ݺߣ

ݺߣShare a Scribd company logo
try! brain.think()
Machine Learning y Aprendizaje
Reforzado en iOS
Ricardo Montesinos
github: richimf
?Que es Reinforcement Learning?
https://www.bbc.com/news/technology-35420579
Aprendizaje reforzado con swift
Reinforcement
Learning
Algorithms
https://medium.com/@jonathan_hui/rl-reinforcement-learning-algorithms-quick-overview-6bf69736694d
Policy EvaluationValue Iteration(Dynamic Programming)
Q-Learning
Deep Q-learning (deep network to estimate Q )
Policy Gradients (uses likelihood)
Actor-Critic Algorithm(Combining policy gradient and value-
function learning).
Model-based Reinforcement Learning(Replan actions in
every action taken).
Q-Learning
Agent Reward Environment
+1 -1
State & Reward
Actions
Agent: Link Environment: Map
La idea de RL es obtener la mayor cantidad
de recompensas.
Algorithm
https://medium.freecodecamp.org/diving-deeper-into-
reinforcement-learning-with-q-learning-c18d0db58efe
Usamos la funcin de Valor-Accin Q para obtener
el Reward esperado dado el State y el Action.
Algorithm
La meta del Q-Learning es maximizar el valor de la
funcin Q. Entre mas Reward obtengamos mejor.
Nuestro Agente utilizar el Q-Table para saber
cuales son las mejores acciones a seguir para
conseguir las recompensas.
Q-Table
https://medium.freecodecamp.org/diving-deeper-into-
reinforcement-learning-with-q-learning-c18d0db58efe
actions
states
Demo
import BrainyRL
Beta
Aprendizaje reforzado con swift
functional programming
functional programming
// some code implementation
// some code implementation
// some code implementation
try! brain.think()
Aprendizaje reforzado con swift
https://link.medium.com/n5GeASQ7fU

More Related Content

Aprendizaje reforzado con swift