Home
Reference Guides
Explore the Extended Definitions, OML Guides, Block Library, API Guide and Glossary.
Reference Guide for OpenMatrix Language Functions
The Reference Guide contains documentation for all functions supported in the OpenMatrix language.
Machine Learning Commands
Evaluation Metrics
logloss
It measures the performance of a classification model where the output of the classification model is in terms of probability. It takes into account the uncertainty of the predictions based on how much it varies from the actual label. Best Value: 0 Worst Value: tends to infinity (it increases as the predicted probability value diverges from actual label).

Welcome
What's New
Discover new features and enhancements.
Get Started
New to Twin Activate? Learn the basics here.
Tutorials
Start using Twin Activate with our interactive tutorials.
User Guide
Learn about the features and functionality available in Twin Activate.
Reference Guides
Explore the Extended Definitions, OML Guides, Block Library, API Guide and Glossary.
- Extended Definitions for Advanced Users
  PDF file with in-depth information on key topics in the User's Guide.
- Scripting Guide for OpenMatrix Language
  OpenMatrix is a mathematical scripting language.
- Reference Guide for OpenMatrix Language Functions
  The Reference Guide contains documentation for all functions supported in the OpenMatrix language.
  - ABF Commands
  - Altair Integration Commands
  - Audio Commands
  - CAE Commands
  - CAE Reader Commands
  - CAE Reader 3D Commands
  - CAE Writer Commands
  - Calculus Commands
  - Computer Vision Commands
  - Communication Commands
  - Control System Commands
  - Core Minimal Interpreter Commands
  - Data Structure Commands
  - Differential Equations Commands
  - Elementary Math Commands
  - File I/O Commands
  - Geometry Commands
  - GUI Commands
  - HDF5 Reader Commands
  - HDF5 Writer Commands
  - hstpyfit Commands
  - Image Commands
  - JSON Commands
  - Library Manager Commands
  - Linear Algebra Commands
  - Logical Commands
  - Machine Learning Commands
    - Apriori
    - Association Rules
    - Classification
    - Clustering
    - Decomposition
    - Distance
    - Evaluation Metrics
      - accuracy
        Measures the over all performance of the classification model in terms of fraction of predictions made by the classifier that are correct. Interpretation: Accuracy is the fraction of number of correct predictions out of all examples, where the best value is 1 and the worst is 0.
      - classificationerror
        Measures the overall error rate of the classification model in terms of fraction of predictions made by the classifier that are not correct. Interpretation: Classification Error Rate is the fraction of number of in correct predictions out of all examples, where the best value is 1 and the worst is 0.
      - confusionmatrix
        Computes confusion matrix for classifier's predictions. The row represents the actual label and the column represents the predicted label. Each cell represents how many examples are actually actual label but predicted as predicted label, where the best value: except left to right diagonal, all are zero, and worst value: left to right diagonal are zero.
      - f1score
        It measures the performance of a classification model in terms of the both precision and recall. F1 score can be interpreted (loosely) as the average of precision and recall, where the best value is 1 and the worst is 0.
      - logloss
        It measures the performance of a classification model where the output of the classification model is in terms of probability. It takes into account the uncertainty of the predictions based on how much it varies from the actual label. Best Value: 0 Worst Value: tends to infinity (it increases as the predicted probability value diverges from actual label).
      - precision
        It measures the performance of a classification model in terms of the classifier’s ability to not label a negative example as positive. Precision score can be interpreted as the probability that a positive prediction made by the classifier is positive, where the best value is 1 and the worst is 0.
      - recall
        It measures the performance of a classification model in terms of the classifier's ability to predict positive examples correctly out of all positive examples. Recall score can be interpreted as the probability that a randomly selected positive example is correctly identified by the classifier, where the best value is 1 and the worst is 0.
      - explainedvariance
        Computes Explained Variance Regression Score from the actual and predicted outputs. It represents the amount of variance explained by the model with respect to variance of the actual output, where the best value is 1 and the worst is 0.
      - hmse
        Computes Half Mean Squared Error between two vectors. It is half of the average error made by the regression model, where the best value is 0 and the worst value tends to infinity (it increases as the deviation between predicted and actual value increases).
      - mae
        Computes Mean Absolute Error between two vectors. It is the average of the error (absolute error) made by the regression model, where the best value is 0 and the worst value tends to infinity (it increases as the deviation between predicted and actual value increases).
      - maxerror
        Computes Maximum Absolute Deviation between two vectors. It is the maximum of the error made by the regression model, where the best value is 0 and the worst value tends to infinity (it increases as the deviation between predicted and actual value increases).
      - mse
        Computes Mean Squared Error between two vectors. It is the average of the error (squared to get rid of negative sign) made by the regression model, where the best value is 0 and the worst value tends to infinity (it increases as the deviation between predicted and actual value increases).
      - r2
        Computes the Coefficient of Determination, which is the proportion of variance in the dependent variable that is predictable from the independent variable(s), where the best value is 1 and the worst value goes towards -Infinity.
      - rmse
        Computes Root Mean Squared Error between two vectors. It is the average of the amount of mistakes made by regression model in prediction, where the best value is 0 and the worst value tends to infinity (it increases as the deviation between predicted and actual value increases).
      - rss
        Computes Residual Sum of Squares between two vectors. It is the sum of the error (squared to get rid of negative sign) made by the regression model, where the best value is 0 and the worst value tends to infinity (it increases as the deviation between predicted and actual value increases).
    - K-Mode Clustering
    - Model Selection
    - Preprocessing
    - Regression
  - Optimization Commands
  - Plotting Commands
  - Polynomial Math Commands
  - Signal Processing Commands
  - Statistical Analysis Commands
  - String Operations Commands
  - System Commands
  - Table Commands
  - Time Commands
  - Trigonometry Commands
  - OML to Python Bridge Commands
  - Python to OML Bridge Commands
- Block Reference Guide
  Describes all of the blocks in the installed Twin Activate library.
- Function Guide for Twin Activate
  Provides information about developing and simulating models through the Twin Activate Application Program Interface.
- Glossary
  Key terms associated with the software.
Frequently Asked Questions
You've got questions? We've got answers!

logloss

It measures the performance of a classification model where the output of the classification model is in terms of probability. It takes into account the uncertainty of the predictions based on how much it varies from the actual label. Best Value: 0 Worst Value: tends to infinity (it increases as the predicted probability value diverges from actual label).

Attention: Available only with Twin Activate commercial edition.

Syntax

Loss = logloss(targets,predictions)

Inputs

targets: Actual label for each observation.; Type: double; Dimension: vector
predictions: Predicted value for each observation.; Type: double; Dimension: vector

Outputs

Loss: Log loss of the classifier.; Type: double; Dimension: scalar

Example

Usage of logloss

targets = [1, 0];
predictions = [0, 1];
score = logloss(targets, predictions);

> score
score = 24.1771435

Comments

When predicted probability becomes 0 or 1, it leads to error or infinity in OML, because of the computation of log(0). So, min and max functions are applied to solve this problem. When the predicted probability is 0, it is replaced with 1e^(-15) and when the predicted probability is 1, it is replaced with 1-1e^(-6).