Causal Discovery on Irregular Time Series
The paper adapts PCMCI+ for event streams that do not arrive on a regular clock.
Rather than using fixed lags, the method aggregates causal influence across predefined temporal windows. The authors test it on synthetic irregular event streams where the causal graph is known. Across different signal-to-noise ratios, they report consistent graph recovery and better results than standard PCMCI+ on irregularly sampled data. ArXiv · AI/CL/LG's note
Rather than using fixed lags, the method aggregates causal influence across predefined temporal windows. The authors test it on synthetic irregular event streams where the causal graph is known. Across different signal-to-noise ratios, they report consistent graph recovery and better results than standard PCMCI+ on irregularly sampled data. ArXiv · AI/CL/LG's note
score 4