I want to…
…use a PSI service
Work through the examples on this site and refer to the PSI Architecture section of the specification for details of the JSON format of requests and responses. A summary of the REST API is also available. Each example below refers to relevant sections of the specification.
- Discover what resources are available at a PSI service
- Examine a relation (dataset) and access its contents using an attribute
- Query a relation
- Transform an attribute’s value
- Discover available learning algorithms
- Construct a learning task and train a predictor
- Examine a predictor and make predictions
- Update a predictor
- Adapt a predictor to a new relation (dataset)
- Learn and make predictions about images, video, etc.
…implement a PSI service
Refer to the specification for full details of the expected behaviour of a PSI service. If you only want to make available a isolated data source, learning algorithm or predictor then consider the following table, which indicates which parts of the specification must be implemented if providing a partial service.
|Service includes||Service must also have||Service should also have|
|many relations||attributes||relation collection, main entry point|
|1 learner||predictors, status resources||predictor collection, main entry point, learner collection (with one entry)|
|many learners||predictors, status resources||learner collection, predictor collection, main entry point|
|many predictors||predictor collection, main entry point|
|many transformers||transformers collection, main entry point|
|custom schema||schema collection||main entry point|