Currently, explainability represents a major barrier that Artificial Intelligence (AI) is facing in regard to its practical implementation various application domains. To combat the lack of understanding AI-based systems, Explainable AI (XAI) aims make black-box models more transparent and comprehensible for humans. Fortunately, plenty XAI methods have been introduced tackle problem from different perspectives. However, due vast search space, it challenging ML practitioners data scientists...

LangXAI is a framework that integrates Explainable Artificial Intelligence (XAI) with advanced vision models to generate textual explanations for visual recognition tasks. Despite XAI advancements, an understanding gap persists end-users limited domain knowledge in artificial intelligence and computer vision. addresses this by furnishing text-based classification, object detection, semantic segmentation model outputs end-users. Preliminary results demonstrate LangXAI's enhanced plausibility,...

This paper introduces EdgeMLOps, a framework leveraging Cumulocity IoT and thin-edge.io for deploying managing machine learning models on resource-constrained edge devices. We address the challenges of model optimization, deployment, lifecycle management in environments. The framework's efficacy is demonstrated through visual quality inspection (VQI) use case where images assets are processed devices, enabling real-time condition updates within an asset system. Furthermore, we evaluate...

Maintaining high data quality is crucial for reliable analysis and machine learning (ML). However, existing management tools often lack automation, interactivity, integration with ML workflows. This demonstration paper introduces DataLens, a novel interactive dashboard designed to streamline automate the process tabular data. DataLens integrates suite of profiling, error detection, repair tools, including statistical, rule-based, ML-based methods. It features user-in-the-loop module rule...

This paper introduces GLLM, an innovative tool that leverages Large Language Models (LLMs) to automatically generate G-code from natural language instructions for Computer Numerical Control (CNC) machining. GLLM addresses the challenges of manual writing by bridging gap between human-readable task descriptions and machine-executable code. The system incorporates a fine-tuned StarCoder-3B model, enhanced with domain-specific training data Retrieval-Augmented Generation (RAG) mechanism....

This paper presents an approach integrating explainable artificial intelligence (XAI) techniques with adaptive learning to enhance energy consumption prediction models, a focus on handling data distribution shifts. Leveraging SHAP clustering, our method provides interpretable explanations for model predictions and uses these insights adaptively refine the model, balancing complexity predictive performance. We introduce three-stage process: (1) obtaining values explain predictions, (2)...

Nowadays, machine learning plays a key role in developing plenty of applications, e.g., smart homes, medical assistance, and autonomous driving. A major challenge these applications is preserving high quality the training serving data. Nevertheless, existing data cleaning methods cannot exploit context information. Thus, they usually fail to track shifts distributions or associated error profiles. To overcome limitations, we introduce, this paper, novel method for automated tabular powered...

Data drifts pose a critical challenge in the lifecycle of machine learning (ML) models, affecting their performance and reliability. In response to this challenge, we present microbenchmark study, called D3Bench, which evaluates efficacy open-source drift detection tools. D3Bench examines capabilities Evidently AI, NannyML, Alibi-Detect, leveraging real-world data from two smart building use cases.We prioritize assessing functional suitability these tools identify analyze drifts....

Machine learning algorithms have become increasingly prevalent in multiple domains, such as autonomous driving, healthcare, and finance. In data preparation remains a significant challenge developing accurate models, requiring expertise time investment to explore the huge search space of well-suited curation transformation tools. To address this challenge, we present AutoCure, novel configuration-free pipeline that improves quality tabular data. Unlike traditional methods, AutoCure...

Nowadays, machine learning (ML) plays a vital role in many aspects of our daily life. In essence, building well-performing ML applications requires the provision high-quality data throughout entire life-cycle such applications. Nevertheless, most real-world tabular suffer from different types discrepancies, as missing values, outliers, duplicates, pattern violation, and inconsistencies. Such discrepancies typically emerge while collecting, transferring, storing, and/or integrating data. To...

Visual quality inspection systems, crucial in sectors like manufacturing and logistics, employ computer vision machine learning for precise, rapid defect detection. However, their unexplained nature can hinder trust, error identification, system improvement. This paper presents a framework to bolster visual by using CAM-based explanations refine semantic segmentation models. Our approach consists of 1) Model Training, 2) XAI-based Explanation, 3) XAI Evaluation, 4) Annotation Augmentation...

LangXAI is a framework that integrates Explainable Artificial Intelligence (XAI) with advanced vision models to generate textual explanations for visual recognition tasks. Despite XAI advancements, an understanding gap persists end-users limited domain knowledge in artificial intelligence and computer vision. addresses this by furnishing text-based classification, object detection, semantic segmentation model outputs end-users. Preliminary results demonstrate LangXAI's enhanced plausibility,...

Visual quality inspection systems, crucial in sectors like manufacturing and logistics, employ computer vision machine learning for precise, rapid defect detection. However, their unexplained nature can hinder trust, error identification, system improvement. This paper presents a framework to bolster visual by using CAM-based explanations refine semantic segmentation models. Our approach consists of 1) Model Training, 2) XAI-based Explanation, 3) XAI Evaluation, 4) Annotation Augmentation...

Machine learning's influence is expanding rapidly, now integral to decision-making processes from corporate strategy the advancements in Industry 4.0. The efficacy of Artificial Intelligence broadly hinges on caliber data used during its training phase; optimal performance tied exceptional quality. Data cleaning tools, particularly those that exploit functional dependencies within ontological frameworks or context models, are instrumental augmenting Nevertheless, crafting these models a...

Addressing data quality issues is a challenging task due to the labor-intensive nature of manual cleaning processes and inadequacy automated tools that lack effective repair strategies. In this paper, we introduce ReClean, novel data-cleaning method, dedicated ML pipelines, employs reinforcement learning (RL) optimize tasks. ReClean treats as sequential decision process, where RL agents learn choose optimal operations improve model convergence predictive performance. Our extensive...

Successful exploitation of wireless sensor networks (WSNs) depends intuitively on the enabling technologies as well provision application-level quality service (QoS). Current research mostly focuses maximising network lifetime without considering predefined task time. In this paper, we firstly provide a survey QoS control approaches for WSNs. Subsequently, propose novel method, referred to planning. Based design-time knowledge, planning provides users/applications with best-effort QoS, while...

Nowadays, machine learning plays a key role in developing plenty of applications, e.g., smart homes, medical assistance, and autonomous driving. A major challenge these applications is preserving high quality the training serving data. Nevertheless, existing data cleaning methods cannot exploit context information. Thus, they usually fail to track shifts distributions or associated error profiles. To overcome limitations, we introduce, this paper, novel method for automated tabular powered...

Recently, several approaches have been proposed to automatically model indoor environments.Most of such efforts principally rely on the crowd sense data as motion traces, images, and WiFi footprints.However, large datasets are usually required derive precise models which can negatively affect energy efficiency mobile devices participating in crowd-sensing system.Furthermore, aforementioned types hardly suitable for deriving 3D models.To overcome these challenges, we propose GraMap, a...

Well drilling and wellbore conditioning typically involve multiple specialized tools. Systematic application of such tools can be a challenge from logistics maintenance perspective, which eventually has an impact on the efficiency process as well resulting condition. This paper presents new multifunctional tool developed to address operational challenges. The tool’s effect logging is illustrated through field case study. CRS able Cut, Ream Stabilize when installing in BHA. It envisioned...

In this paper, we present our vision of differentiable ML pipelines called DiffML to automate the construction in an end-to-end fashion. The idea is that allows jointly train not just model itself but also entire pipeline including data preprocessing steps, e.g., cleaning, feature selection, etc. Our core formulate all steps a way such can be trained using backpropagation. However, non-trivial problem and opens up many new research questions. To show feasibility direction, demonstrate...