Rapid adoption of the Internet of Things (IoT) devices has transformed industries by enabling automation and seamless connectivity, but it also introduces significant security challenges, particularly for malware threats. To address these challenges, numerous AI-based solutions have been proposed to enhance IoT device security. Additionally, federated learning (FL) based solutions have been proposed to enhance security and data privacy guarantees. However, these methods require access to labeled data, but data annotation is an arduous task in security. Further, storing such a large pool of labeled data in memory-constrained IoT devices is difficult. We propose 10T-FedMaIDetect, a novel semi-supervised FL framework for dynamic malware detection in IoT edge devices based on network traces to address these issues. Our approach is a two-stage training framework that combines unsupervised federated learning with supervised fine-tuning. In the first stage, clients train feature learning models independently, and the server aggregates them to build a global model. In the second stage, this global model is paired with a classifier and fine-tuned on a publicly available labeled dataset of mal ware. The resulting model is then used as the deployable detection system at the IoT devices. Our solution achieves comparable detection performance with a ROC-AUC of 0.96 and a PR-AUC of 0.943, with the existing semi-supervised FL method, FedMSE, on the N-BaIoT dataset without requiring any client-side labels. Additionally, we deployed our framework on a real-world IoT testbed to evaluate its deployment feasibility and observed comparable detection performance with reduced memory and CPU resources.

Loss of plasticity is a phenomenon where neural networks can become more difficult to train over the course of learning. Continual learning algorithms seek to mitigate this effect by sustaining good performance while maintaining network trainability. We develop a new technique for improving continual learning inspired by the observation that the singular values of the neural network parameters at initialization are an important factor for trainability during early phases of learning. From this perspective, we derive a new spectral regularizer for continual learning that better sustains these beneficial initialization properties throughout training. In particular, the regularizer keeps the maximum singular value of each layer close to one. Spectral regularization directly ensures that gradient diversity is maintained throughout training, which promotes continual trainability, while minimally interfering with performance in a single task. We present an experimental analysis that shows how the proposed spectral regularizer can sustain trainability and performance across a range of model architectures in continual supervised and reinforcement learning settings. Spectral regularization is less sensitive to hyperparameters while demonstrating better training in individual tasks, sustaining trainability as new tasks arrive, and achieving better generalization performance..

Pressure measurement within microfluidic devices is crucial for a variety of applications, including precise flow control, measurement of mechanical properties of cells, and clog detection. The integration of pressure sensors within microfluidic channels is predicted to improve device performance. This work presents a simple, cleanroom-free technique to integrate pressure sensors in microfluidic devices. Initially, we present an innovative technique for patterning Ti3C2-MXene on PDMS membranes using inkjet printing. The resulting inkjet-printed strain sensor exhibits a high gauge factor of 6256 and can detect minimal strain values as low as 3 x 10-4. The sensor demonstrates a rapid response time around 200 ms and exhibits consistent behavior throughout durability tests comprising 9000 cycles. Furthermore, we demonstrated a straightforward process for fabricating pressure sensors by overlaying a thin PDMS layer onto microfluidic channels, followed by sensor fabrication via inkjet printing. Through this approach, we illustrate the integration of a printed Wheatstone bridge pressure sensor with a microfluidic device. An ultra-sensitive pressure sensor is achieved through the utilization of an 8 mu m thin PDMS membrane. This technique facilitates localized pressure sensing within microfluidic devices, featuring a simple electrical readout. It opens the prospect of measuring local pressures while evaluating the deformability of cells in microfluidic flow cytometry and similar applications.

An agent that efficiently accumulates knowledge to develop increasingly sophisticated skills over a long lifetime could advance the frontier of artificial intelligence capabilities. The design of such agents, which remains a long-standing challenge of artificial intelligence, is addressed by the subject of continual learning. This monograph clarifies and formalizes concepts of continual learning, introducing a framework and set of tools to stimulate further research.

In continual learning, plasticity refers to the ability of an agent to quickly adapt to new information. Neural networks are known to lose plasticity when processing non-stationary data streams. In this paper, we propose L2 Init, a simple approach for maintaining plasticity by incorporating in the loss function L2 regularization toward initial parameters. This is very similar to standard L2 regularization (L2), the only difference being that L2 regularizes toward the origin. L2 Init is simple to implement and requires selecting only a single hyper-parameter. The motivation for this method is the same as that of methods that reset neurons or parameter values. Intuitively, when recent losses are insensitive to particular parameters, these parameters should drift toward their initial values. This prepares parameters to adapt quickly to new tasks. On problems representative of different types of nonstationarity in continual supervised learning, we demonstrate that L2 Init most consistently mitigates plasticity loss compared to previously proposed approaches.

The objective of this paper is to explore the capabilities to implement closed-loop supply chain (CLSC). In this regard, a theoretical model grounded in natural resource-based view is proposed, which depicts inter-relationships among the capabilities and CLSC. The model is tested using survey data from Indian manufacturing firms by partial least squares (PLS) approach. The findings show information technology and organizational learning as important lower-order capabilities, and internal integration, demand management and product design as significant higher-order capabilities for CLSC implementation. To the best of authors' knowledge, this is the first study to examine key capabilities for CLSC. The study contributes to CLSC literature by providing an integrative framework, classifying capabilities into lower-order capabilities and higher-order capabilities, and empirically examining the key capabilities for CLSC. The findings provide managers with insights about the hierarchical levels of capabilities for CLSC, which will help them to accordingly deploy the appropriate resources to build capabilities for CLSC.

A default assumption in the design of reinforcement-learning algorithms is that a decision-making agent always explores to learn optimal behavior. In sufficiently complex environments that approach the vastness and scale of the real world, however, attaining optimal performance may in fact be an entirely intractable endeavor and an agent may seldom find itself in a position to complete the requisite exploration for identifying an optimal policy. Recent work has leveraged tools from information theory to design agents that deliberately forgo optimal solutions in favor of sufficiently-satisfying or satisficing solutions, obtained through lossy compression. Notably, such agents may employ fundamentally different exploratory decisions to learn satisficing behaviors more efficiently than optimal ones that are more data intensive. While supported by a rigorous corroborating theory, the underlying algorithm relies on model-based planning, drastically limiting the compatibility of these ideas with function approximation and high-dimensional observations. In this work, we remedy this issue by extending an agent that directly represents uncertainty over the optimal value function allowing it to both bypass the need for model-based planning and to learn satisficing policies. We provide simple yet illustrative experiments that demonstrate how our algorithm enables deep reinforcement-learning agents to achieve satisficing behaviors. In keeping with previous work on this setting for multi-armed bandits, we additionally find that our algorithm is capable of synthesizing optimal behaviors, when feasible, more efficiently than its non-information-theoretic counterpart.

The "small agent, big world" frame offers a conceptual view that motivates the need for continual learning. The idea is that a small agent operating in a much bigger world cannot store all information that the world has to offer. To perform well, the agent must be carefully designed to ingest, retain, and eject the right information. To enable the development of performant continual learning agents, a number of synthetic environments have been proposed. However, these benchmarks suffer from limitations, including unnatural distribution shifts and a lack of fidelity to the "small agent, big world" framing. This paper aims to formalize two desiderata for the design of future simulated environments. These two criteria aim to reflect the objectives and complexity of continual learning in practical settings while enabling rapid prototyping of algorithms on a smaller scale.

PurposeBlockchain technology has been labeled as the most disruptive technological innovation of the current decade due to its impact on almost every major industry. Based on privacy calculus theory and prior adoption literature on emerging technologies, this research investigates the impact of blockchain technology in the consumer technology segment. It elaborated on the mechanism through which blockchain technology influences users' willingness to share information with technology products enabled by blockchain.Design/methodology/approachTaking a heterogeneous pool of users, this study conducted multiple experiments with the application of blockchain (vs. regular database) technology to high (vs. low) sensitive data to study the impact of blockchain perception on users' information-sharing tendencies.FindingsThrough a mediated moderation analysis, the result shows that the use of blockchain technology enhances the sense of security among users. However, the impact of this heightened sense of security only develops a higher willingness to share information when the data is highly sensitive.Practical implicationsThe research reflects on the perception of blockchain technology and the leading impact on willingness to share information with firms. This could be a critical criterion for determining investment in blockchain technologies for consumer products, particularly based on the sensitivity of the data the consumer is sharing.Originality/valueThis research focuses on the perception of blockchain technology among consumers and its impact on consumers' decision-making related to their data sharing. People have a higher sense of safety when it comes to blockchain-enabled products. However, we find that it would not be the same for all contexts, and the sensitivity of the data collected would have an impact on this relationship and consumers' data-sharing decisions.

The implementation of Industry 4.0 (I4) technologies has gained momentum due to several inherent benefits associated with their adoption. However, the benefits of I4 technologies are yet to be realised to the full potential, specifically in the case of emerging economies. Managers need to focus on certain critical factors for the successful implementation of I4 technologies. Though some studies have proposed factors for implementing I4 technologies, empirical examination of critical factors still lacks in the published literature. This study proposes and empirically analyses the critical factors for adopting I4 technologies in the following Indian manufacturing industries: electrical/electronics, automotive, textiles, paper and plastics. The key factors across six different categories (organisational, workforce management, external support, technological infrastructure, usage of data and regulations) are examined. Further, the contingency effects of firm size and industry sector are also examined. The results are useful for managers in manufacturing industries as it can help them to understand the key factors for adopting I4 technologies. The results are equally useful for managers who are planning to implement I4 technologies in their firms or are in the early phase of I4 implementation.

Electrocatalysts with a synergistic combination of low-cost, eco-friendly, and highly electrochemical active properties are frequently used in hydrogen evolution reactions (HERs) without utilizing any noble or toxic metals. Herein, a simple chemical oxidative polymerization method has been used to synthesize a set of nanocomposites of polypyrrole (PPy) and Ni-doped NASICON-structured Na3NixFe(2-x)(PO4)(SO4)2 [NFPS(Nix)], which shows remarkable HER activity. Among them, as-synthesized uniform spherical-shaped PPy/NFPS(Ni0.5) (optimized ratio) was found to show the lowest onset overpotential, -13 mV versus reversible hydrogen electrode (RHE), with a lesser Tafel slope of 58 mV dec-1, which is comparable to Pt/C and superior among the majority of the polymer-based HER catalysts. Also, it offers a current density of 10 mA/cm2 at a noticeably lower overpotential of -111 mV versus RHE. Moreover, the nanocomposite also exhibits notable long-sustaining environmental and catalytic stability along with exceptional durability. In PPy/NFPS(Ni0.5), improved electrocatalytic activity comes from the synergistic effect of the NFPS(Ni0.5) and PPy matrix, which results in enormous electrocatalytically active sites and facilitates easy charge transportation. Thus, real-world hydrogen production may hold possibilities for the as-synthesized PPy/NFPS(Ni0.5) nanocomposite as well as an alternative to a Pt-based electrocatalyst for the HER.

Polyanionic-type electrode material made of (PO4)n- polyhedral covalently bonded with Ni-O linkage is envisaged as a novel electrode material for intercalative battery-type hybrid supercapacitors. Highly porous, flake-type KNiPO4 showed robust electrochemical performances as a result of its open framework structure and active participation of the Ni2+/3+ redox couple that results in superior pseudocapacitive intercalating charge storage in the aqueous KOH electrolyte. The KNiPO4 electrode shows the specific charge storage equivalent to 168.5 mAh/g (capacitance of 935 F/g) at 1 A/g current rate in the potential window of 0.65 V in the aqueous 2 M KOH electrolyte. KNiPO4 electrodes exhibit excellent long-term cycle stability at 10 A/g for 5000 cycles with 87% of the initial capacity retention of the electrode and coulombic efficiency (eta = td/tc) equivalent to 95.1% after 5000 cycles. Further, in full cell hybrid supercapacitor (HSC) mode in which porous KNiPO4 acted as the positive electrode and activated carbon (AC) functioned as the negative electrode, in the voltage window of 1.6 V, the highest energy density equivalent to 200 Wh/kg and power density equivalent to similar to 819 W/kg were obtained at 1 A/g current rate. At a higher current rate (10 A/g), the hybrid supercapacitor attains a very high power density equivalent to 7981 W/kg with a retention of energy density close to 75 Wh/kg with superior cyclic stability. Coulombic efficiency of the full cell [asymmetric supercapacitor (ASC) mode] has lost only 3.4% with excellent capacity retention (92.3%) of its initial value after 2200 cycles. The robust performance and long cycle life of the electrode in full cells confirm the applicability of the material to power implantable biomedical devices. Further, high power performance coupled with superior cyclic stability coupled with strong electrochemical energy storage properties of the KNiPO4 electrode makes it suitable for bulk, grid-level charge storage applications.

Approaches to policy optimization have been motivated from diverse principles, based on how the parametric model is interpreted (e.g. value versus policy representation) or how the learning objective is formulated, yet they share a common goal of maximizing expected return. To better capture the commonalities and identify key differences between policy optimization methods, we develop a unified perspective that re-expresses the underlying updates in terms of a limited choice of gradient form and scaling function. In particular, we identify a parameterized space of approximate gradient updates for policy optimization that is highly structured, yet covers both classical and recent examples, including PPO. As a result, we obtain novel yet well motivated updates that generalize existing algorithms in a way that can deliver benefits both in terms of convergence speed and final result quality. An experimental investigation demonstrates that the additional degrees of freedom provided in the parameterized family of updates can be leveraged to obtain non-trivial improvements both in synthetic domains and on popular deep RL benchmarks.

Recent global concerns over continuously increasing air pollution and the related health risks due to automobile exhaust have shifted our attention towards green transportation. Recent decades have witnessed a revolution in portable energy-storage systems, mainly lithium-based energy-storage devices. However, the uneven distribution of global lithium reserves and its scarcity lead to huge price differences and geopolitical imbalances, and hence the research in energy-storage materials has shifted towards the development of cost-effective, abundant electrode materials. Here, NaCr(SO4)2, a transition metal-based polyanionic layered material with low cost and high stability during the charge/discharge process vs. Na, operating on the basis of the Cr3+/2+ redox couple, is presented. The test materials were characterized by techniques like XRD, FTIR, SEM, UV, XPS, TGA-DTA, and a detailed electrochemical analysis of the charge/discharge capacity of the materials is presented here. Here, the findings provide insights towards achieving a Cr3+/Cr2+ redox-couple-based sodium-ion battery with a specific capacity of 75 mA h g-1 and 150 mA h g-1 at operating voltages of 0.95 V vs. Na and 1.05 V vs. Li, respectively, with 100% coulombic efficiency. Cr2+ is a very special oxidation of Cr that cannot be obtained easily and CrTa2O6 is the only known oxide where Cr exists in the 2+ state. Here, a shift in the redox energy of the Cr3+/2+ couple was obtained due to its bonding with (SO4)2- polyanions in eldfellite that made the accessibility of Cr3+/2+ possible, resulting in the superior intercalation/deintercalation of Na and Li and the superior energy-storage capacity of the NaCr(SO4)2vs. Na/Li cell.

With an increase in waste generation, resource scarcity and deterioration of environment, circular economy is being given wide attention. Closed-loop supply chain can help to achieve the goals of circular economy through maximizing the use of materials. This study explores the critical success factors for closed-loop supply chain operations in Indian small and medium manufacturing enterprises. A survey-study approach is followed for the collection of data, and partial least squares approach is applied to examine the critical factors. The findings show ‘Green innovation’ to be the most critical factor, followed by ‘Management support and coordination’, ‘Design for recovery’ and ‘Managing product returns’. The study guides the managers in Indian small and medium manufacturing enterprises on critical factors for closed-loop supply chain operations. The study also extends the knowledge on closed-loop supply chain by identifying the critical factors in Indian small and medium manufacturing enterprises.

The synergistic effect of low-cost, noble metal-free, environmentally friendly, and highly electrochemically active electrocatalysts is often used for the hydrogen evolution reaction (HER). Herein, nanocomposites of polypyrrole (PPy) and NASICON-structured Na3Fe2(SO4)(2)(PO4) (NFS) for an efficient HER have been synthesized by simple chemical oxidative polymerization. The PPy/NFS-5% (optimized ratio) nanocomposite exhibits excellent HER activity in a highly acidic medium, having a considerably low onset overpotential of -26 mV along with a Tafel slope of 101 mV dec(-1) and a comparatively low overpotential of -206 mV corresponding to a current density of 10 mA/cm(2). The as-synthesized PPy/NFS-5% nanocomposite also offers long-term stability and excellent durability. The superior electrocatalytic activity of PPy/NFS-5% nanocomposite results from the synergism between the conductivity of the PPy matrix and the hydrophilicity of NFS, providing an abundant electrocatalytic active site. The as-synthesized PPy/NFS-5% nanocomposite may find potential in real-world hydrogen production.

The performance of a zero-shot sketch-based image retrieval (ZS-SBIR) task is primarily affected by two challenges. The substantial domain gap between image and sketch features needs to be bridged, while at the same time the side information has to be chosen tactfully. Existing literature has shown that varying the semantic side information greatly affects the performance of ZS-SBIR. To this end, we propose a novel graph transformer based zero-shot sketch-based image retrieval (GTZSR) framework for solving ZS-SBIR tasks which uses a novel graph transformer to preserve the topology of the classes in the semantic space and propagates the context-graph of the classes within the embedding features of the visual space. To bridge the domain gap between the visual features, we propose minimizing the Wasserstein distance between images and sketches in a learned domain-shared space. We also propose a novel compatibility loss that further aligns the two visual domains by bridging the domain gap of one class with respect to the domain gap of all other classes in the training set. Experimental results obtained on the extended Sketchy, TU-Berlin, and QuickDraw datasets exhibit sharp improvements over the existing state-of-the-art methods in both ZS-SBIR and generalized ZS-SBIR.

Purpose Digital transformation (DT) leverages digital technologies to change current processes and introduce new processes in any organisation’s business model, customer/user experience and operational processes (DT pillars). Artificial intelligence (AI) plays a significant role in achieving DT. As DT is touching each sphere of humanity, AI led DT is raising many fundamental questions. These questions raise concerns for the systems deployed, how they should behave, what risks they carry, the monitoring and evaluation control we have in hand, etc. These issues call for the need to integrate ethics in AI led DT. The purpose of this study is to develop an “AI led ethical digital transformation framework”. Design/methodology/approach Based on the literature survey, various existing business ethics decision-making models were synthesised. The authors mapped essential characteristics such as intensity and the individual, organisational and opportunity factors of ethics models with the proposed AI led ethical DT. The DT framework is evaluated using a thematic analysis of 23 expert interviews with relevant AI ethics personas from industry and society. The qualitative data of the interviews and opinion data has been analysed using MAXQDA software. Findings The authors have explored how AI can drive the ethical DT framework and have identified the core constituents of developing an AI led ethical DT framework. Backed by established ethical theories, the paper presents how DT pillars are related and sequenced to ethical factors. This research provides the potential to examine theoretically sequenced ethical factors with practical DT pillars. Originality/value The study establishes deduced and induced ethical value codes based on thematic analysis to develop guidelines for the pursuit of ethical DT. The authors identify four unique induced themes, namely, corporate social responsibility, perceived value, standard benchmarking and learning willingness. The comprehensive findings of this research, supported by a robust theoretical background, have substantial implications for academic research and corporate applicability. The proposed AI led ethical DT framework is unique and can be used for integrated social, technological and economic ethical research.

Internal friction is often sensitive to microstructural features. However, there is a clear absence of rational approach for decoupling internal friction spectra for diverse microstructural inputs. In this study, a robust multi-scale atomistic computational framework, combining atomistic kinetic Monte Carlo and molecular dynamics simulations, has been proposed. Predictions from our simulations were then compared with careful experiments on engineered microstructures in bcc steel. Specifically, theoretical contributions from interstitial solute type and concentration, crystallographic orientation, and residual stress (RS), were compared to actual experimental results. The atomistic computational framework successfully demonstrated that the overall internal friction response was composed, almost entirely, of Snoek relaxations from interstitial atoms. Ideal single-crystal simulations correctly predicted peak dissipation temperatures and Snoek peak height, $${\text{tan}}{\delta }_{\text{max}}$$ , when compared with available single-crystal experimental data. The simulations also captured the correct experimental trends with residual stress and crystallographic orientation in polycrystalline bcc steel. In particular, both RS and crystallographic orientation affected internal friction response by altering diffusion barriers for interstitial migration. Our study, thus, established that an integrated computational framework, supported with careful experiments, can be extremely effective in decoupling various microstructural inputs to complex experimental internal friction spectrum.

In the recent few years, with an increase in focus on sustainability, firms have been actively pursuing different strategies to contribute towards sustainability. Industry 4.0 (I4) technologies can help organizations to achieve superior environmental as well as economic performance. Through the lens of stakeholder theory (ST) and Schumpeterian view of competition (SCV), this paper examines whether stakeholder and competitive pressures towards sustainability stimulate organizations to implement I4 technologies and commensurate performance outcomes. The study further tests the mediating role of environmental commitment and green process innovation (GPI) on these relationships. The proposed hypotheses are examined using the survey data from 173 manufacturing firms in India by partial least squares (PLS) approach. Findings show that environmental commitment mediates the effect of stakeholder and competitive pressures on I4 technologies. Further, results also show that GPI mediates between I4 technologies and performance. The findings provide insights for managers on how they can best respond to stakeholder and competitive pressures on sustainability and contribute towards sustainable development.