Each agent can be instantiated using:

agent = AgentDQL(name="DQL4", continueTraining=False, type="vsEveryone", initialEpsilon=1, verbose=True, logDirectory="")  # training agent

where the initialization parameters are:

Parameter	Description
name	name of the agent, must be unique in a game
continueTraining	if the agent will learn during the game
type	type of the agent, see table above
initialEpsilon	when learning, initial exploration value
saveFolder	folder that the agent will be saved in
verbose	verbose when learning
logDirectory	The directory where the agent log will be located

Karma Chameleon Club

Implementation of an Adversarial Inverse Reinforcement Learning Agent (AIRL), and instances of trained agents based on human data collected from Chef's Hat Online.

Type	Type	Type	Type	Type
lil_abcd_	lilAbsol	lilAle	lilAna	lilArkady
lilAuar	lilBlio1	lilBlio2	lilChu	lilDa48
lilDana	lilDJ	lilDomi948	lilEle	lilFael
lilGeo	lilIlzy	lilJba	lilLeandro	lilLena
lilLordelo	lilMars	lilNathalia	lilNik	lilNilay
lilRamsey	lilRaos	lilThecube	lilThuran	lilTisantana
lilToran	lilWinne	lilYves	lilYves2

Each agent can be instantiated using:

agent = AgentAIRL(name="AIRL", continueTraining=False, type="lilDJ", initialEpsilon=1, verbose=True, logDirectory="")  # training agent

where the initialization parameters are:

Parameter	Description
name	name of the agent, must be unique in a game
continueTraining	if the agent will learn during the game. Must have a demonstration if continue learning is True.
demonstrations	must be a npy with a list of ([state, action, possibleActions])
type	type of the agent, see table above
initialEpsilon	when learning, initial exploration value
saveFolder	folder that the agent will be saved in
verbose	verbose when learning
logDirectory	The directory where the agent log will be located

Chef`s Hat Cup 1

Implementations of the agents that won the Chef`s Hat Cup 1. For more information please visit: https://github.com/yvesmendes/ChefsHatPlayersClubAgents/

Agent	Description
AIACIMP	use a custom loss function based on the Hub Loss and a custom reward
AINSA	Use a custom reward
ALLIN	Custom reward
AMYG4	Only winning reward

where the initialization parameters are:

Parameter	Description
name	name of the agent, must be unique in a game
continueTraining	if the agent will learn during the game. Must have a demonstration if continue learning is True.
demonstrations	must be a npy with a list of ([state, action, possibleActions])
initialEpsilon	when learning, initial exploration value
saveFolder	folder that the agent will be saved in
verbose	verbose when learning
logDirectory	The directory where the agent log will be located

agent = AIACIMP(name="01", continueTraining=False, initialEpsilon=0.2, loadNetwork="", saveFolder="", verbose=True, logDirectory=logDirectory)
agent = AINSA(name="01", continueTraining=False, initialEpsilon=0.2, loadNetwork="", saveFolder="", verbose=True, logDirectory=logDirectory)
agent = ALLIN(name="01", continueTraining=False, initialEpsilon=0.2, loadNetwork="", saveFolder="", verbose=True, logDirectory=logDirectory)
agent = AMYG4(name="01", continueTraining=False, initialEpsilon=0.2, loadNetwork="", saveFolder="", verbose=True, logDirectory=logDirectory)

Chef`s Hat Cup 2

Implementations of the agents that won the Chef`s Hat Cup 2.

Agent	Description
Larger_Value	Discard the larger amount of high cards, avoiding to pass
BOOM	Discard the larger amount of high cards
PPO_V2	PPO algorithm trained with a custom reward based on finishing position

agent = Bloom(name="001", saveModelIn="", verbose=True, savelogDirectory=logDirectory)
agent = AgentLargerValue(name="002")
agent = PPO_V2(name="003"continueTraining=False, initialEpsilon=1, loadNetwork="", saveFolder="", verbose=False, logDirectory=logDirectory)

Citations

• Barros, P., Yalçın, Ö. N., Tanevska, A., & Sciutti, A. (2023). Incorporating rivalry in reinforcement learning for a competitive game. Neural Computing and Applications, 35(23), 16739-16752.

• Barros, P., & Sciutti, A. (2022). All by Myself: Learning individualized competitive behavior with a contrastive reinforcement learning optimization. Neural Networks, 150, 364-376.

• Barros, P., Yalçın, Ö. N., Tanevska, A., & Sciutti, A. (2022). Incorporating Rivalry in reinforcement learning for a competitive game. Neural Computing and Applications, 1-14.

• Barros, P., Tanevska, A., & Sciutti, A. (2021, January). Learning from learners: Adapting reinforcement learning agents to be competitive in a card game. In 2020 25th International Conference on Pattern Recognition (ICPR) (pp. 2716-2723). IEEE.

• Barros, P., Sciutti, A., Bloem, A. C., Hootsmans, I. M., Opheij, L. M., Toebosch, R. H., & Barakova, E. (2021, March). It's Food Fight! Designing the Chef's Hat Card Game for Affective-Aware HRI. In Companion of the 2021 ACM/IEEE International Conference on Human-Robot Interaction (pp. 524-528).

• Barros, P., Tanevska, A., Cruz, F., & Sciutti, A. (2020, October). Moody Learners-Explaining Competitive Behaviour of Reinforcement Learning Agents. In 2020 Joint IEEE 10th International Conference on Development and Learning and Epigenetic Robotics (ICDL-EpiRob) (pp. 1-8). IEEE.

• Barros, P., Sciutti, A., Bloem, A. C., Hootsmans, I. M., Opheij, L. M., Toebosch, R. H., & Barakova, E. (2021, March). It's food fight! Designing the chef's hat card game for affective-aware HRI. In Companion of the 2021 ACM/IEEE International Conference on Human-Robot Interaction (pp. 524-528).

Contact

Pablo Barros - [e-mail]

Alexandre Rodolfo - [e-mail]

• Twitter
• Google Scholar