<ul><li>Real-world machine learning applications often involve data from multiple modalities that must be integrated effectively for robust predictions.</li><li>Cohort-Based Active Modality Acquisition (CAMA) is introduced as a novel test-time setting to prioritize samples for acquiring additional modalities when resources are limited.</li><li>The method combines generative imputation and discriminative modeling to estimate the benefit of acquiring missing modalities.</li><li>Experiments show that CAMA strategies outperform unimodal information, entropy guidance, and random selection in guiding modality acquisition for improved resource utilization.</li></ul>

Cohort-Based Active Modality Acquisition

Discover more