B1805
Title: Efficiency coresets techniques with multivariate conditional transformation models
Authors: Zeyu Ding - TU Dortmund (Germany) [presenting]
Abstract: In the complex realm of big data, achieving efficiency in large-scale regression analysis is essential. An innovative method is presented that integrates Coresets techniques with multivariate conditional transformation models. Through this integration, effective sample size compression is realized. A distinctive feature of the approach is its ability to maintain the likelihood within a $1\pm\epsilon$ error range. This precision in likelihood, combined with the reduced sample size, empowers the model to handle larger and more intricate datasets. As data scales and diversifies, the method stands as a solution, ensuring rigorous and scalable regression analyses.