Artelys (France)

Abstract This study explores the implications of technology availability constraints on the decarbonisation pathways of the EU power sector, drawing on scenario results from five European energy supply models: MEESA, LIMES, ENERTILE, ACSG, and OSeMBE. The analysis evaluates how limiting the deployment of key low-carbon generation technologies, namely carbon capture and storage (CCS), nuclear power, bioenergy, solar photovoltaics, and wind energy, affects the electricity generation mix, emissions reduction, investment needs, and power system costs by 2030 and 2050. Results within the model ensemble indicate that Variable Renewable Energy sources (VRE), wind and solar, are indispensable for deep decarbonisation. Constraints on solar or wind power substantially increase electricity generation costs and require major shifts in technology portfolios, often resulting in greater reliance on remaining renewable options or fossil fuel generation with CCS. The unavailability of CCS leads to higher system emissions and increased investment in renewables and storage. In contrast, removing nuclear or bioenergy has a more moderate impact, though some regional effects are significant. All models show that achieving ambitious emission reductions in the power sector remains technically feasible under individual technology constraints, but the mitigation effort shifts across generation technologies, and system costs rise considerably in low VRE futures. Policy implications include the need for robust support for wind and solar deployment, cross-border system integration, flexible technologies, and backup capacity. The findings underscore the value of a diversified technology portfolio, strategic infrastructure investments, and EU-level coordination to preserve cost efficiency and ensure stable power system performance under uncertainty in future technology availability.

Neural metrics for machine translation (MT) evaluation have become increasingly prominent due to their superior correlation with human judgments compared to traditional lexical metrics. Researchers have therefore utilized neural metrics through quality-informed decoding strategies, achieving better results than likelihood-based methods. With the rise of Large Language Models (LLMs), preference-based alignment techniques have gained attention for their potential to enhance translation quality by optimizing model weights directly on preferences induced by quality estimators. This study focuses on Contrastive Preference Optimization (CPO) and conducts extensive experiments to evaluate the impact of preference-based alignment on translation quality. Our findings indicate that while CPO consistently outperforms Supervised Fine-Tuning (SFT) on high-quality data with regard to the alignment metric, it may lead to instability across downstream evaluation metrics, particularly between neural and lexical ones. Additionally, we demonstrate that relying solely on the base model for generating candidate translations achieves performance comparable to using multiple external systems, while ensuring better consistency across downstream metrics.

Abstract We are interested in the design of robust (or resilient) capacitated rooted Steiner networks in the case of terminals with uniform demands. Formally, we are given a graph, capacity, and cost functions on the edges, a root, a subset of vertices calledterminals, and a bound on the number of possible edge failures. We first study the problem where and the network that we want to design must be a tree covering the root and the terminals: we give complexity results and propose models to optimize both the cost of the tree and the number of terminals disconnected from the root in the worst case of an edge failure, while respecting the capacity constraints on the edges. Secondly, we consider the problem of computing a minimum‐cost survivable network, that is, a network that covers the root and terminals even after the removal of any edges, while still respecting the capacity constraints on the edges. We also consider the possibility of protecting a given number of edges. We propose three different formulations: a bilevel formulation (with an attacker and a defender), a cutset‐based formulation and a flow‐based one. We compare the formulations from a theoretical point of view, and we propose algorithms to solve them and compare their efficiency in practice.

We prove the almost-sure convergence of a class of sampling-based nested decomposition algorithms for multistage stochastic convex programs in which the stage costs are general convex functions of the decisions and uncertainty is modelled by a scenario tree. As special cases, our results imply the almost-sure convergence of stochastic dual dynamic programming, cutting-plane and partial-sampling (CUPPS) algorithm, and dynamic outer-approximation sampling algorithms when applied to problems with general convex cost functions.

In the standard version of the UCT algorithm, in the case of a continuous set of decisions, the exploration of new decisions is done through blind search. This can lead to very inefficient exploration, particularly in the case of large dimension problems, which often happens in energy management problems, for instance. In an attempt to use the information gathered through past simulations to better explore new decisions, we propose a method named Blind Value (BV). It only requires the access to a function that randomly draws feasible decisions. We also implement it and compare it to the original version of continuous UCT. Our results show that it gives a significant increase in convergence speed, in dimensions 12 and 80.

Dynamic functional imaging with slow rotating SPECT systems (dSPECT) is a relatively new development in emission tomography introduced in (M.N. Limber et al., ibid., vol. 42, p. 1249-56, 1995), and further developed in a number of contributions. The present paper aims at a wider distribution of the method and demonstrates its applicability in the clinical setting. The authors present a patient study, and a simulated case study to test the performance of dSPECT, and to show its abilities in terms of spatial and temporal resolution of the reconstructed dynamic images.

Lagrangian relaxation is often an efficient tool to solve (large-scale) optimization problems, even nonconvex. However it introduces a duality gap, which should be small for the method to be really efficient. Here we make a geometric study of the duality gap. Given a nonconvex problem, we formulate in a first part a convex problem having the same dual. This formulation involves a convexification in the product of the three spaces containing respectively the variables, the objective and the constraints. We apply our results to several relaxation schemes, especially one called “Lagrangean decomposition” in the combinatorial-optimization community, or “operator splitting” elsewhere. We also study a specific application, highly nonlinear: the unit-commitment problem.