|
|
|
 |
|
|
Seminarium 15.11.2022
dodano: 10/11/2022
przez: Dariusz Zawisza
|
- Referent : Jarosław Duda
- Tytuł referatu : Hierarchical correlation reconstruction – between statistics and machine learning
- Abstrakt : While machine learning techniques are very powerful, they have some weaknesses, like iterative optimization with many local minimums, large freedom of parameters, lack of interpretability and accuracy control. From the other side we have classical statistics based on moments not having these issues, but providing only a rough description. I will introduce and show on various applications (e.g. financial, medical) HCR family of methods combining their advantages: with MSE-optimal moment-like coefficients, but designed such that we can reconstruct (joint) probability distributions from them, also modeled in adaptive/evolving way, or predicted from other information
- Slajdy: Link
- Termin: Wtorek 15.11.2022, 12:15, sala 1093
|
|
Seminarium 08.11.2022
dodano: 04/11/2022
przez: Dariusz Zawisza
|
- Referent : Kewin Pączek
- Tytuł referatu : Modele hybrydowe w prognozowaniu szeregów czasowych.
- Termin: Wtorek 04.11.2022, 12:15, sala 1093
|
|
Seminarium 25.10.2022
dodano: 21/10/2022
przez: Dariusz Zawisza
|
- Referent : Jarosław Duda
- Tytuł referatu : Electron diffusion model of semiconductor p-n junction (diode) using Maximal Entropy Random Walk
- Abstrakt : Standard diffusion incorrectly predicts nearly uniform stationary probability distribution -incorrect e.g. for electrons or neutrons, for example making semiconductor a conductor with linear Ohm law. To repair it, there was instead used diffusion chosen accordingly to the maximal entropy principle (Maximal Entropy Random Walk) – getting the same stationary probability distribution as quantum ground state, with Anderson-like localization. Including mean-field self-interaction between electrons, there was obtained proper asymmetric non-linear Ohm law for model of semiconductor p-n junction (diode) – with conductance easy only in one direction
- Slajdy: Link
- Termin: Wtorek 25.10.2022, 12:15, sala 1093
|
|
|
|