 |
||||||||
Canonical Response Function/Canonical Link Function - Python and Machine Learning for Integrated Circuits - - An Online Book - |
||||||||
| Python and Machine Learning for Integrated Circuits http://www.globalsino.com/ICs/ | ||||||||
| Chapter/Index: Introduction | A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Y | Z | Appendix | ||||||||
================================================================================= For each exponential family distribution, there is a particular response function called the canonical response function that has nice mathematical properties. The canonical response function can be given by, E[y; η] = g(η) ------------------------------------------------------ [3858a] Canonical link function is given by, η = g-1(μ) -------------------------------------------------- [3858b] We also have, The canonical parameters are parameters specific to the probability distribution chosen for the GLM. They are often denoted by θ (theta) and are related to the variance of the response variable. The canonical parameters are derived from the natural parameters and the probability distribution's properties. They are used to define the probability distribution's shape and characteristics. ============================================
|
||||||||
| ================================================================================= | ||||||||
|
|
||||||||
------------------------------------------- [3858c]