Algorithm

An Algorithm describes a machine learning algorithm and its hyperparameters. This resource is used internally by the Modela data plane to sample algorithms relevant to the current task type (i.e. classification, regression, etc) that will be trained on the cluster.

Algorithms can be accessed by selecting the Algorithms section when viewing the details of an ML Framework. All Algorithms are included as part of the Modela Helm Charts and should not be changed by the user.

Examples

Random Forest Regressor, SGD Classifier, Logistic Regression Classifier

Relationships to Other Resources

An Algorithm is part of a MLFramework

Algorithm API Reference

Algorithm define the metadata about a machine learning algorithm.

Name	Type	Description	Required
apiVersion	string	catalog.modela.ai/v1alpha1	true
kind	string	Algorithm	true
metadata	object	Refer to the Kubernetes API documentation for the fields of the `metadata` field.	true
spec	object	AlgorithmSpec describes the attributes of an machine learning algorithm	true

Algorithm.spec

^{↩ Parent}

AlgorithmSpec describes the attributes of an machine learning algorithm

Name	Type	Description	Required
categoricalParameters	[]object	CategoricalParameters is the list of categorical hyper parameters ranges	false
description	string	Description is the description of the algorithm Default:	false
floatParameters	[]object	FloatParameters is the list of float hyper parameters ranges	false
frameworkName	string	FrameworkName is a reference to library	false
integerParameters	[]object	IntegerParameters is the list of integer hyper parameter ranges.	false
sparse	boolean	Sparse indicate if this algorithm can deal with sparse data Default: false	false
task	enum	Task is the machine learning task (binary classification / regression) Enum: binary-classification, multi-classification, multi-label-classification, forecasting, regression, clustering, recommendation, outlier-detection, novelty-detection, topic-modeling, video-action-recognition, video-classification, video-object-tracking, image-classification, image-multi-classification, image-object-detection, image-segmentation, auto, text-ner, text-classification, text-summarization, text-qa, text-sentiment-analysis, text-generation, text-code-generation, text-translation, text-lang-detection, text-grammer-correction, text-paraphrasing, text-intent-classification, text-semantic-similarity, text-keyword-extraction, text-pos, text-tokenization, text-lemma, unknown	false
url	string	URL is a url to the html page describing the algorithm Default:	false

Algorithm.spec.categoricalParameters[index]

^{↩ Parent}

CategoricalParameter contain the specification of an categorical hyper parameter

Name	Type	Description	Required
defaultValue	string	DefaultValue if the default categorical value	false
enums	[]string	Enums defines the list of values	false
name	string	Name is the name of the hyper parameter	false

Algorithm.spec.floatParameters[index]

^{↩ Parent}

FloatParameter contain the specification of an float hyper parameter

Name	Type	Description	Required
defaultValue	number	Default float value	false
log	boolean	Log specify if this hyper paramer is logartimic	false
max	number	Max is the maximum value	false
min	number	Min is the minimum value	false
name	string	Name is the name of the hyper parameter	false

Algorithm.spec.integerParameters[index]

^{↩ Parent}

IntParameter contain the specification of an integer hyper parameter

Name	Type	Description	Required
defaultValue	integer	DefaultValue is the default value of the parameter Format: int32	false
max	integer	Max is the maximum value of the hyper parameter range Format: int32	false
min	integer	Min is the minimum value of the hyper parameter range Format: int32	false
name	string	Name is the name of the hyper parameter	false
type	enum	Type is the datatype of the parameter Enum: boolean, datetime, number, categorical, ordinal, text, json, number-list, categorical-list, text-list	false