Search

Users interact with modern applications and devices through graphical user interfaces (GUIs). To ensure intuitive and easy usability, the GUIs need to be tested, where developers aim at finding possible bugs and inconsistent functionality. Manual GUI testing requires time and effort, and thus, its efficiency can be improved with automation. Conventional automation tools for GUI testing reduce the burden of manual testing but also introduce challenges in the maintenance of test cases. In order to overcome these issues, we propose a deep-reinforcement-learning-based (DRL) solution for automated and adaptive GUI testing. Specifically, we propose and evaluate the performance of an image-based DRL solution. We adapt the asynchronous advantage actor-critic (A3C) algorithm to GUI testing inspired by how a human uses a GUI. We feed screenshots of the GUI as the input and let the algorithm decide how to interact with GUI components. We observe that our solution can achieve up to six times higher exploration efficiency compared to selected baseline algorithms. Moreover, our solution is more efficient than inexperienced human users and almost as efficient as an experienced human user in our experimental GUI testing scenario. For these reasons, image-based DRL exploration can be considered as a viable GUI testing method.

We benchmark Q-learning methods, with various action selection strategies, in intelligent orchestration of the network edge. Q-learning is a reinforcement learning technique that aims to find optimal action policies by taking advantage of the experiences in the past without utilizing a model that describes the dynamics of the environment. With experiences, we refer to the observed causality between the action and the corresponding impact to the environment. In this paper, the environment for Q-learning is composed of virtualized networking resources along with their dynamics that are monitored with Spindump, an in-network latency measurement tool with support for QUIC and TCP. We optimize the orchestration of these networking resources by introducing Q-learning as part of the machine learning driven, intelligent orchestration that is applicable in the edge. Based on the benchmarking results, we identify which action selection strategies support network orchestration that provides low latency and packet loss by considering network resource allocation in the edge.

Machine learning requires a relevant amount of computational resources and it is usually executed in high-capacity centralized cloud infrastructures (e.g., data centers). In such infrastructures, resources are shared in a scalable manner through instantiation and orchestration of multiple virtualized services. Emerging trends in machine learning are distribution and parallelization of model training, which allows the execution of model training tasks in multiple distributed computational domains, with the aim of reducing the overall training time. A possible drawback in decentralization of machine learning is that performance latency issues may arise when the computation of training is geographically distributed to nodes with long distance from each other. One way to reduce latency is to utilize edge computing infrastructure, i.e., to distribute computation near the origin of the request. As edge resources can be scarce, it is important to orchestrate the model training in a parallelized manner. To this extent, in order to effectively ease the use of parallelization both in centralized and in distributed scenarios, we propose and implement a concept that we refer to Intelligent Agent (IA). An IA is responsible for instantiating and scheduling of the machine learning tasks (e.g., model training), and deriving inferences. In our solution, model training is distributed to multiple IAs in parallel. Each IA is packaged into a Linux container in order to take advantage of container portability across heterogenous deployments and to reuse existing container orchestration tools. We validate our proposal by deploying and instantiating multiple IAs across a distributed cloud environment, where each IA is accounting for a fixed amount of computational resources. Keywords - Intelligent agent, Model parallelism, Regression training, Intelligent cloud