Sökning: "Non-asymptotic estimation"

Hittade 3 avhandlingar innehållade orden Non-asymptotic estimation.

1. 1. Data driven modeling in the presence of time series structure: : Improved bounds and effective algorithms

   Författare :Othmane Mazhar; Boualem Djehiche; Munther Dahleh; KTH; []
   Nyckelord :NATURVETENSKAP; NATURAL SCIENCES; Time series; Non-asymptotic estimation; Minimax; Change point detection; Hidden Markov model; State space model; Least square; Penalized Regression; Random covariance matrix; Concentration inequality; Chaining integral; Self-normalized martingale inequality; Cramér-Rao inequality; van Trees inequality; Matematisk statistik; Mathematical Statistics;

   Sammanfattning : This thesis consists of five appended papers devoted to modeling tasks where the desired models are learned from data sets with an underlying time series structure. We develop a statistical methodology for providing efficient estimators and analyzing their non-asymptotic behavior. LÄS MER

3. 2. Case studies in omniparametric simulation

   Författare :Fredrik Lundin; Göteborgs universitet; []
   Nyckelord :NATURVETENSKAP; NATURAL SCIENCES; growth model; Ising model; Markov chain; omnithermal simulation; omniparametric simulationpercolatiion; Potts model; parameter estimation; partial observations; random cluster model; Richardson model; simulation driven parameter estimation; two-type Richardson model; omnithermal simulation;

   Sammanfattning : In the eld of particle systems and growths models simulation is an important tool. When explicit calculations are too complex or impossible to perform we may use simulations instead. We adapt a new technique here denoted omniparametric simulation, to the two-type Richardson, Ising and Potts models. LÄS MER

4. 3. Statistical Learning in Linearly Structured Systems: Identification, Control, and Reinforcement Learning

   Författare :Yassir Jedra; Alexandre Proutiere; Alexander Rakhlin; KTH; []
   Nyckelord :TEKNIK OCH TEKNOLOGIER; ENGINEERING AND TECHNOLOGY; Machine Learning; Statistical Learning; Control Theory; Reinforcement Learning; System Identification; Electrical Engineering; Elektro- och systemteknik;

   Sammanfattning : In this thesis, we investigate the design and statistical efficiency of learning algorithms in systems with a linear structure. This study is carried along three main domains, namely identification, control, and reinforcement learning, and is presented as a collection of five papers. LÄS MER