analiza aiqbench
This paper uses convolutional neural networks followed by MLPs to map five-minute return data directly to a monthly skewness estimate, aiming to reduce the finite-sample bias of conventional realized-skewness estimators. The evidence is mostly simulation-based (stochastic volatility models with and without jumps) plus a portfolio-sorting exercise, and the consistency claims rest on stated regularity conditions. For ML readers, the useful point is the design choice of fixing a temporal aggregation architecture before learning, which treats a learned estimator as a statistical tool with explicit bias and risk properties rather than a black-box predictor.
Estimating "Realized" Skewness using Convolutional Neural Network
arXiv:2611.04598v1 Announce Type: cross Abstract: We propose a new estimator of low-frequency skewness (monthly) that exploits high-frequency data (5 min) through a direct functional mapping consisting of layers of Convolutional Neural Networks followed by Multi Layer Perceptrons. Monte Carlo experiments involving stochastic volatility models, with and without jumps, show that the estimator reduces finite-sample bias relative to existing realized-skewness estimators and remains accurate under model misspecification. Empirically, our estimator exhibits temporal stability and delivers cross-sectional performance in actual skewness-sorted portfolios. We characterize the finite-sample bias of conventional plug-in skewness estimators and establish risk consistency properties of the proposed NN estimator under standard regularity conditions.
Abstract:We propose a new estimator of low-frequency skewness (monthly) that exploits high-frequency data (5 min) through a direct functional mapping consisting of layers of Convolutional Neural Networks followed by Multi Layer Perceptrons. Monte Carlo experiments involving stochastic volatility models, with and without jumps, show that the estimator reduces finite-sample bias relative to existing realized-skewness estimators and remains accurate under model misspecification. Empirically, our estimator exhibits temporal stability and delivers cross-sectional performance in actual skewness-sorted portfolios. We characterize the finite-sample bias of conventional plug-in skewness estimators and establish risk consistency properties of the proposed NN estimator under standard regularity conditions.
This paper first proves that bias exist given a fixed-horizon of high-frequency data to estimate the low-frequency skewness by using the plug-in method. Then it provides an innovation machine learning method to approximate the optimal estimator by fixing a temporal aggregation architecture for neural network before learning
Statistics Theory (math.ST); Machine Learning (cs.LG)
arXiv:2611.04598 [math.ST]
(or arXiv:2611.04598v1 [math.ST] for this version)
https://doi.org/10.48550/arXiv.2611.04598 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 167 cuvinte
Subiecte conexe
FasTac: A Curved Multispectral Vision-Based Tactile Sensor for High-Speed High-Precision 3D Shape and Force Perception
arXiv · 31 iul. 2026
WAND: Learning Robust Navigation under Complex Wind Disturbances and Dense Obstacles for Quadrotors
arXiv · 9 oct. 2026
Time-warping estimation via stationarity-based learning of the de-warped signal
arXiv · 16 sept. 2026
Factorized Spatio-Temporal Convolutions for Human Pose Estimation from Planar Lidar
arXiv · 24 iul. 2026