The rapid advancement of deep learning models that can generate and synthesis hyper-realistic videos known as Deepfakes and their ease of access to the general public have raised concern from all concerned bodies to their possible malicious intent use. Deep learning techniques can now generate faces, swap faces between two subjects in a video, alter facial expressions, change gender, and alter facial features, to list a few. These powerful video manipulation methods have potential use in many fields. However, they also pose a looming threat to everyone if used for harmful purposes such as identity theft, phishing, and scam. In this work, we propose a Convolutional Vision Transformer for the detection of Deepfakes. The Convolutional Vision Transformer has two components: Convolutional Neural Network (CNN) and Vision Transformer (ViT). The CNN extracts learnable features while the ViT takes in the learned features as input and categorizes them using an attention mechanism. We trained our model on the DeepFake Detection Challenge Dataset (DFDC) and have achieved 91.5 percent accuracy, an AUC value of 0.91, and a loss value of 0.32. Our contribution is that we have added a CNN module to the ViT architecture and have achieved a competitive result on the DFDC dataset.

Existing Scholarly Question Answering (QA) methods typically target homogeneous data sources, relying solely on either text or Knowledge Graphs (KGs). However, scholarly information often spans heterogeneous sources, necessitating the development of QA systems that integrate information from multiple heterogeneous data sources. To address this challenge, we introduce Hybrid-SQuAD (Hybrid Scholarly Question Answering Dataset), a novel large-scale QA dataset designed to facilitate answering questions incorporating both text and KG facts. The dataset consists of 10.5K question-answer pairs generated by a large language model, leveraging the KGs DBLP and SemOpenAlex alongside corresponding text from Wikipedia. In addition, we propose a RAG-based baseline hybrid QA model, achieving an exact match score of 69.65 on the Hybrid-SQuAD test set.

Large Language Models (LLMs) are transforming Natural Language Processing (NLP), but their benefits are largely absent for Africa's 2,000 low-resource languages. This paper comparatively analyzes African language coverage across six LLMs, eight Small Language Models (SLMs), and six Specialized SLMs (SSLMs). The evaluation covers language coverage, training sets, technical limitations, script problems, and language modelling roadmaps. The work identifies 42 supported African languages and 23 available public data sets, and it shows a big gap where four languages (Amharic, Swahili, Afrikaans, and Malagasy) are always treated while there is over 98\% of unsupported African languages. Moreover, the review shows that just Latin, Arabic, and Ge'ez scripts are identified while 20 active scripts are neglected. Some of the primary challenges are lack of data, tokenization biases, computational costs being very high, and evaluation issues. These issues demand language standardization, corpus development by the community, and effective adaptation methods for African languages.

Hate speech and abusive language are global phenomena that need socio-cultural background knowledge to be understood, identified, and moderated. However, in many regions of the Global South, there have been several documented occurrences of (1) absence of moderation and (2) censorship due to the reliance on keyword spotting out of context. Further, high-profile individuals have frequently been at the center of the moderation process, while large and targeted hate speech campaigns against minorities have been overlooked. These limitations are mainly due to the lack of high-quality data in the local languages and the failure to include local communities in the collection, annotation, and moderation processes. To address this issue, we present AfriHate: a multilingual collection of hate speech and abusive language datasets in 15 African languages. Each instance in AfriHate is annotated by native speakers familiar with the local culture. We report the challenges related to the construction of the datasets and present various classification baseline results with and without using LLMs. The datasets, individual annotations, and hate speech and offensive language lexicons are available on this https URL

Large language models (LLMs) have gained popularity recently due to their outstanding performance in various downstream Natural Language Processing (NLP) tasks. However, low-resource languages are still lagging behind current state-of-the-art (SOTA) developments in the field of NLP due to insufficient resources to train LLMs. Ethiopian languages exhibit remarkable linguistic diversity, encompassing a wide array of scripts, and are imbued with profound religious and cultural significance. This paper introduces EthioLLM -- multilingual large language models for five Ethiopian languages (Amharic, Ge'ez, Afan Oromo, Somali, and Tigrinya) and English, and Ethiobenchmark -- a new benchmark dataset for various downstream NLP tasks. We evaluate the performance of these models across five downstream NLP tasks. We open-source our multilingual language models, new benchmark datasets for various downstream tasks, and task-specific fine-tuned language models and discuss the performance of the models. Our dataset and models are available at the this https URL repository.

In a generalized Turán problem, we are given graphs $H$ and $F$ and seek to maximize the number of copies of $H$ in an $F$-free graph of order $n$. We consider generalized Turán problems where the host graph is planar. In particular we obtain the order of magnitude of the maximum number of copies of a fixed tree in a planar graph containing no even cycle of length at most $2\ell$, for all $\ell$, $\ell \geq 1$. We obtain the order of magnitude of the maximum number of cycles of a given length in a planar $C_4$-free graph. An exact result is given for the maximum number of $5$-cycles in a $C_4$-free planar graph. Multiple conjectures are also introduced.

An edge-colored graph is said to contain a rainbow-$F$ if it contains $F$ as a subgraph and every edge of $F$ is a distinct color. The problem of maximizing edges among $n$-vertex properly edge-colored graphs not containing a rainbow-$F$, known as the rainbow Turán problem, was initiated by Keevash, Mubayi, Sudakov and Verstraëte. We investigate a variation of this problem with the additional restriction that the graph is planar, and we denote the corresponding extremal number by $\ex_{\p}^*(n,F)$. In particular, we determine $\ex_{\p}^*(n,P_5)$, where $P_5$ denotes the $5$-vertex path.

An important part of data science is the use of visualisations to display data in a way that is easy to digest. Visualisations often rely on underlying statistical or machine learning models -- ranging from basic calculations like category means to advanced methods such as principal component analysis of multidimensional datasets -- to convey insights. We introduce an analogous concept for word descriptions of data, which we call wordalisations. Wordalisations describe data in easy to digest words, without necessarily reporting numerical values from the data. We show how to create wordalisations using large language models, through prompt templates engineered according to a task-agnostic structure which can be used to automatically generate prompts from data. We show how to produce reliable and engaging texts on three application areas: scouting football players, personality tests, and international survey data. Using the model cards framework, we emphasise the importance of clearly stating the model we are imposing on the data when creating the wordalisation, detailing how numerical values are translated into words, incorporating background information into prompts for the large language model, and documenting the limitations of the wordalisations. We argue that our model cards approach is a more appropriate framework for setting best practices in wordalisation of data than performance tests on benchmark datasets.

This paper presents a constraint-guided deep learning framework for developing physically consistent health indicators in bearing prognostics and health management. Conventional data-driven methods often lack physical plausibility, while physics-based models are limited by incomplete system knowledge. To address this, we integrate domain knowledge into deep learning using constraints to enforce monotonicity, bound output values between 1 and 0 (representing healthy to failed states), and ensure consistency between signal energy trends and health indicator estimates. This eliminates the need for complex loss term balancing. We implement constraint-guided gradient descent within an autoencoder architecture, creating a constrained autoencoder. However, the framework is adaptable to other architectures. Using time-frequency representations of accelerometer signals from the Pronostia dataset, our constrained model generates smoother, more reliable degradation profiles compared to conventional methods, aligning with expected physical behavior. Performance is assessed using three metrics: trendability, robustness, and consistency. Compared to a conventional baseline, the constrained model improves all three. Another baseline, incorporating monotonicity via a soft-ranking loss function, outperforms in trendability but falls short in robustness and consistency. An ablation study confirms that the monotonicity constraint enhances trendability, the boundary constraint ensures consistency, and the energy-health consistency constraint improves robustness. These findings highlight the effectiveness of constraint-guided deep learning in producing reliable, physically meaningful health indicators, offering a promising direction for future prognostic applications.

The electrification of road transport, as the predominant mode of transportation in Africa, represents a great opportunity to reduce greenhouse gas emissions and dependence on costly fuel imports. However, it introduces major challenges for local energy infrastructures, including the deployment of charging stations and the impact on often fragile electricity grids. Despite its importance, research on electric mobility planning in Africa remains limited, while existing planning tools rely on detailed local mobility data that is often unavailable, especially for privately owned passenger vehicles. In this study, we introduce a novel framework designed to support private vehicle electrification in data-scarce regions and apply it to Addis Ababa, simulating the mobility patterns and charging needs of 100,000 electric vehicles. Our analysis indicate that these vehicles generate a daily charging demand of approximately 350 MWh and emphasize the significant influence of the charging location on the spatial and temporal distribution of this demand. Notably, charging at public places can help smooth the charging demand throughout the day, mitigating peak charging loads on the electricity grid. We also estimate charging station requirements, finding that workplace charging requires approximately one charging point per three electric vehicles, while public charging requires only one per thirty. Finally, we demonstrate that photovoltaic energy can cover a substantial share of the charging needs, emphasizing the potential for renewable energy integration. This study lays the groundwork for electric mobility planning in Addis Ababa while offering a transferable framework for other African cities.

Currently, the wide spreading of real-time applications such as VoIP and videos-based applications require more data rates and reduced latency to ensure better quality of service (QoS). A well-designed traffic classification mechanism plays a major role for good QoS provision and network security verification. Port-based approaches and deep packet inspections (DPI) techniques have been used to classify and analyze network traffic flows. However, none of these methods can cope with the rapid growth of network traffic due to the increasing number of Internet users and the growth of real time applications. As a result, these methods lead to network congestion, resulting in packet loss, delay and inadequate QoS delivery. Recently, a deep learning approach has been explored to address the time-consumption and impracticality gaps of the above methods and maintain existing and future traffics of real-time applications. The aim of this research is then to design a dynamic traffic classifier that can detect elephant flows to prevent network congestion. Thus, we are motivated to provide efficient bandwidth and fast transmision requirements to many Internet users using SDN capability and the potential of Deep Learning. Specifically, DNN, CNN, LSTM and Deep autoencoder are used to build elephant detection models that achieve an average accuracy of 99.12%, 98.17%, and 98.78%, respectively. Deep autoencoder is also one of the promising algorithms that does not require human class labeler. It achieves an accuracy of 97.95% with a loss of 0.13 . Since the loss value is closer to zero, the performance of the model is good. Therefore, the study has a great importance to Internet service providers, Internet subscribers, as well as for future researchers in this area.

In a generalized Turán problem, we are given graphs $H$ and $F$ and seek to maximize the number of copies of $H$ in an $F$-free graph of order $n$. We consider generalized Turán problems where the host graph is planar. In particular we obtain the order of magnitude of the maximum number of copies of a fixed tree in a planar graph containing no even cycle of length at most $2\ell$, for all $\ell$, $\ell \geq 1$. We obtain the order of magnitude of the maximum number of cycles of a given length in a planar $C_4$-free graph. An exact result is given for the maximum number of $5$-cycles in a $C_4$-free planar graph. Multiple conjectures are also introduced.

The World Wide Web has come to be a great part of our daily life, yet user observed latency is still a problem that needs a proper means of handling. Even though earlier attempts focused on caching as the chief solution to tackling this issue, its success was extremely limited. Prefetching has come to be the primary technique in supplementing caching towards soothing the latency problem associated with the contemporary Internet. However, existing approaches in prefetching are extremely limited in their ability to employ application level web document relationship which is often visible only to the content developer. This is because most approaches are access history based schemes that make future users' access prediction only based on past user access. Attempts to incorporate prefetching schemes that utilize semantic information with those that use users past access history are extremely limited in their extensibility. In this work we present a novel framework that enables integration of schemes from both worlds of prefetching without the need for a major modification to the algorithms. When there is a need/possibility to capture new application level context, a new algorithm could be developed to do so and then it can be integrated into the framework. Since each participating scheme is merely viewed as an algorithm that produces a list of candidate objects that are likely to be accessed in the near future, the framework can entertain any one of the existing prefetching schemes. With its adaptive weight management technique the framework adjusts the effect of each algorithm in the overall prediction to parallel with its observed performance so far. We have found this formwork to be less aggressive than its contemporary counterparts which is extremely important for resource constrained mobile devices that have come to be the major means of access by users of the current web.

An edge-colored graph is said to contain a rainbow-$F$ if it contains $F$ as a subgraph and every edge of $F$ is a distinct color. The problem of maximizing edges among $n$-vertex properly edge-colored graphs not containing a rainbow-$F$, known as the rainbow Turán problem, was initiated by Keevash, Mubayi, Sudakov and Verstraëte. We investigate a variation of this problem with the additional restriction that the graph is planar, and we denote the corresponding extremal number by $\ex_{\p}^*(n,F)$. In particular, we determine $\ex_{\p}^*(n,P_5)$, where $P_5$ denotes the $5$-vertex path.

We compute the electromagnetic fields generated in relativistic heavy-ion collisions using the iEBE-VISHNU framework. We calculated the incremental drift velocity from the possible four sources of the electric force (coulomb, Lorentz, Faraday, and Plasma-based) on the particles created. The effect of this external electromagnetic field on the flow harmonics of particles was investigated, and we found out that the flow harmonics values get suppressed and rouse in a non-uniform fashion throughout the evolution. More precisely, a maximum of close to three percent increase in elliptic flow was observed. We also found mass more dominant factor than charges for the change in flow harmonics due to the created electromagnetic field. On the top of that, the magnetic field perpendicular to the reaction plane is found to be sizable while the different radial electric forces were found to cancel out each other. Finally, we found out that the inclusion of an electromagnetic field affects the flow of particles by suppressing or rising it in a non-uniform fashion throughout the evolution.

Machine learning has proved useful in many software disciplines, including computer vision, speech and audio processing, natural language processing, robotics and some other fields. However, its applicability has been significantly hampered due its black-box nature and significant resource consumption. Performance is achieved at the expense of enormous computational resource and usually compromising the robustness and trustworthiness of the model. Recent researches have been identifying a lack of interactivity as the prime source of these machine learning problems. Consequently, interactive machine learning (iML) has acquired increased attention of researchers on account of its human-in-the-loop modality and relatively efficient resource utilization. Thereby, a state-of-the-art review of interactive machine learning plays a vital role in easing the effort toward building human-centred models. In this paper, we provide a comprehensive analysis of the state-of-the-art of iML. We analyze salient research works using merit-oriented and application/task oriented mixed taxonomy. We use a bottom-up clustering approach to generate a taxonomy of iML research works. Research works on adversarial black-box attacks and corresponding iML based defense system, exploratory machine learning, resource constrained learning, and iML performance evaluation are analyzed under their corresponding theme in our merit-oriented taxonomy. We have further classified these research works into technical and sectoral categories. Finally, research opportunities that we believe are inspiring for future work in iML are discussed thoroughly.

In this paper, we study composition operators on Hilbert space of complex-valued harmonic functions. In particular, we explore isometries, the type of self-map that generate bounded composition operator, and characterize the boundedness of composition operator in terms of Poisson integral. Furthermore, we establish the relation between reproducing kernels and composition operators on Hilbert space of complex-valued harmonic functions.

We investigate the period changes of 13 short-period Type II Cepheids using the O-C method over a century-long data baseline. The O-C diagrams for these stars exhibit a parabolic trend, indicating both increasing and decreasing period changes over time. These observed period changes are consistent with recent theoretical models based on horizontal branch evolutionary models for short-period BL Her stars. The pulsation stability test proposed by Lombard and Koen confirms that the period changes are signficant, indicating evolutionary shifts. We identify seven BL Her stars with decreasing periods, expanding the existing sample size of short-period Type II Cepheids. This contributes to a deeper understanding of stellar evolution and the processes governing low-mass stars.

Model-based geostatistics (MBG) is a subfield of spatial statistics focused on predicting spatially continuous phenomena using data collected at discrete locations. Geostatistical models often rely on the assumptions of stationarity and isotropy for practical and conceptual simplicity. However, an alternative perspective involves considering non-stationarity, where statistical characteristics vary across the study area. While previous work has explored non-stationary processes, particularly those leveraging covariate information to address non-stationarity, this research expands upon these concepts by incorporating multiple covariates and proposing different ways for constructing non-stationary processes. Through a simulation study, the significance of selecting the appropriate non-stationary process is demonstrated. The proposed approach is then applied to analyse malaria prevalence data in Mozambique, showcasing its practical utility