セミナー

談話会

タイトル	Parsimonious statistical analysis for multivariate models of high dimension
開催日時	3 月 16 日（月）16:30-17:30
主催者	先端数理科学研究センター
講演者	ポワニャール ベンジャミン 氏（慶應義塾大学理工学部）
場所	日吉独立館D410
内容	Inherent to most multivariate models, the curse of dimensionality refers to the explosive increase in the number of parameters to be estimated as the dimension of the underlying stochastic vector grows. Within this framework, balancing a model that is sufficiently rich to capture complex relationships while remaining parsimonious enough to avoid overfitting is a key challenge. This issue can be addressed using sparsity-based methods, which focus on variable selection to recover the true signal when the model admits a sparse representation. Penalization techniques are typically applied to the objective function (e.g., likelihood or least squares) during estimation to produce an estimator with many coefficients set to zero. The following topics will be considered: illustrations of multivariate models affected by the curse of dimensionality; theoretical properties of sparsity-based estimators; optimization methods. ================================================================== 談話会終了後の懇親会を開催します。ご出席される方は 3/9（月）までに以下のリンクからお知らせください。 https://schd.adst.keio.ac.jp/ksa/join-event?accessId=432ee6e9417b42739dacbdf517172335
資料
URL