Mixer filter

From Piki

The mixer filter repartitions the training and validation sets of the data unit.

Mixer
Name	Mixer
Deployable	No
Static	Yes/No
Sample modifier	Yes
Feature modifier	No

Contents 1 Usage 2 Settings 3 General advice 4 See also

Usage

The mixer filter is used to set aside a certain percentage of data for validation or to move validation data back into the training set.

Settings

The settings can be modified using the settings browser.

Mixer filter settings

(Mode) Mode: Mix Mode: Shuffle - Samples at random positions are moved from training to validation. Do not use this with time series data. Continuous - Samples at the end of the training set are moved to the validation set. Split: Split mode: Percentage - Makes the validation set a percentage of the total data. FixedTraining - Makes the training set a fixed number of samples large and moves the rest into the validation set. FixedValidation - Makes the validation set a fixed number of samples large and moves the rest into the training set. (If Percentage) Percent: The percentage of total data as validation. Must be between 0 and 100. (If Fixed) Fixed: The number of samples in the training or validation set. Seed: Random number initialization used for Shuffle mode. If the same seed is entered, the same samples will be moved.

General advice

• When using a static system (i.e. not time series), it may be a good idea to shuffle the data using this filter, making sure that the validation set has a representative data distributions. The input formats in order not to destroy time series data generally just take the last part of the data and put it in the validation channel. This may cause problems if the data is sorted in some way.

See also

• Filter - Article covering general filter principles.
• List of Filter components - List of all available filters.
• Data unit - Article explaining data units in detail.
• Data - Article explaining essential data concepts.