Search Results for « Generative ML »

Generative ML models are increasingly popular in networking for tasks such as telemetry imputation, prediction, and synthetic trace generation. Despite their capabilities, they suffer from two shortcomings: \emph{(i)} their output is often visibly violating well-known networking rules, which undermines their trustworthiness; and \emph{(ii)} they are difficult to control, frequently requiring retraining even for minor changes. To address these limitations and unlock the benefits of generative models for networking, we propose a new paradigm for integrating explicit network knowledge, in the form of first-order logic rules, into ML models used for networking tasks. Rules capture well-known relationships among observed signals, e.g., that increased latency precedes packet loss. While the idea is conceptually straightforward, its realization is challenging: networking knowledge is rarely formalized into rules, and naively injecting rules into ML models often hampers their effectiveness. This paper introduces NetNomos, a multi-stage framework that \emph{(i)} learns rules directly from data (e.g., measurements); \emph{(ii)} filters them to select semantically meaningful ones; and \emph{(iii)} enforces them through collaborative generation between an ML model and a Satisfiability Modulo Theories (SMT) solver. %We evaluate NetNomos both component-wise and end-to-end across four diverse network datasets. We show that NetNomos learns diverse, meaningful rules from four real-world datasets and is 1.6--6.5$\times$ more scalable than DuoAI, a state-of-the-art (SOTA) rule-learning method. By enforcing these rules on a generic GPT-2 model, NetNomos achieves performance on par with or even surpassing specialized SOTA systems such as Zoom2Net and NetShare across three networking tasks: telemetry imputation, traffic forecasting, and synthetic data generation.

Calculation of binding energies for protein-ligand molecular systems requires accurate treatment of the electronic structure, a quantum chemistry problem that scales exponentially on classical hardware, while current quantum hardware remains too noisy for the required circuit depths. This report presents a hybrid quantum-classical workflow performed on the Fujitsu FX700 ideal state-vector simulator using QARP that addresses two structural inefficiencies in quantum-sampling-based diagonalization workflows. First, we integrate the Linear Scaling CNOT UCCSD (LCNot-UCCSD) ansatz into the QSCI framework, replacing the $\mathcal{O}(N^6)$ CCSD parameter initialization of the competing LUCJ ansatz approach with $\mathcal{O}(N^4)$ MP2-amplitude initialization. Second, we introduce QSCI-RBM, a variant that replaces the configuration recovery of the SQD framework with a Restricted Boltzmann Machine (RBM) acting as a compact generative subspace expansion model. Both are evaluated on eight different molecules in STO-3G across 14 controlled artificial error levels with 100 independent runs each, validated on potential energy surface scans of the N$_2$ molecule in cc-pVDZ, and embedded within DMET to treat the FDA-approved antiviral Amantadine (C$_{10}$H$_{17}$N, 11 DMET fragments) and the active region of the SARS-CoV-2 main protease complexed with its covalent inhibitor Carmofur (PDB: 7BUY, C$_{15}$H$_{28}$N$_4$O$_5$S, 10 fragments). To our knowledge, this is the first deployment of LCNot-UCCSD within QSCI on a quantum computing simulator, and the first DMET-QSCI(LCNot-UCCSD)-RBM application to an industry-relevant protein-ligand system. By utilizing a fraction of the classical computing resources required by the current state-of-the-art work by Cleveland Clinic, RIKEN, and IBM Quantum, this approach enables more efficient and economical drug discovery simulations for the industry.

Artificial Intelligence (AI), Machine Learning (ML), and Generative AI are transforming the financial services industry by enabling smarter decision-making, improved operational efficiency, and enhanced customer experiences. Financial institutions increasingly leverage these technologies across a wide spectrum of use cases, including fraud detection, risk management, regulatory compliance, customer personalization, algorithmic trading, and intelligent automation. ML models analyze vast volumes of structured and unstructured data to uncover patterns, predict risks, and optimize processes, while Generative AI extends these capabilities by synthesizing insights, automating content creation, and enabling conversational intelligence. In banking and capital markets, AI-driven solutions support credit scoring, anti-money laundering (AML) monitoring, and predictive analytics for market trends. Insurance firms use AI/ML for underwriting, claims automation, and fraud prevention, while wealth management platforms integrate AI-powered advisory services to deliver personalized investment strategies. Generative AI further enhances productivity through automated report generation, document summarization, and intelligent virtual assistants, reducing manual intervention in compliance and customer servicing workflows. Despite significant benefits, challenges remain in ensuring data privacy, model explainability, regulatory compliance, and governance. Financial institutions must adopt responsible AI frameworks to mitigate risks associated with bias, security, and transparency. Overall, the integration of AI/ML and Generative AI is reshaping financial ecosystems, enabling innovation, improving resilience, and driving competitive advantage in an increasingly digital and data-driven landscape. This paper explores the transformative impact of Artificial Intelligence (AI), Machine Learning (ML), and Generative AI (GenAI) on the financial services industry. It highlights real-world applications, benefits, challenges, and regulatory considerations, offering a roadmap for institutions aiming to harness these technologies for strategic advantage

To improve real-world applications of machine learning, experienced modelers develop intuition about their datasets, their models, and how the two interact. Manual inspection of raw data - of representative samples, of outliers, of misclassifications - is an essential tool in a) identifying and fixing problems in the data, b) generating new modeling hypotheses, and c) assigning or refining human-provided labels. However, manual data inspection is problematic for privacy sensitive datasets, such as those representing the behavior of real-world individuals. Furthermore, manual data inspection is impossible in the increasingly important setting of federated learning, where raw examples are stored at the edge and the modeler may only access aggregated outputs such as metrics or model parameters. This paper demonstrates that generative models - trained using federated methods and with formal differential privacy guarantees - can be used effectively to debug many commonly occurring data issues even when the data cannot be directly inspected. We explore these methods in applications to text with differentially private federated RNNs and to images using a novel algorithm for differentially private federated GANs.

This paper examines the evolution of cybersecurity assault prediction methods, which have evolved from basic Machine Learning (ML) methods to more intricate Generative AI methods. In the past, standard machine learning models such as Neural Networks, Decision Trees, and Random Forests were frequently employed to identify deviant behavior. The complexity of cyberthreats has increased over time. Generative AI technologies, such as GANs and large language models, are employed to anticipate and mitigate hazards prior to their occurrence. Modern cybersecurity systems are more effective at safeguarding against cyberattacks due to the implementation of new technologies that enhance the accuracy of predictions, automate hazard analysis, and enhance the detection of unusual behavior.

The automotive industry stands on the brink of a transformative era, driven by advancements in artificial intelligence (AI), machine learning (ML), and generative AI technologies.This paper explores how these cutting-edge technologies are reshaping automotive manufacturing, focusing on their integration into the design, production, and operation of next-generation autonomous vehicles.AI and ML are revolutionizing automotive manufacturing by enabling predictive maintenance, enhancing quality control, and optimizing supply chains through data-driven insights.Generative AI, in particular, is pushing the boundaries of design innovation, allowing for the creation of novel vehicle designs and features that were previously inconceivable.The convergence of these technologies promises not only to streamline manufacturing processes but also to accelerate the development of autonomous vehicles, making them safer, more efficient, and more adaptable to consumer needs.By examining current applications and future potentials, this paper highlights the significant impact of AI and ML on improving operational efficiencies and driving innovation in vehicle design and production.Furthermore, it addresses the challenges and ethical considerations associated with these technologies, including data privacy, cybersecurity, and the potential for job displacement.Through a comprehensive review of emerging trends and case studies, the paper provides a forward-looking perspective on how the integration of AI, ML, and generative AI will shape the future of automotive manufacturing and drive the industry towards a new paradigm of automation and intelligence.

The ongoing climate crisis, caused by the annual release of 37 billion metric tons of CO2 emissions, is putting pressure on the advancement of Carbon Capture and Storage (CCS) and Direct Air Capture (DAC) technologies. Metal–Organic Frameworks (MOFs), with their high surface areas and modular pore topologies, present a very attractive class of sorbents for CO2 capture; however, it currently remains computationally prohibitive to explore their extensive chemical design space. Herein, we provide a thorough evaluation of how machine learning (ML) (as an emerging technology) has played an increasing role in furthering our understanding of CO2 capture from MOFs. Through an organized investigation, we provide evaluations of the latest generation of models across four key areas: application at a process level, mechanistic interpretable modelling; physically relevant descriptors, and predictive performance metrics. Recent work with Machine Learning Interatomic Potentials (MLPs) shows that traditional assumptions about rigid frameworks are being challenged by the fact that diffusion properties and adsorption thermodynamics are heavily influenced by the flexibility of the framework. The use of physics-informed descriptor engineering yields R2 values of 0.81-0.97 across gas species and pressure regimes, while the generative nature of Deep Reinforcement Learning and transformer-based architectures has been shown to allow for the inverse design of frameworks with high affinities for gas species. The trend in this sector is moving towards optimization of multiple scales simultaneously and integrating processes to achieve an optimized property prediction. Current work with machine learning is focusing on using a combination of material properties and operational indicators (such as how much gas is recovered through pressure swing adsorption) to make predictions. As these techniques improve, there will be a similar need for a design that is both physically informed and understandable, thus allowing for a link between molecular discoveries and water-stable materials that have been experimentally verified and are suitable for use in commercial applications.

Alternative investments can deliver diversification and risk premia, but for LPs the most dangerous feature in stress periods is not mark-to-market valuation drawdowns, but "same-direction cash-flow squeeze": distributions fall sharply or even freeze, while capital calls become more concentrated and more frequent, making liquidity management passive. The latent-factor cash flow model proposed by Cao and van Beek introduces systematic latent factors and heavy-tailed idiosyncratic terms into the TA framework, and links call rates, distribution rates, and growth rates to macro variables, making stress testing estimable, transmissible, and reproducible. Building on this, this paper proposes Generative Scenario Modeling (GSM): using the latent-factor model as an interpretable backbone, and under the cash flow–NAV identity and business-boundary constraints, introducing conditional GANs/conditional VAEs to learn tail dependence, nonlinear regime shifts, and scenario-cluster structure, generating a stress-scenario library of quarterly net cash flows, NAV paths, valuation adjustments, and distribution changes, for shortfall distribution estimation, commitment pacing and rebalancing evaluation, and tiered trigger-based contingency planning. The paper concludes by engaging the discussion of "whether LPs should use AI to assist decision-making," emphasizing definition consistency, confidentiality and compliance, model governance, and human review, to avoid treating AI as an automatic adjudicator. Keywords: private markets; cash flow risk; stress testing; generative models; scenario library; LP liquidity management.