Sample Complexity of Estimation in Logistic Regression - Lecture

Explore the intricacies of sample complexity in logistic regression parameter estimation through this comprehensive lecture presented at IPAM's EnCORE Workshop. Delve into the non-asymptotic analysis of sample complexity, examining its relationship with error and inverse temperature in logistic regression models. Discover the three distinct temperature regimes—low, moderate, and high—and understand how they impact the sample complexity curve. Gain insights into the challenges of estimating parameters with a given ℓ2 error, considering factors such as dimension and inverse temperature with standard normal covariates. Compare this approach to traditional generalization bounds and asymptotic performance analyses of maximum-likelihood estimators. Enhance your understanding of binary classification problems and the logistic regression model's role in noisy data generation processes.

Arya Mazumdar - Sample complexity of estimation in logistic regression - IPAM at UCLA