Anki Deck Changes

Commit: 0cdada61 - jonas sag danke ich fix hier iml shit lol

Author: obrhubr <obrhubr@gmail.com>

Date: 2026-02-23T21:30:46+01:00

Changes: 1 note(s) changed (0 added, 1 modified, 0 deleted)

Note 1: ETH::Major: Information and Data Processing::IML

Deck: ETH::Major: Information and Data Processing::IML
Note Type: Horvath Cloze
GUID: xwS2i87ME6
modified

Before

Front

ETH::Majors::Information_and_Data_Processing::IML::1._Regression::1._Loss_Functions
The asymmetric loss is \(\ell_\tau(r) = {{c1::\tau \max\{r,0\} + (1-\tau)\max\{-r,0\}}}\). A higher \(\tau\) means steeper penalty for over-shooting (positive residual), lower \(\tau\) means steeper penalty for under-shooting.

Back

ETH::Majors::Information_and_Data_Processing::IML::1._Regression::1._Loss_Functions
The asymmetric loss is \(\ell_\tau(r) = {{c1::\tau \max\{r,0\} + (1-\tau)\max\{-r,0\}}}\). A higher \(\tau\) means steeper penalty for over-shooting (positive residual), lower \(\tau\) means steeper penalty for under-shooting.

Like absolute loss, but with two different slopes on each side of the y-axis.

After

Front

ETH::Majors::Information_and_Data_Processing::IML::1._Regression::1._Loss_Functions
The asymmetric loss is \(\ell_\tau(r) = {{c1::\tau \max\{r,0\} + (1-\tau)\max\{-r,0\} }}\). A higher \(\tau\) means steeper penalty for over-shooting (positive residual), lower \(\tau\) means steeper penalty for under-shooting.

Back

ETH::Majors::Information_and_Data_Processing::IML::1._Regression::1._Loss_Functions
The asymmetric loss is \(\ell_\tau(r) = {{c1::\tau \max\{r,0\} + (1-\tau)\max\{-r,0\} }}\). A higher \(\tau\) means steeper penalty for over-shooting (positive residual), lower \(\tau\) means steeper penalty for under-shooting.

Like absolute loss, but with two different slopes on each side of the y-axis.
Field-by-field Comparison
Field Before After
Text The asymmetric loss is \(\ell_\tau(r) = {{c1::\tau \max\{r,0\} + (1-\tau)\max\{-r,0\}}}\). A higher \(\tau\) means {{c2::steeper penalty for over-shooting (positive residual)}}, lower \(\tau\) means {{c3::steeper penalty for under-shooting}}. The asymmetric loss is \(\ell_\tau(r) = {{c1::\tau \max\{r,0\} + (1-\tau)\max\{-r,0\} }}\). A higher \(\tau\) means {{c2::steeper penalty for over-shooting (positive residual)}}, lower \(\tau\) means {{c3::steeper penalty for under-shooting}}.
Tags: ETH::Majors::Information_and_Data_Processing::IML::1._Regression::1._Loss_Functions
↑ Top