ORANGE : A visual programming platform

Orateur : Janez DEMSAR (University of Ljubljana, Slovenia)

Orange (http://orange.biolab.si/) is a data mining platform, which allows non-programmers to write data processing and analysis programs visually. The "program" is presented as a collection of interconnected functional units called "widgets". There are widgets for loading the data, perform various operations (preprocessing, filtering, analysis), printing and visualizing the data in various ways, and constructing machine learning and statistical models. Each widget can have one or more inputs and outputs. To "program" a schema, we put the widgets on the Canvas and connect them as we wish. Orange is fully integrated with Python, and programming Orange widget and adapting user-codes and Python scripts is very easy. In this talk, we will present Orange from two perspectives. First, we will briefly show its capabilities as a data mining tool, with the purpose of demonstrating the versatility of this approach. Second - and more interesting for the ESRF - we will focus on the concept of data-flow-based programming : by replacing the existing set of widgets by new ones related to another field of science, we can easily use Orange as infrastructure for a specific simulation or data analysis process. As an example, a data-flow schema based on Orange Canvas is well adapted for optics simulation. We are preparing a visual tool to perform simulations of Optics instruments, like a synchrotron beamline. A set of widgets that match the different components (optics, monitors, detectors) are used to construct the instrument and define their parameters. The system is simulated via the Python API of the SHADOW ray tracing code.

Télécharger dans mon agenda