C-HDNet: A Fast Hyperdimensional Computing Based Method for Causal Effect Estimation from Networked Observational...

Area of Science:

• Causal Inference
• Network Analysis
• Hyperdimensional Computing

Background:

• Observational data analysis is challenged by network confounding, where network structures bias treatment and outcome assignments.
• Traditional causal inference methods struggle with network interference, leading to inaccurate effect estimations.

Purpose of the Study:

• To develop a novel method for estimating causal effects in the presence of network confounding.
• To improve the reliability of causal effect estimates by incorporating network structure information.

Main Methods:

• A novel matching-based approach utilizing hyperdimensional computing principles.
• Encoding and incorporating structural network information for identifying comparable individuals.

Main Results:

• The proposed method achieves performance comparable to or better than state-of-the-art approaches, including computationally intensive deep learning models.
• Demonstrated significant reduction in runtime (nearly an order of magnitude) without compromising accuracy.

Conclusions:

• The novel hyperdimensional computing-based matching approach effectively addresses network confounding in causal inference.
• This method offers a computationally efficient and accurate solution for large-scale or time-sensitive causal effect estimation from observational network data.