Basic Instrument for Experimental Probes in Machine Learning

Abstract

We propose an instrument based on a abstract experiment model that covers very different set-ups but still yields enough formal grip to allow useful services to be implemented in advance. The model demands the user to entirely decompose the experiments into named compounds. In return the services help with the presentation of results, parallelization on computing grids, or handling of the results containers. In contrast to other machine learning frameworks the tool does not include any algorithms or data-types. Instead it exspect the user to call foreign