Skip to main content

Ctrl+K

Introduction

Overview

Overview

Motivating Examples

Causal Structure Learning (CSL)
Causal Effect Learning (CEL)
Causal Policy Learning (CPL)

Preliminary

Causal Inference Preliminary

Causal Structure Learning (CSL)

Preliminaries of Causal Graphs
Causal Discovery
Causal Mediation Analysis

Causal Effect Learning (CEL)

Single Stage – Paradigm 1
Markov Decision Processes – Paradigm 2
Panel Data – Paradigm 3

Causal Policy Learning (CPL)--Paradigm 1

Single Stage
Discrete Action Space
Continuous Action Space
Plan To Do
- Entropy learning

Causal Policy Learning (CPL)--Paradigm 2

Preliminary: Off-policy Evaluation and Optimization in Markov Decision Processes
Policy Evaluation–Value Estimation
Policy Evaluation--Mediation Analysis
Policy Optimization
- Fitted-Q Iteration

Causal Policy Learning (CPL)--Paradigm 3

Multiple Stages (DTR)
Q-Learning
A-Learning

Causal Policy Learning (CPL)--Paradigm 4

Overview: Bandits ALgorithm
Multi-Armed Bandits (MAB)
Contextual Bandits
- LinUCB
- LinTS
Meta Bandits
- Meta Thompson Sampling
- Multi-Task Thompson Sampling (MTTS)
Structured Bandit (Slate Recommendation)
Online Policy Evaluation

Causal Policy Learning (CPL)--Paradigm 5

Online Policy Learning and Evaluation in Markovian Environments

Causal Policy Learning (CPL)--Paradigm 6

Ooline Policy Learning in Non-Markovian Environments

Case Studies

Mimic3
MovieLens

Repository
Open issue

.md

Discrete Action Space

Discrete Action Space#

In this section, our focus is on single-stage policy learning that involves actions which are either binary (0 or 1) or multinomial (choices A, B, C, or D).

previous

Single Stage

next

Q-Learning

By Causal Decision Making Team

© Copyright 2022.