Abstract Anomaly detection in high dimensional data is becoming a fundamental research problem that has various applications the real world. However, many existing anomaly techniques fail to retain sufficient accuracy due so-called “big data” characterised by high-volume, and high-velocity generated variety of sources. This phenomenon having both problems together can be referred “curse big dimensionality,” affect terms performance accuracy. To address this gap understand core problem, it...

Software engineers are increasingly adding semantic search capabilities to applications using a strategy known as Retrieval Augmented Generation (RAG). A RAG system involves finding documents that semantically match query and then passing the large language model (LLM) such ChatGPT extract right answer an LLM. systems aim to: a) reduce problem of hallucinated responses from LLMs, b) link sources/references generated responses, c) remove need for annotating with meta-data. However, suffer...

The Hierarchical Navigable Small World (HNSW) algorithm is widely used for approximate nearest neighbor (ANN) search, leveraging the principles of navigable small-world graphs. However, it faces some limitations. first local optima problem, which arises from algorithm's greedy search strategy, selecting neighbors based solely on proximity at each step. This often leads to cluster disconnections. second limitation that HNSW frequently fails achieve logarithmic complexity, particularly in...

Object Tracking is essential for many computer vision applications, such as autonomous navigation, surveillance, and robotics. Unlike Passive (POT), which relies on static camera viewpoints to detect track objects across consecutive frames, Active (AOT) requires a controller agent actively adjust its viewpoint maintain visual contact with moving target in complex environments. Existing AOT solutions are predominantly single-agent-based, struggle dynamic scenarios due limited information...

The widespread use of Multi-layer perceptrons (MLPs) often relies on a fixed activation function (e.g., ReLU, Sigmoid, Tanh) for all nodes within the hidden layers. While effective in many scenarios, this uniformity may limit networks ability to capture complex data patterns. We argue that employing same at every node is suboptimal and propose leveraging different functions each increase flexibility adaptability. To achieve this, we introduce Local Control Networks (LCNs), which leverage...

Decentralized Multi-Agent Reinforcement Learning (Dec-MARL) has emerged as a pivotal approach for addressing complex tasks in dynamic environments. Existing (MARL) methodologies typically assume shared objective among agents and rely on centralized control. However, many real-world scenarios feature with individual goals limited observability of other agents, complicating coordination hindering adaptability. Dec-MARL strategies prioritize either communication or coordination, lacking an...

Next-frame prediction in videos is crucial for applications such as autonomous driving, object tracking, and motion prediction. The primary challenge next-frame lies effectively capturing processing both spatial temporal information from previous video sequences. transformer architecture, known its prowess handling sequence data, has made remarkable progress this domain. However, transformer-based models face notable issues: (a) multi-head self-attention (MHSA) mechanism requires the input...

Neural Architecture Search (NAS) aims to automate the design of deep neural networks. However, existing NAS techniques often focus on maximising accuracy, neglecting model efficiency. This limitation restricts their use in resource-constrained environments like mobile devices and edge computing systems. Moreover, current evaluation metrics prioritise performance over efficiency, lacking a balanced approach for assessing architectures suitable constrained scenarios. To address these...

Software engineers are increasingly adding semantic search capabilities to applications using a strategy known as Retrieval Augmented Generation (RAG). A RAG system involves finding documents that semantically match query and then passing the large language model (LLM) such ChatGPT extract right answer an LLM. systems aim to: a) reduce problem of hallucinated responses from LLMs, b) link sources/references generated responses, c) remove need for annotating with meta-data. However, suffer...

Large language models (LLMs) enable state-of-the-art semantic capabilities to be added software systems such as search of unstructured documents and text generation. However, these are computationally expensive. At scale, the cost serving thousands users increases massively affecting also user experience. To address this problem, caches used check for answers similar queries (that may have been phrased differently) without hitting LLM service. Due nature cache techniques that rely on query...

Several approaches have applied Deep Reinforcement Learning (DRL) to Unmanned Aerial Vehicles (UAVs) do autonomous object tracking. These methods, however, are resource intensive and require prior knowledge of the environment, making them difficult use in real-world applications. In this paper, we propose a Lightweight Vision (LDVRL) framework for dynamic tracking that uses camera as only input source. Our employs several techniques such stacks frames, segmentation maps from simulation,...

Drilling and blasting operations are pivotal for productivity safety in hard rock surface mining. These restricted due to complexities such as site-specific uncertainties, risks, environmental economic constraints. Machine Learning (ML) is a transformative approach tackle these resulting significant cost reductions. ML applications can reduce overall costs by up 23% decrease the amount of explosives much 89% compared traditional methods. This survey presents comprehensive review how be...

Research interest in autonomous agents is on the rise as an emerging topic. The notable achievements of Large Language Models (LLMs) have demonstrated considerable potential to attain human-like intelligence agents. However, challenge lies enabling these learn, reason, and navigate uncertainties dynamic environments. Context awareness emerges a pivotal element fortifying multi-agent systems when dealing with situations. Despite existing research focusing both context-aware systems, there...

Large language models (LLMs) enable state-of-the-art semantic capabilities to be added software systems such as search of unstructured documents and text generation. However, these are computationally expensive. At scale, the cost serving thousands users increases massively affecting also user experience. To address this problem, caches used check for answers similar queries (that may have been phrased differently) without hitting LLM service. Due nature cache techniques that rely on query...

Abstract Recent technological advancements have led to a significant increase in digital documents. A document’s key information is generally represented by the keyphrases that provide abstract description contained therein. With traditional keyphrase techniques, however, it difficult identify relevant based on context. Several studies literature explored graph-based unsupervised extraction techniques for automatic extraction. However, there only limited existing work embeds contextual To...

Many application areas such as biodiversity monitoring and speech recognition produce several gigabytes of audio data per day, think weeks, months accumulated available to analyse. However, current methods frameworks can only process a few megabytes at time. Moreover, these require lot manual processing before they give relevant results for researchers. In this paper, we present novel scalable framework called B2P2 handling large files them in an efficient manner. The proposed handle that...

High-dimensional data is becoming more and available due to the advent of big IoT. Having dimensions makes analysis cumbersome increasing sparsity points problem called "curse dimensionality". To address this problem, global dimensionality reduction techniques are used; however, these ineffective in revealing hidden outliers from high-dimensional space. This behaviour being subspace where they belong; hence, a locally relevant needed reveal outliers. In paper, we present technique that...

Anomaly detection is an important research area in data mining and has been studied intensively recent years. The increasing number of features, non-informative noises other irrelevant makes it challenging to detect anomalies high-dimensional data. When analyzing data, are difficult identify due the sparsity caused by curse dimensionality. One most commonly used algorithm for reducing dimensionality Principal Component Analysis (PCA), however, known be sensitive identifying anomalies....

Engineering safety-critical systems such as medical devices and digital health intervention is complex, where long-term engagement with subject-matter experts (SMEs) needed to capture the systems' expected behaviour. In this paper, we present a novel approach that leverages Large Language Models (LLMs), GPT-3.5 GPT-4, potential world model accelerate engineering of software systems. This involves using LLMs generate logic rules, which can then be reviewed informed by SMEs before deployment....

The rapid growth of both the Industrial Internet Things (IIoT) and Artificial Intelligence (AI) results in a high demand for AI applications devices. To achieve levels accuracy, typically require large amount annotated data. Accessing such data is challenging various as healthcare, finance information security. Federated learning (FL) one strategies that was proposed to overcome this challenge. Specifically, FL enables model centralized system be trained without any prior knowledge on Recent...

Machine Learning (ML) techniques in clinical decision support systems are scarce due to the limited availability of clinically validated and labelled training data sets. We present a framework (1) enable quality controls at submission toward ML appropriate data, (2) provide in-situ algorithm assessments, (3) prepare dataframes for robust stochastic analysis. developed evaluated PiMS (Pandemic Intervention Monitoring Systems): remote monitoring solution patients that Covid-positive. The...

Machine reading comprehension (MRC) is a challenging task in natural language processing that demonstrates the understanding of machine. An approach to tackle this challenge requires machine answer question about given context when needed and abstain from answering there no answer. Recent works attempted solve with various comprehensive neural network architectures for sequences such as SAN, U-Net, EQuANt, others were trained on SQuAD 2.0 dataset containing unanswerable questions. However,...

In many real-world applications such as the inspection of powerlines, automated detection anomalies can minimise damage and reduce costs that result from presence unknown anomalies. Technologies LiDAR scans obtained Unmanned Aerial Vehicles (UAV) are becoming prominent due to data depth they provide. context powerline transmission, investigators must search for anomalous elements line defects or obstructions. Such occurrences not always apparent detecting them requires extensive analysis...