Saad Shafiq

Syeda Tasnim Fabiha, Saad Shafiq, Wesley Klewerton Guez Assunção et al.

Service-based architecture (SBA) has gained attention in industry and academia as a means to modernize legacy systems. It refers to a design style that enables systems to be developed as suites of small, loosely coupled, and autonomous components (services) that encapsulate functionality and communicate via language-agnostic APIs. However, defining appropriately sized services that capture cohesive subsets of system functionality remains challenging. Existing work often relies on the availability of documentation, access to project personnel, or a priori knowledge of the target number of services, assumptions that do not hold in many real-world scenarios. Our work addresses these limitations using a deep reinforcement learning-based approach to identify appropriately sized services directly from implementation artifacts. We present Rake, a reinforcement learning-based technique that leverages available system documentation and source code to guide service decomposition at the level of implementation methods. Rake does not require specific documentation or access to project personnel and is language-agnostic. It also supports a customizable objective function that balances modularization quality and business capability alignment, i.e., the degree to which a service covers the targeted business capability. We applied Rake to four open-source legacy projects and compared it with two state-of-the-art techniques. On average, Rake achieved 7-14 percent higher modularization quality and 18-22 percent stronger business capability alignment. Our results further show that optimizing solely for business context can degrade decomposition quality in tightly coupled systems, highlighting the need for balanced objectives.

Abid Hassan, Tuan Ngo, Saad Shafiq et al.

Detecting out-of-distribution (OOD) inputs is a critical safeguard for deploying machine learning models in the real world. However, most post-hoc detection methods operate on penultimate feature representations derived from global average pooling (GAP) -- a lossy operation that discards valuable distributional statistics from activation maps prior to global average pooling. We contend that these overlooked statistics, particularly channel-wise variance and dominant (maximum) activations, are highly discriminative for OOD detection. We introduce DAVIS, a simple and broadly applicable post-hoc technique that enriches feature vectors by incorporating these crucial statistics, directly addressing the information loss from GAP. Extensive evaluations show DAVIS sets a new benchmark across diverse architectures, including ResNet, DenseNet, and EfficientNet. It achieves significant reductions in the false positive rate (FPR95), with improvements of 48.26\% on CIFAR-10 using ResNet-18, 38.13\% on CIFAR-100 using ResNet-34, and 26.83\% on ImageNet-1k benchmarks using MobileNet-v2. Our analysis reveals the underlying mechanism for this improvement, providing a principled basis for moving beyond the mean in OOD detection.

Abid Hassan, Tuan Ngo, Saad Shafiq et al.

Out-of-distribution (OOD) detection is critical for the safe deployment of deep neural networks. State-of-the-art post-hoc methods typically derive OOD scores from the output logits or penultimate feature vector obtained via global average pooling (GAP). We contend that this exclusive reliance on the logit or feature vector discards a rich, complementary signal: the raw channel-wise statistics of the pre-pooling feature map lost in GAP. In this paper, we introduce Catalyst, a post-hoc framework that exploits these under-explored signals. Catalyst computes an input-dependent scaling factor ($γ$) on-the-fly from these raw statistics (e.g., mean, standard deviation, and maximum activation). This $γ$ is then fused with the existing baseline score, multiplicatively modulating it -- an ``elastic scaling'' -- to push the ID and OOD distributions further apart. We demonstrate Catalyst is a generalizable framework: it seamlessly integrates with logit-based methods (e.g., Energy, ReAct, SCALE) and also provides a significant boost to distance-based detectors like KNN. As a result, Catalyst achieves substantial and consistent performance gains, reducing the average False Positive Rate by 32.87 on CIFAR-10 (ResNet-18), 27.94% on CIFAR-100 (ResNet-18), and 22.25% on ImageNet (ResNet-50). Our results highlight the untapped potential of pre-pooling statistics and demonstrate that Catalyst is complementary to existing OOD detection approaches.

Tuan Ngo, Abid Hassan, Saad Shafiq et al.

Deep Neural Networks (DNNs) tend to accrue technical debt and suffer from significant retraining costs when adapting to evolving requirements. Modularizing DNNs offers the promise of improving their reusability. Previous work has proposed techniques to decompose DNN models into modules both during and after training. However, these strategies yield several shortcomings, including significant weight overlaps and accuracy losses across modules, restricted focus on convolutional layers only, and added complexity and training time by introducing auxiliary masks to control modularity. In this work, we propose MODA, an activation-driven modular training approach. MODA promotes inherent modularity within a DNN model by directly regulating the activation outputs of its layers based on three modular objectives: intra-class affinity, inter-class dispersion, and compactness. MODA is evaluated using three well-known DNN models and five datasets with varying sizes. This evaluation indicates that, compared to the existing state-of-the-art, using MODA yields several advantages: (1) MODA accomplishes modularization with 22% less training time; (2) the resultant modules generated by MODA comprise up to 24x fewer weights and 37x less weight overlap while (3) preserving the original model's accuracy without additional fine-tuning; in module replacement scenarios, (4) MODA improves the accuracy of a target class by 12% on average while ensuring minimal impact on the accuracy of other classes.

Saad Shafiq, Atif Mashkoor, Christoph Mayr-Dorn et al.

In this paper, we propose a recommendation approach -- TaskAllocator -- in order to predict the assignment of incoming tasks to potential befitting roles. The proposed approach, identifying team roles rather than individual persons, allows project managers to perform better tasks allocation in case the individual developers are over-utilized or moved on to different roles/projects. We evaluated our approach on ten agile case study projects obtained from the Taiga.io repository. In order to determine the TaskAllocator's performance, we have conducted a benchmark study by comparing it with contemporary machine learning models. The applicability of the TaskAllocator was assessed through a plugin that can be integrated with JIRA and provides recommendations about suitable roles whenever a new task is added to the project. Lastly, the source code of the plugin and the dataset employed have been made public.

Saad Shafiq, Atif Mashkoor, Christoph Mayr-Dorn et al.

Context: The software development industry is rapidly adopting machine learning for transitioning modern day software systems towards highly intelligent and self-learning systems. However, the full potential of machine learning for improving the software engineering life cycle itself is yet to be discovered, i.e., up to what extent machine learning can help reducing the effort/complexity of software engineering and improving the quality of resulting software systems. To date, no comprehensive study exists that explores the current state-of-the-art on the adoption of machine learning across software engineering life cycle stages. Objective: This article addresses the aforementioned problem and aims to present a state-of-the-art on the growing number of uses of machine learning in software engineering. Method: We conduct a systematic mapping study on applications of machine learning to software engineering following the standard guidelines and principles of empirical software engineering. Results: This study introduces a machine learning for software engineering (MLSE) taxonomy classifying the state-of-the-art machine learning techniques according to their applicability to various software engineering life cycle stages. Overall, 227 articles were rigorously selected and analyzed as a result of this study. Conclusion: From the selected articles, we explore a variety of aspects that should be helpful to academics and practitioners alike in understanding the potential of adopting machine learning techniques during software engineering projects.