Convolutional classifiers operate under the closed-set assumption: every input is taken to belong to a category seen during training. Deployed surveillance systems violate that assumption routinely, and a classifier confronted with an unfamiliar object assigns it to the nearest known category with high confidence, giving no indication that the classification is unwarranted. This paper evaluates an energy-based open-set detection layer built on a ResNet-50 backbone against three baselines, namely a conventional closed-set classifier, a softmax confidence threshold and OpenMax. The contribution is a comparison protocol in which the four detectors share a single trained checkpoint, a single calibration procedure restricted to the validation split, and a single decision rule, so that any measured difference is attributable to the detection layer alone. Twenty-four evaluation cells were produced from three training runs, spanning two regimes of unknown samples and three random seeds. With unknown samples drawn from held-out CIFAR-10 categories, OpenMax reached an AUROC of 0.845 against 0.815 for the energy detector, a significant difference under a paired test; the energy detector nevertheless reduced the false positive rate at 95 percent true positive rate by 0.062 relative to softmax thresholding while remaining indistinguishable from it on AUROC. With unknown samples drawn from TinyImageNet, all methods improved and no difference between the energy detector and OpenMax reached significance, the energy detector showing a dispersion across seeds four to five times larger. The largest effect measured was neither of these: changing the origin of the unknown samples raised the AUROC of the energy detector by 0.109 and lowered its false positive rate by 0.305, more than three times the span between the strongest and weakest detectors within either regime. Both regimes are built from public object-recognition corpora rather than from surveillance footage, so the results characterise the detection layers themselves; their transfer to operational surveillance imagery remains to be established. Keywords— energy-based detection: open-set recognition: out-of-distribution detection: reproducibility: ResNet-50: video surveillance
The exponential proliferation of digital content across personal repositories—encompassing academic manuscripts, annotated lecture transcripts, technical project documentation, and domain-specific study materials—has introduced substantial challenges in efficient knowledge retrieval. Conventional information retrieval paradigms, predominantly reliant on lexical or keyword-based matching, are fundamentally ill-equipped to resolve queries that demand semantic comprehension. This paper proposes a Retrieval-Augmented Conversational Framework (RACF) for intelligent personal document understanding, integrating multi-stage document ingestion, semantic chunking, dense vector embedding, and large language model (LLM)-driven conversational synthesis. The architecture transforms heterogeneous personal document collections into dynamic, queryable knowledge structures, enabling users to interact with their information through natural language dialogues rather than rigid query syntax. Empirical evaluations demonstrate that context-aware retrieval coupled with generative conversational response synthesis yields superior information accessibility, response coherence, and user satisfaction compared to traditional keyword-search baselines. The proposed framework provides a principled foundation for building next-generation personal knowledge management systems capable of transforming passive document archives into interactive, semantically rich conversational assistants. Keywords— Retrieval-Augmented Generation (RAG); Conversational AI; Personal Document Understanding; Semantic Embedding; Vector Databases; Natural Language Processing; Knowledge Management
Traditional manual screening of resumes is rapidly giving way to automated recruitment systems, yet the growing body of literature on the subject remains heavily fragmented across isolated technical, managerial, and ethical perspectives. To bridge these gaps, this systematic literature review synthesizes existing research to identify best technical architectures, examine responsible artificial intelligence considerations, and outline practical design recommendations. A systematic search was conducted across multiple digital repositories. Guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses framework, 69 primary studies were ultimately retained for detailed evaluation. The findings reveal a clear evolutionary shift away from rule-based parsers towards deep learning models, transformer-based architectures, generative large language models, and hybrid systems. However, while tokenization and named entity recognition remain the most foundational document-reading mechanisms, less than one-third of the literature explicitly operationalizes ethical fairness audits or explainable artificial intelligence metrics. This review concludes that while resume parsing technologies are highly mature algorithmically, the domain itself remains methodologically and ethically underdeveloped. This immaturity stems from severe fragmentation in dataset availability, a lack of standardized output formats, and an absence of clear accountability mechanisms. Ultimately, future research must shift towards establishing open-access, multilingual benchmark datasets and embedding transparent evaluation protocols directly within production-level applicant tracking systems. Keywords— artificial intelligence; automated recruitment systems; document reading techniques; named entity recognition; resume parsing; rule-based techniques.
Nonstationary time-series forecasting in critical applications needs an efficient accuracy level in the presence of concept drift; however, metaheuristics such as Multi-Population Quantum-inspired Particle Swarm Optimization (multiPQPSO) algorithm work in the form of a black box algorithm using static parameter settings. This work proposes a Landscape-Aware Explainable Multi-Population PSO (LAE-multiPQPSO) framework to tackle inefficient adaptation and lack of interpretability resulting from such non-transparent algorithms with static parameter settings. The proposed approach entails an online landscape analysis module, which calculates lightweight metrics of spatial ruggedness, normalized dimensional diversity, and convergence speed using swarm particle positions. These metrics adaptively tune hyperparameter limits (like fitness degradation limit, quantum contraction factor, and subswarm assignments) and generate rule-based explanations triggered via a JSON record structure. The proposed framework was evaluated using electric load, compute workload, epidemiological, and financial data sets, LAE-multiPQPSO outperforms benchmark models. The findings show that online landscape awareness gives dynamic parameter stability, fast concept drift adaptation, and clear explainability. Keywords— Concept Drift: Explainable Metaheuristics: Fitness Landscape Analysis: Particle Swarm Optimization: Time Series Forecasting
— This paper explores the theory of reinforcement learning as a cloud optimisation and allocation technique within a Zimbabwean telco-oriented multi-tenant environment. Conventional allocation algorithms and approaches, including heuristic, rule-based, and load-driven algorithms, are already in use in modern cloud implementations such as VMware and AWS. Still, they are often unable to handle dynamic, bursty traffic patterns. This research, therefore, examines the existing literature on reinforcement learning-based allocation techniques and compares them with known scheduling models. A multi-tenant cloud model is conceptualized to frame comparisons and outline the major factors, namely tenant variability, allocation goals, infrastructure, limitations, and decision-making behaviour. It compares the two families of algorithms at the theoretical level, including scalability, fairness, responsiveness, resource efficiency, and scale responsiveness to dynamic demand. The research adds to a polished perspective on the potential advantages and drawbacks of reinforcement learning in multifaceted cloud setups and offers a theoretical framework that identifies the circumstances in which RL methods could provide a better allocation functionality. The results will give researchers, cloud architects, and telecommunications stakeholders a baseline for future empirical, simulation-based, or implementation-driven assessments of intelligent cloud resource allocation strategies. Keywords—Cloud Resource Allocation: Conceptual Framework: Multi-Tenant Environments: Reinforcement Learning: SLA Performance: Traditional Algorithms
