Using machine learning to infer constraints for product lines

Abstract

Variability intensive systems may include several thousand features allowing for an enormous number of possible configurations, including wrong ones (e.g. the derived product does not compile). For years, engineers have been using constraints to a priori restrict the space of possible configurations, i.e. to exclude configurations that would violate these constraints. The challenge is to find the set of constraints that would be both precise (allow all correct configurations) and complete (never allow a wrong configuration with respect to some oracle). In this paper, we propose the use of a machine learning approach to infer such product-line constraints from an oracle that is able to assess whether a given product is correct. We propose to randomly generate products from the product line, keeping for each of them its resolution model. Then we classify these products accord-ing to the oracle, and use their resolution models to infer crosstree constraints over the product-line. We validate our approach on a product-line video generator, using a simple computer vision algorithm as an oracle. We show that an interesting set of cross-tree constraint can be generated, with reasonable precision and recall.

Learning Contextual-Variability Models

“…We used a Lua video generator that synthesizes video variants together with their expected results (ground truths) [3]. In this way, we can measure the performance (precision and accuracy of the tracking of objects of interests, execution time, etc.)…”

“…Case Study. We have specialized an industrial video generator written in Lua [3]. Our learning approach allows one to constrain the generator and only build video variants of a certain quality (size, noise frequency, etc.).…”

“…A sample of 500 configurations was used and we obtained a classification accuracy of 80% (see Table 1). Constraints have proven to be readable: domain experts were able to validate and incorporate them into the variability model [3].…”

“…Our experience shows that modeling context and software variability (and their mutual relation) is labour-intensive, subjective and error-prone, even with extensive knowledge [3]. With our method, we aim to synthesize the constraints of Figure 1 as well as identifying new ones.…”

“…A configuration is composed of contextual features and software features. Numerous sampling strategies (e.g., random, t-wise) can be used to automatically select valid configurations [3,4]. The number of configurations N in the sample can be controlled by the user.…”