Abstract:Causal Machine Learning (CausalML) is an umbrella term for machine learning methods that formalize the data-generation process as a structural causal model (SCM). This perspective enables us to reason about the effects of changes to this process (interventions) and what would have happened in hindsight (counterfactuals). We categorize work in CausalML into five groups according to the problems they address: (1) causal supervised learning, (2) causal generative modeling, (3) causal explanations, (4) causal fairness, and (5) causal reinforcement learning. We systematically compare the methods in each category and point out open problems. Further, we review data-modality-specific applications in computer vision, natural language processing, and graph representation learning. Finally, we provide an overview of causal benchmarks and a critical discussion of the state of this nascent field, including recommendations for future work.

What problem does this paper attempt to address?

Based on the provided text content, the main problem that this paper "Causal Machine Learning: A Survey and Open Problems" attempts to solve is the lack of causal formalization in modern machine - learning systems. Specifically, the paper points out that current machine - learning techniques are excellent at finding associations in independent and identically distributed (i.i.d.) data, but face the following challenges in practical applications: 1. **Decline in generalization performance**: When the data distribution changes, the generalization performance of the model declines significantly. 2. **Insufficient sample control in generative models**: There is insufficient fine - grained control of samples in generative models. 3. **Prediction bias**: There are unfair biases in the prediction results of certain subgroups. 4. **Explanatory problems**: The explanations are too abstract and not related to specific problems. 5. **Instability of reinforcement - learning methods**: The application of reinforcement - learning methods in practical problems is unstable. These challenges are partly due to the lack of causal formalization in modern machine - learning systems. Therefore, the paper focuses on causal machine learning (CausalML), that is, machine - learning methods that use causal knowledge to model systems, aiming to solve these problems through structural causal models (SCM). SCM allows us to estimate what will happen after intervening in the data - generation process and can model counterfactual consequences given what has occurred. The main contributions of the paper include: 1. **Brief introduction of causal concepts**: It provides a completely self - contained introduction to causal concepts, assuming that the reader has no prior causal knowledge. 2. **Classification of CausalML**: It classifies the existing causal machine - learning work into five categories: causal supervised learning, causal generative modeling, causal explanation, causal fairness, and causal reinforcement learning, and compares the methods in each category. 3. **Applications in specific data modalities**: It reviews causal applications in computer vision, natural language processing, and graph - representation learning. 4. **Causal benchmarks**: It discusses causal benchmark testing. 5. **Analysis of advantages and disadvantages**: It discusses the advantages, challenges faced by causal machine learning, and the inevitable costs of using these techniques. Overall, this paper attempts to solve the above challenges by systematically reviewing and classifying the methods of causal machine learning and providing guidance for future research.

Causal Machine Learning: A Survey and Open Problems

Causal machine learning for healthcare and precision medicine

A survey of deep causal models and their industrial applications

From Identifiable Causal Representations to Controllable Counterfactual Generation: A Survey on Causal Generative Modeling

A Survey of Methods, Challenges and Perspectives in Causality

Methods and tools for causal discovery and causal inference

Evaluation Methods and Measures for Causal Learning Algorithms

CausalML: Python Package for Causal Machine Learning

Causal Reinforcement Learning: A Survey

What if? Causal Machine Learning in Supply Chain Risk Management

Special Issue on Causal Discovery and Causality-Inspired Machine Learning

Open problems in causal structure learning: A case study of COVID-19 in the UK

Causality Learning: A New Perspective for Interpretable Machine Learning

A Survey on Causal Reinforcement Learning

A Survey of Learning Causality with Data

On the Opportunity of Causal Deep Generative Models: A Survey and Future Directions

A Survey on Causal Representation Learning and Future Work for Medical Image Analysis

Improving Causal Reasoning in Large Language Models: A Survey

Towards Understanding How Machines Can Learn Causal Overhypotheses

Causal Feature Selection for Responsible Machine Learning