File formats

There are four file formats specific to the ShapeKit Suite. This chapter provides information about them.

Xml Set

Those files contains data set stored in Xml. Data sets are used for training a model or for testing a model (so-called validation set). A data set is characterized by a number of inputs, a number of outputs, and a number of experiments. For each experiment, there is a input vector and an output (or result) vector, with consistence sizes. There are at least one input and one output. The file also stores a name for the set, and names for each input and output.

Binary Set

This contains exactly the same thing as the Xml Set, but in a compact form. It is stored using a more efficient way (“binary”), and then compressed.

Using the ShapeKit Loader, it is possible to convert between Xml Set and Binary Set, in both directions.

Sketch

This file records work of progress in ShapeKit Studio. This contains the training data set and all parts of the design already added/created. The design can be incomplete.

Design

This is the result of the data modeling process. The Design files contains all that is needed to compute outputs from inputs. This includes preprocessing steps/parameters, models, and links between all of them.

The training data set is NOT included in this file. As such it is usually very small (typically a few dozens of kilobytes). This is important because most of our customers have very sensitive data and very strict rules on how those data can/may be shared.

Summary

The following table summarizes the most important properties of file formats handled by the ShapeKit Suite:

Format	Xml Set	Binary Set	Sketch	Design
Extension	*.XmlSet	*.BinarySet	*.Sketch	*.Design
Read/write by scripts	Yes	No	No	No
Read/write by human	Yes	No	No	No
Compressed	No	Yes	Yes	Yes
Contains the training DataSet	Yes	Yes	Yes	No

This table shows interaction between programs from the ShapeKit Suite and file formats:

Format	Xml Set	Binary Set	Sketch	Design
ShapeKit Loader	Read/Write	Read/Write	None	None
ShapeKit Studio	Read	Read	Read/Write	Write
ShapeKit Exploiter	Read(for tests)	Read(for tests)	None	Read