Logarithmic scoring rule - Revision history

Vipul: /* Relation with logarithmic loss functions in logistic regression */

2017-09-10T15:18:16Z

Relation with logarithmic loss functions in logistic regression

← Older revision		Revision as of 15:18, 10 September 2017
Line 15:		Line 15:
	==Relation with logarithmic loss functions in logistic regression==		==Relation with logarithmic loss functions in logistic regression==

	The logarithmic loss function used in [[~~wikipedia~~:logistic regression\|logistic regression]] relies on the logarithmic scoring rule, but with the following twist: the probabilities are not fixed numbers, but themselves depend on parameters that vary with the random instances that we are presented.		The logarithmic loss function used in [[machinelearning:logistic regression\|logistic regression]] relies on the logarithmic scoring rule, but with the following twist: the probabilities are not fixed numbers, but themselves depend on parameters that vary with the random instances that we are presented.

Vipul at 16:11, 31 May 2014

2014-05-31T16:11:57Z

← Older revision		Revision as of 16:11, 31 May 2014
Line 1:		Line 1:
	==Definition==		==Definition==

	The '''logarithmic scoring rule''' is a scoring rule used to measure how well a given assignment of probabilities to values of a random variable performs on some real-world instances of the random variable. Explicitly, consider a random variable <math>X</math> that can take <math>n</math> distinct values <math>1,2,\dots,n</math>. Suppose we estimate probabilities <math>p_1,p_2,\dots,p_n</math> for these values respectively (with <math>\sum_{i=1}^n p_i = 1</math>. The logarithmic scoring rule works as follows: for every instance of the random variable <math>X</math>, we assign score equal to the negative of the [[logarithm]] of the corresponding probability <math>p_i</math>. Explicitly, if the instances are <math>X_1,X_2,\dots,X_m</math>, the total score is:		The '''logarithmic scoring rule''' is a scoring rule used to measure how well a given assignment of probabilities to values of a random variable performs on some real-world instances of the random variable. Explicitly, consider a random variable <math>X</math> that can take <math>n</math> distinct values <math>1,2,\dots,n</math>. Suppose we estimate probabilities <math>p_1,p_2,\dots,p_n</math> for these values respectively (with <math>p_i \in [0,1]</math>, <math>\sum_{i=1}^n p_i = 1</math>). The logarithmic scoring rule works as follows: for every instance of the random variable <math>X</math>, we assign score equal to the negative of the [[logarithm]] of the corresponding probability <math>p_i</math>. Explicitly, if the instances are <math>X_1,X_2,\dots,X_m</math>, the total score is:

	<math>\sum_{j=1}^m -\ln(p_{X_j})</math>		<math>\sum_{j=1}^m -\ln(p_{X_j})</math>

Vipul at 16:10, 31 May 2014

2014-05-31T16:10:47Z

← Older revision		Revision as of 16:10, 31 May 2014
Line 1:		Line 1:
	==Definition==		==Definition==

	The '''logarithmic scoring rule''' is a scoring rule used to measure how well a given assignment of probabilities to values of a random variable performs on some real-world instances of the random variable. Explicitly, consider a random variable <math>X</math> that can take <math>n</math> distinct values <math>1,2,\dots,n</math>. Suppose we estimate probabilities <math>p_1,p_2,\dots,p_n</math> for these values respectively. The logarithmic scoring rule works as follows: for every instance of the random variable <math>X</math>, we assign score equal to the negative of the [[logarithm]] of the corresponding probability <math>p_i</math>. Explicitly, if the instances are <math>X_1,X_2,\dots,X_m</math>, the total score is:		The '''logarithmic scoring rule''' is a scoring rule used to measure how well a given assignment of probabilities to values of a random variable performs on some real-world instances of the random variable. Explicitly, consider a random variable <math>X</math> that can take <math>n</math> distinct values <math>1,2,\dots,n</math>. Suppose we estimate probabilities <math>p_1,p_2,\dots,p_n</math> for these values respectively (with <math>\sum_{i=1}^n p_i = 1</math>. The logarithmic scoring rule works as follows: for every instance of the random variable <math>X</math>, we assign score equal to the negative of the [[logarithm]] of the corresponding probability <math>p_i</math>. Explicitly, if the instances are <math>X_1,X_2,\dots,X_m</math>, the total score is:

	<math>\sum_{j=1}^m -\ln(p_{X_j})</math>		<math>\sum_{j=1}^m -\ln(p_{X_j})</math>

Vipul: Created page with "==Definition== The '''logarithmic scoring rule''' is a scoring rule used to measure how well a given assignment of probabilities to values of a random variable performs on so..."

2014-05-31T16:07:05Z

Created page with "==Definition== The '''logarithmic scoring rule''' is a scoring rule used to measure how well a given assignment of probabilities to values of a random variable performs on so..."

New page

==Definition==

The '''logarithmic scoring rule''' is a scoring rule used to measure how well a given assignment of probabilities to values of a random variable performs on some real-world instances of the random variable. Explicitly, consider a random variable <math>X</math> that can take <math>n</math> distinct values <math>1,2,\dots,n</math>. Suppose we estimate probabilities <math>p_1,p_2,\dots,p_n</math> for these values respectively. The logarithmic scoring rule works as follows: for every instance of the random variable <math>X</math>, we assign score equal to the negative of the [[logarithm]] of the corresponding probability <math>p_i</math>. Explicitly, if the instances are <math>X_1,X_2,\dots,X_m</math>, the total score is:

<math>\sum_{j=1}^m -\ln(p_{X_j})</math>

Note that the base of the logarithms does not matter: it could be any fixed number greater than 1.

The smaller the value of the score with the logarithmic scoring rule, the better the assignment of probabilities has performed according to the rule.

==Facts==

* [[Logarithmic scoring rule is proper]]: In expectation, if the actual probabilities of the values are <math>q_1,q_2,\dots,q_n</math> respectively, then the choice of values that minimizes our expected score is <math>p_1 = q_1, p_2 = q_2, \dots, p_n = q_n</math>.

==Relation with logarithmic loss functions in logistic regression==

The logarithmic loss function used in [[wikipedia:logistic regression|logistic regression]] relies on the logarithmic scoring rule, but with the following twist: the probabilities are not fixed numbers, but themselves depend on parameters that vary with the random instances that we are presented.