Artificial Intelligence Tasks Research Articles

The role of student motivation in integrating AI into web design education: A longitudinal study

Amidst the current wave studies of artificial intelligence (AI) in education, this longitudinal case study, spanning Spring 2023 to Spring 2024, delves into the integration of AI in the UI/UX web design classroom. By introducing both text-based and image-based AI tools to students with varying levels of skill in introductory web design and user experience (UX) courses, the study observed a significant enhancement in student creative capabilities and project outcomes. The utilization of text-based generators markedly improved writing efficiency and coding, while image-based tools facilitated better ideation and color selection. These findings underscore the potential to augment traditional educational methods, providing students with novel avenues for creativity and innovation. At the same time, the goal of this study was also to ascertain the factors that led to the adoption of AI tools in the educational workflow, specifically focusing on student major and background, thereby illuminating how AI can be tailored to meet diverse educational needs and foster a more adaptive and innovative learning environment. The findings reveal that students were more receptive to integrating AI tasks into their workflows when these tasks did not directly relate to their major field of study. For example, Computer Science students exhibited less resistance to using AI for selecting color palettes, a task outside their primary focus, compared to utilizing AI for coding. Additionally, the study observed a significant growth in both awareness and usage of AI tools among students throughout the duration of the research. This trend suggests an increasing incorporation of AI technologies into their standard toolkit, highlighting a broader acceptance and integration of AI in educational practices.

Harnessing ferroic ordering in thin film devices for analog memory and neuromorphic computing applications down to deep cryogenic temperatures

The future computing beyond von Neumann era relies heavily on emerging devices that can extensively harness material and device physics to bring novel functionalities and can perform power-efficient and real time computing for artificial intelligence (AI) tasks. Additionally, brain-like computing demands large scale integration of synapses and neurons in practical circuits that requires the nanotechnology to support this hardware development, and all these should come at an affordable process complexity and cost to bring the solutions close to market rather soon. For bringing AI closer to quantum computing and space technologies, additional requirements are operation at cryogenic temperatures and radiation hardening. Considering all these requirements, nanoelectronic devices utilizing ferroic ordering has emerged as one promising alternative. The current review discusses the basic architectures of spintronic and ferroelectric devices for their integration in neuromorphic and analog memory applications, ferromagnetic and ferroelectric domain structures and control of their dynamics for reliable multibit memory operation, synaptic and neuronal leaky-integrate-and-fire (LIF) functions, concluding with their large-scale integration possibilities, challenges and future research directions.

The accelerated tensor Kaczmarz algorithm with adaptive parameters for solving tensor systems

Solving tensor systems is a common task in scientific computing and artificial intelligence. In this paper, we propose a tensor randomized average Kaczmarz method with adaptive parameters that exponentially converges to the unique least Frobenius norm solution of a given consistent tensor system under the t-product structure. In order to accelerate convergence, a tensor average Kaczmarz method based on stochastic heavy ball momentum technique (tAKSHBM) is proposed. The tAKSHBM method utilizes iterative information to update parameters instead of relying on prior information, addressing the problem in the adaptive learning of parameters. Additionally, the tAKSHBM method based on Fourier transform is proposed, which can be effectively implemented in a distributed environment. It is proven that the iteration sequences generated by all the proposed methods are convergent for given consistent tensor systems. Finally, we conduct experiments on both synthetic data and practical applications to support our theoretical results and demonstrate the effectiveness of the proposed algorithms.

리처드 파워스의 소설『갈라테아 2.2』에 기술된 인공 지능(AI) 헬렌(Helen)의 특징적인 언어양상 연구

Richard Power’s novel Galatea 2.2 presents the task and prospect of Artificial Intelligence’s language although the novel is based on story-telling and the author’s imagination. In particular, it is worth to note not only terms of address and terms of reference but also terms of body concerning Artificial Intelligence Helen. She is mostly addressed by her first name (Helen) and other terms such as machine, apparatus, bag of networks, fancy tin-can telephone, distributed process, multidimentional shape, etc. Also, Artificial Intelligence Helen is a symbol of A (a twenty-two year old woman and post-graduate student) and imitating A, some terms like ‘a gifted cockatiel’ and ‘twin’ are used for Helen in terms of reference. In comparison, Rick is addressed in various ways, that is first name, surname, affectionate names and other nicknames. In addition, Artificial Intelligence Helen has no nose, mouth or fingers and only she has partially functioning eyes and ears. Thus Helen is not able to fully understand metaphorical language expressions concerning emotion. For instance, in some expressions like ‘your stomach in knots’, ‘embarrasment shrinks you’, or ‘amazement strikes you dead’. Helen could not feel the meanings and thus she has a lack of knowledge in similar linguistic circumstances. Lastly, Artificial Intelligence Helen is Imp. H. It means that Helen has been evolved from Imp. A, Imp. B, Imp. C, and so on. So the shut down of herself in the novel suggests and hints at the revival of a new version of Artificial Intelligence may be named “Galatea 3.3”.

American Society of Retina Specialists Artificial Intelligence Task Force Report

American Society of Retina Specialists Artificial Intelligence Task Force Report

IN-MEMORY INTELLIGENT COMPUTING

Context. Processed big data has social significance for the development of society and industry. Intelligent processing of big data is a condition for creating a collective mind of a social group, company, state and the planet as a whole. At the same time, the economy of big data (Data Economy) takes first place in the evaluation of processing mechanisms, since two parameters are very important: speed of data processing and energy consumption. Therefore, mechanisms focused on parallel processing of large data within the data storage center will always be in demand on the IT market.
 Objective. The goal of the investigation is to increase the economy of big data (Data Economy) thanks to the analysis of data as truth table addresses for the identification of patterns of production functionalities based on the similarity-difference metric.
 Method. Intelligent computing architectures are proposed for managing cyber-social processes based on monitoring and analysis of big data. It is proposed to process big data as truth table addresses to solve the problems of identification, clustering, and classification of patterns of social and production processes. A family of automata is offered for the analysis of big data, such as addresses. The truth table is considered as a reasonable form of explicit data structures that have a useful constant – a standard address routing order. The goal of processing big data is to make it structured using a truth table for further identification before making actuator decisions. The truth table is considered as a mechanism for parallel structuring and packing of large data in its column to determine their similarity-difference and to equate data at the same addresses. Representation of data as addresses is associated with unitary encoding of patterns by binary vectors on the found universe of primitive data. The mechanism is focused on processorless data processing based on read-write transactions using in-memory computing technology with significant time and energy savings. The metric of truth table big data processing is parallelism, technological simplicity, and linear computational complexity. The price for such advantages is the exponential memory costs of storing explicit structured data.
 Results. Parallel algorithms of in-memory computing are proposed for economic mechanisms of transformation of large unstructured data, such as addresses, into useful structured data. An in-memory computing architecture with global feedback and an algorithm for matrix parallel processing of large data such as addresses are proposed. It includes a framework for matrix analysis of big data to determine the similarity between vectors that are input to the matrix sequencer. Vector data analysis is transformed into matrix computing for big data processing. The speed of the parallel algorithm for the analysis of big data on the MDV matrix of deductive vectors is linearly dependent on the number of bits of the input vectors or the power of the universe of primitives. A method of identifying patterns using key words has been developed. It is characterized by the use of unitary coded data components for the synthesis of the truth table of the business process. This allows you to use read-write transactions for parallel processing of large data such as addresses.
 Conclusions. The scientific novelty consists in the development of the following innovative solutions: 1) a new vector-matrix technology for parallel processing of large data, such as addresses, is proposed, characterized by the use of read-write transactions on matrix memory without the use of processor logic; 2) an in-memory computing architecture with global feedback and an algorithm for matrix parallel processing of large data such as addresses are proposed; 3) a method of identifying patterns using keywords is proposed, which is characterized by the use of unitary coded data components for the synthesis of the truth table of the business process, which makes it possible to use the read-write transaction for parallel processing of large data such as addresses. The practical significance of the study is that any task of artificial intelligence (similarity-difference, classification-clustering and recognition, pattern identification) can be solved technologically simply and efficiently with the help of a truth table (or its derivatives) and unitarily coded big data . Research prospects are related to the implementation of this digital modeling technology devices on the EDA market. KEYWORDS: Intelligent

Deep Learning on Graphs: A Data-Centric Exploration

Many learning tasks in Artificial Intelligence (AI) require dealing with graph data, ranging from biology and chemistry to finance and education. As powerful deep learning tools for graphs, graph neural networks (GNNs) have demonstrated remarkable performance in various graph-related applications. Despite the significant accomplishments of GNNs, recent studies have highlighted that their efficiency and effectiveness face significant challenges such as adversarial robustness and scalability, which are fundamentally linked to data. While major attention has been devoted to improving GNNs from the model perspective, the potential of directly enhancing data has often been overlooked. It underscores a critical gap in GNN research---while model improvements are undoubtedly important, we also need to recognize and address the data-related factors contributing to the challenges. Hence, my research is to investigate solutions for these challenges from the data perspective, employing strategies such as data characterization, reduction, augmentation, transformation, and detection.

Abstract 3449: Is AI-enhanced breast ultrasound ready for breast cancer screening in low-resource environments? A systematic review

Abstract Purpose. Screening mammography is unavailable in many low-resource areas. We ask if the state-of-the-art in artificial intelligence (AI)-enhanced breast ultrasound (BUS) is sufficiently accurate to be used for primary breast cancer screening in low-resource regions. Background. Since the 1980s, high-income countries have implemented mammographic screening programs, leading to breast cancer mortality reduction in screened women.1 Mammography is unavailable in many low-resource regions, such as the USAPI. Furthermore, travel difficulties and lack of radiologists hinder implementation. AI combined with portable BUS may address limitations of the high-income paradigm. In this systematic review, we ask if AI-enhanced BUS can detect/segment lesions (Objective 1) and classify lesions as cancerous (Objective 2). Methods. Two reviewers independently assessed articles from 1/1/2016 to 8/6/2023 from PubMed, Google Scholar, and citation searching. Studies which report on AI development and report performance on a patient-wise, held-out test set met the inclusion criteria. Studies were characterized by AI task and clinical application time. Dataset composition and performance were examined via narrative data synthesis. QUADAS-2 bias assessment was performed using criteria for each AI task. Success in (2) is defined by meeting minimum screening performance guidelines.2,3 Results. PubMed yielded 281 studies, Google Scholar yielded 225 studies, and a manual citation search yielded 41 studies. From 382 unique full texts evaluated, 52 articles met all inclusion criteria: 3 frame selection, 2 real-time detection, 2 combination, 14 segmentation-only, and 31 classification-only. Lesion segmentation-only models achieved a 90th percentile Dice similarity coefficient of 0.913 on generally small datasets. The best evidence for lesion cancer classification reported 0.976 area under the curve. All studies faced elevated bias and applicability concerns under QUADAS-2. Conclusion. Reported performance for (1) is insufficient to introduce AI-enhanced BUS for breast cancer screening. Evidence supporting AI-enhanced BUS for (2) is dependent on few studies relying on internal datasets, limiting generalizability. Geographically diverse clinical trials are needed to confirm and improve robustness of performance of AI-enhanced BUS for (1) and (2). References. 1. Marmot MG, et al. The benefits and harms of breast cancer screening: an independent review. British journal of cancer. 2013;108(11):2205-2240. 2. Lehman CD, et al. National Performance Benchmarks for Modern Screening Digital Mammography: Update from the Breast Cancer Surveillance Consortium. Radiology. 2017-04-01 2017;283(1):49-58. doi:10.1148/radiol.2016161174 3. Rosenberg RD, et al. Performance Benchmarks for Screening Mammography. Radiology. 2006-10-01 2006;241(1):55-66. doi:10.1148/radiol.2411051504 Citation Format: Arianna Bunnell, Dustin Valdez, Fredrik Strand, Yannik Glaser, Peter Sadowski, John A. Shepherd. Is AI-enhanced breast ultrasound ready for breast cancer screening in low-resource environments? A systematic review [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3449.

Embedded machine learning of IoT streams to promote early detection of unsafe environments

Indoor particulate matter (PM) are small solid and liquid particles present in the air, and its monitoring is one of the key challenges regarding workplace safety because of its impact on human health. To address this issue, the Internet of Things (IoT) paradigm allows the implementation of hyperlocal monitoring systems, typically using traditional cloud architectures, which can be enhanced using edge computing architectures. For this reason, we propose an IoT-Edge-Cloud architecture for a platform which promotes the early detection of unsafe environments through machine learning, composed of a sensing layer that collects all the data, an edge layer that performs the artificial intelligence tasks and a cloud layer orchestrating. This architecture is based on the FogFlow framework and the FIWARE components. Our solution proposes an embedded model that can predict the occurrence of PM values higher than the recommended ones—according to the Occupational Safety and Health Administration (OSHA) indicators—with an 87 % of accuracy and a reduction of latency of 26 %. Our platform is innovative because it is based on the FogFlow framework for edge computing and supported by the Smart Spot device validated in a field test. This step is missing from similar state-of-the-art platforms. Thus, we believe that this work contributes to demonstrating the usefulness of AIoT to monitor workplace safety and make trustable predictions, avoiding risky environments.

Engineering Ferroelectric‐/Ion‐Modulated Conductance in 2D vdW CuInP2S6 for Non‐Volatile Digital Memory and Artificial Synapse

AbstractTwo‐dimensional (2D) ferroionics is appealing in performing complex artificial intelligence tasks due to the interesting property of coexistence of ferroelectricity and ionic activities. CuInP2S6 (CIPS), as a typical 2D ferroionic material, is highly conducive to rich functions of information devices due to the displacement of Cu+ inducing both ferroelectricity and ionic conductivity. However, the coupling and modulation of polarization and ion migration in CIPS for multifunctional information devices has not been fully explored. Here, this study demonstrates that digital memory and synaptic simulations are realized in Au/CIPS/Au device via engineering ferroelectric polarization reversal and the long‐distance migration of the Cu+ to change conductive modes. Steep resistive switching behavior based on ion‐migration is observed with a high on/off ratio of over 108, long retention time (>2 × 104 s), and current compliance engineered multilevel resistance states, demonstrating reliable nonvolatile high‐density memory characteristics. Based on the continuous modulation of polar order, the key synaptic behaviors are successfully simulated. Moreover, by the co‐modulation of polarization state and ions migration, the paired‐pulse facilitation, paired‐pulse depression, and potentiation following depression are achieved. These results suggest that CIPS is a promising candidate for constructing high‐performance, function‐enriched devices for data storage, information processing, and neuromorphic computing.

Toward the novel AI tasks in infection biology.

Machine learning and artificial intelligence (AI) are becoming more common in infection biology laboratories around the world. Yet, as they gain traction in research, novel frontiers arise. Novel artificial intelligence algorithms are capable of addressing advanced tasks like image generation and question answering. However, similar algorithms can prove useful in addressing advanced questions in infection biology like prediction of host-pathogen interactions or inferring virus protein conformations. Addressing such tasks requires large annotated data sets, which are often scarce in biomedical research. In this review, I bring together several successful examples where such tasks were addressed. I underline the importance of formulating novel AI tasks in infection biology accompanied by freely available benchmark data sets to address these tasks. Furthermore, I discuss the current state of the field and potential future trends. I argue that one such trend involves AI tools becoming more versatile.

Exploring the Promise and Challenges of Artificial Intelligence in Biomedical Research and Clinical Practice.

Artificial intelligence (AI) is poised to revolutionize how science, and biomedical research in particular, are done. With AI, problem solving and complex tasks using massive data sets can be performed at a much higher rate and dimensionality level compared to humans. With the ability to handle huge data sets and self-learn, AI is already being exploited in drug design, drug repurposing, toxicology, and material identification. AI could also be used in both basic and clinical research in study design, defining outcomes, analyzing data, interpreting findings, and even identifying the most appropriate areas of investigation and funding sources. State-of-the-art AI-based large language models (LLM), such as ChatGPT and Perplexity, are positioned to change forever how science is communicated and how scientists interact with one another and their profession, including post-publication appraisal and critique. Like all revolutions, upheaval will follow and not all outcomes can be predicted, necessitating guardrails at the onset, especially to minimize the untoward impact of the many drawbacks of LLMs, which include lack of confidentiality, risk of hallucinations, and propagation of mainstream albeit potentially mistaken opinions and perspectives. In this review, we highlight areas of biomedical research that are already being reshaped by AI and how AI is likely to impact it further in the near future. We discuss the potential benefits of AI in biomedical research and address possible risks, some surrounding the creative process, that warrant further reflection.

End to End AI Architecture for Next Generation Network

Targeting 6G network architecture with native intelligence, a holistic view on the end to end (E2E) Artificial Intelligence (AI) flow across Operation Administration and Maintenance (OAM), Core Network (CN), Edge, Radio Access Network (RAN), and User Equipment (UE) domains is presented in this article. The AI Task Management (AITM) is introduced as the enabler in such network, and a User Equipment Data Analytics (UEDA) is proposed to take advantage of the increasing computing capability of UEs while taking their power consumption limitation into consideration. Conventional control plane and user plane connections assume fixed one-to-one end points and a single chosen transmission mode. Depending on the goal of the AI task and what network entities need to participate, the endpoints of the AI task connection is uncertain, with a group of participant UEs on one side and base station, core network or application server on the other. The transmission mode of such connection may need to be dynamically reconfigured as well. This article proposes an approach of AI task-oriented connections, where logical connections among UE, RAN, CN, and applications are established via AITM for AI model distribution, data collection, as well as parameter synchronization across different domains. A new wireless bearer type, named AI Radio Bearer (AIRB), for AI task-oriented connections is proposed with due design considerations and its associated performance.

Learning Relation-Enhanced Hierarchical Solver for Math Word Problems.

Automatically solving math word problems (MWPs) is a challenging task for artificial intelligence (AI) and machine learning (ML) research, which aims to answer the problem with a mathematical expression. Many existing solutions simply model the MWP as a sequence of words, which is far from precise solving. To this end, we turn to how humans solve MWPs. Humans read the problem part-by-part and capture dependencies between words for a thorough understanding and infer the expression precisely in a goal-driven manner with knowledge. Moreover, humans can associate different MWPs to help solve the target with related experience. In this article, we present a focused study on an MWP solver by imitating such procedure. Specifically, we first propose a novel hierarchical math solver (HMS) to exploit semantics in one MWP. First, to imitate human reading habits, we propose a novel encoder to learn the semantics guided by dependencies between words following a hierarchical "word-clause-problem" paradigm. Next, we develop a goal-driven tree-based decoder with knowledge application to generate the expression. One step further, to imitate human associating different MWPs for related experience in problem-solving, we extend HMS to the Relation-enHanced Math Solver (RHMS) to utilize the relation between MWPs. First, to capture the structural similarity relation, we develop a meta-structure tool to measure the similarity based on the logical structure of MWPs and construct a graph to associate related MWPs. Then, based on the graph, we learn an improved solver to exploit related experience for higher accuracy and robustness. Finally, we conduct extensive experiments on two large datasets, which demonstrates the effectiveness of the two proposed methods and the superiority of RHMS.

Analysis the Principle of AI Chip Principle and the State-of-art Applications

In recent years, AI chips are widely used in advance IC manufacturing. On this basis, this study provides an in-depth analysis of the principles and state-of-the-art applications of AI chips. This research first briefly introduces the history of artificial intelligence and the emergence of artificial intelligence chips as specialized hardware components aimed at effectively performing complex calculations required for artificial intelligence tasks. The article then delves into the basic definition and description of AI chips, the differences between AI chips and ordinary chips, the architecture and manufacturing process of AI chips, application scenarios and results, and the limitations of current AI chips. At the same time, the study also emphasizes the impact of artificial intelligence chips on the field of artificial intelligence research and their potential to shape the future of computing. Overall, the result of the study is a valuable resource for anyone interested in understanding the principles and applications of AI chips.

Analysis the Principle of AI Chips and the State-of-art Applications

In recent years, the artificial intelligence (AI) chips have emerged as indispensable components specialized for executing AI computations, which is considered as a new era for integrate circuit industry. On this basis, this study comprehensively examines the underlying principles, performance characteristics, diverse applications, limitations, and prospective directions in AI chip research. The parallel architecture, sparsity, quantization, and adaptability of AI chips empower the expeditious and efficient processing of AI tasks. These chips find versatile applications in computer vision, healthcare, and intelligent robotics, amplifying their transformative impact. Nevertheless, challenges persist in terms of scalability, versatility, and the rapid pace of technological evolution. As a matter of fact, ongoing research endeavors should prioritize the symbiotic optimization of hardware and software, the creation of energy-efficient designs tailored to specific domains, and the seamless integration of emerging technologies. These results offer a lucid depiction of the prevailing state and the anticipated trajectory of AI chip development.

The legal regulation of artificial intelligence security in Ukrainian banking

The need to expand the range of banking services in Ukraine is stipulated with technological progress, the European integration processes and the legal regime of martial law introduced in the country. Under the conditions of war, the need to strengthen the security of banking activities and protect the banking system from the influence of any internal and external factors gains meaning. The topical direction of economic and legal research of scientists today is the possibility to introduce digital technologies with elements of artificial intelligence (AI) into the banking activity in Ukraine to improve its protection. The AI law as an independent branch of the Ukrainian law has not been developed so far. The sources of AI law, its functions, tasks, scope, risks and limits of legal responsibility for prohibited practices of artificial intelligence have not been defined. The purpose of the article is to analyze the theoretical and legal provisions that underpin the regulation of AI application in Ukrainian banking. The comparative legal method made it possible, considering the provisions of the draft law on AI of the European Union, to determine the trends in the development of the legal regulation of AI in Ukraine. Following the study, proposals to the legislation of Ukraine were formulated, which will contribute to the legal regulation of banking activities using digital technologies with elements of AI.

The Method Of Logical Networks For Modeling Adaptive Knowledge Testing Systems

The subject of the research is the development of mathematical and algorithmic support of the intellectual toolkit for the analysis of sets of test tasks and the modeling of the process of interpreting the quality of sets of test tasks, which allows for objective and comprehensive continuous control of the knowledge of subjects of training, subject to the implementation of the concept of virtual distributed training (retraining).
 The purpose of the research is to improve the effectiveness of monitoring the knowledge of subjects of education in the distance form of education through the use of adaptive computer testing methods based on models of logical networks and the algebra of finite predicates.
 The following tasks are solved in the article: the formation of a testing model in a distributed virtual learning environment and a model of validity assessment based on the content of sets of test tasks.
 The following methods are used: methods of logical networks and algebraic programming based on the algebra of finite predicates and predicate operations, intellectual analysis of information.
 The following results were obtained: the principles of intellectual analysis were formulated in the development of a model of a universal logical network and its application to actual tasks of artificial intelligence in the field of informal information processing, namely, in the construction of knowledge testing systems for distributed virtual learning
 Conclusions. Algorithms for optimal multi-stage adaptive testing of knowledge as part of distributed virtual learning models and methods for analyzing the success of training subjects have been improved. The use of conjunctive decomposition with binary predicates achieves the goal of the research, because in this way any multi-place predicate can be represented by a logical network simulating the process of knowledge testing, the model of the subject of learning is described.

Compression and Transmission of Big AI Model Based on Deep Learning

In recent years, big AI models have demonstrated remarkable performance in various artificial intelligence (AI) tasks. However, their widespread use has introduced significant challenges in terms of model transmission and training. This paper addresses these challenges by proposing a solution that involves the compression and transmission of large models using deep learning techniques, thereby ensuring the efficiency of model training. To achieve this objective, we leverage deep convolutional networks to design a novel approach for compressing and transmitting large models. Specifically, deep convolutional networks are employed for model compression, providing an effective means to reduce the size of large models without compromising their representational capacity. The proposed framework also includes carefully devised encoding and decoding strategies to guarantee the restoration of model integrity after transmission. Furthermore, a tailored loss function is designed for model training, facilitating the optimization of both the transmission and training performance within the system. Through experimental evaluation, we demonstrate the efficacy of the proposed approach in addressing the challenges associated with large model transmission and training. The results showcase the successful compression and subsequent accurate reconstruction of large models, while maintaining their performance across various AI tasks. This work contributes to the ongoing research in enhancing the practicality and efficiency of deploying large models in real-world AI applications.

Handling Data Handoff of AI-Based Applications in Edge Computing Systems

Edge computing aims at better supporting low-latency applications. One of its key techniques is computation offloading, the process that outsources computing tasks from resourced-constrained mobile devices and moves them to edge data centers. In this paper, we tackle an emerging problem within the umbrella of computation offloading, i.e., migration of offloaded inference tasks of Artificial Intelligence (AI) trained models. Such context tailors migration aspects of data-sensitive services where i) the value of the updates is inversely proportional to the data age and ii) outage is highly detrimental to accuracy. To tackle this challenge, we propose Mobile Edge Data-handoff (MED) a framework able to relocate inference or online training tasks from one edge datacenter to another by moving only the necessary data to minimize any accuracy drop during the process. We implemented MED in a well-known edge computing emulator, openLEON, and experimentally verified its performance with an AI-based Industry 4.0 application that forecasts the gas flow in a chemical plant. For our experiments, we use a real, open-source dataset that contains sensors readings. Collected results show that MED, employing proactive data handoff algorithms, is able to minimize the packet loss during the handoff thereby providing guarantees on the inference accuracy.