Example process - ML-Schema/core GitHub Wiki
Create a representation of a simple machine learning modelling process, i.e.
- Open algorithm implementation (e.g. weka.J48)
- Load dataset (e.g. Iris.arff)
- Build model (e.g. decision tree)
- Store model (e.g. a .model file)
You can do this from the perspective of your favourite machine learning environment, but don't overspecialize. In enough detail so that anyone else can repeat the same thing and has all relevant information. Mention the core concepts that you need. Optionally, also state their relations (useful for later).
OpenML Example
OpenML starts out by creating a task (e.g. Classification in dataset Iris):
http://www.openml.org/api_new/v1/task/59 (You need to log in on OpenML.org first.)
It also automatically gets a webpage: http://www.openml.org/t/59
The task contains the dataset, which is described as follows: http://www.openml.org/api_new/v1/data/61
Webpage: http://www.openml.org/d/61
You can then upload any algorithm, like this: http://www.openml.org/api_new/v1/flow/1720
Webpage: http://www.openml.org/f/1720
An example run (J48 on iris):
Webpage: http://www.openml.org/r/501579
XML description: http://openml.org/data/download/1745358/weka_generated_run7769371227014322202.xml
A run can also include the (instance-level) predictions:
http://openml.org/data/download/1745359/weka_generated_predictions5123198195447015004.arff
And the model (serialized and/or human-readable):
http://openml.org/data/download/1745360/WekaSerialized_weka.classifiers.trees.J487848219619191034547.model
http://openml.org/data/download/1745361/WekaModel_weka.classifiers.trees.J481306807653080981411.model
MEX Example
- iris-mex-output.ttl => mex file (metadata)
- iris-weka-output.txt => weka output
- iris.arff => dataset
- j48-iris.model => model file
ML Schema - RapidMiner Examples
ML-Schema-Example1 (SVM)
- Workflow
- Model
-
ModelRepresentation
-
Workflow representation in XML RapidMiner workflow in XML
Representation of ML-Schema-Example1 (SVM) using ML-Schema terms:
1. Level of Specification:
- //Samples/data/Golf->
mls:Data
- //Local Repository/processes/ML-Schema-Example1->
mls:Workflow
- Retrieve Golf ->
mls:Implementation
- Select Attributes ->
mls:Implementation
- SVM ->
mls:Implementation
- Store ->
mls:Implementation
2. Level of execution:
- SVM.model->
mls:Model
- golfModelSerialization->
mls:ModelRepresentation
DMOP Example
Example machine learning experiment annotated with DMOP (v5.3) terms
DMOP DM-Experiment Sandbox (Iris)
Example of representation with OntoDM terms
I plan to update this with a more detailed picture. The current one is just for the sake of a discussion Example representation with OntoDM terms