Adaptive kernel deconvolution: new approach by Penalized Comparison to Overfitting method

par Jade Nguyen (Univ. of sciences, Ho Chi Minh City)

mardi 12 mars 2024, 14:00 → 15:00 Europe/Paris

Salle K. Johnson, 1er étage (1R3)

Kernel density estimation is a well-known method involving a smoothing parameter (the bandwidth)  that needs to be tuned by the user. We present a recently developed bandwidth selection method, called Penalized Comparison to Overfitting (PCO), a new data-driven bandwidth selection proposed by Lacour et al. (2017).Then we consider a convolution model for which we propose an adaptive kernel estimator of a target density measured with additive error. We employ the PCO method in order to construct our adaptive estimator. Some theoretical results are established.