Numerical weather prediction (NWP) models remain the cornerstone of atmospheric sciences. Yet, deep learning (DL) is challenging this paradigm by its ability to capture intricate spatio-temporal patterns and deliver ultra-fast predictions. Analogous to the foundation models (e.g., ChatGPT) in natural language processing, foundation models in the weather/climate domain have also been developed. This paper reviews DL and foundation models for weather prediction by highlighting their strengths and limitations. In particular, we carefully examine them from the perspective of their training paradigms: deterministic predictive learning, probabilistic generative learning, and pre-training & fine-tuning. For each paradigm, we summarize the underlying model architectures, training methods, and respective features. To facilitate further study, we provide a curated repository featuring categorized papers, open-source code, and benchmark datasets. Finally, we discuss and suggest potential research directions across new tasks and models in weather data storage and management, and operational deployment, further inspiring innovations in this rapidly evolving field. GitHub: https://github.com/JimengShi/DL-Foundation-Models-Weather.
论文检索
输入标题、作者或关键词，从 1,767 篇学术成果中精准定位
Multivariate Time Series Forecasting (MTSF) plays a crucial role across diverse fields, ranging from economics to energy to traffic. In recent years, deep learning has demonstrated outstanding performance in MTSF tasks. In MTSF, modeling the correlations among different channels is critical, as leveraging information from other related channels can significantly improve the prediction accuracy of a specific channel. This study systematically reviews the channel modeling strategies for time series and proposes a taxonomy organized into three hierarchical levels: the strategy perspective, the mechanism perspective, and the characteristic perspective. On this basis, we provide a structured analysis of these methods and conduct an in-depth examination of the advantages and limitations of different channel strategies. Finally, we summarize and discuss some future research directions to provide useful research guidance. Moreover, we maintain an up-to-date GitHub repository which includes all the papers discussed in the survey.
Autonomous AI agents on embedded platforms require real-time, risk-aware scheduling under resource and thermal constraints. Classical heuristics struggle with workload irregularity, tabular regressors discard structural information, and model-free reinforcement learning (RL) risks overheating. We introduce GraphPerf-RT, an AI technology achieving deep learning accuracy at heuristic speeds (2-7ms). GraphPerf-RT is, to our knowledge, the first graph-grounded infrastructure unifying task DAG topology, CFG-derived code semantics, and runtime context (per-core DVFS, thermal state, utilization) in a heterogeneous graph with typed edges encoding precedence, placement, and contention. The architecture supports multi-task evidential heads with Normal-Inverse-Gamma uncertainty; we validate on makespan prediction for risk-aware scheduling. Experiments on three ARM platforms (Jetson TX2, Orin NX, RUBIK Pi) achieve R^2 = 0.81 on log-transformed makespan with Spearman rho = 0.95 and conservative uncertainty calibration (PICP = 99.9% at 95% confidence). Integration with four RL methods demonstrates that multi-agent model-based RL with GraphPerf-RT as the world model achieves 66% makespan reduction and 82% energy reduction versus model-free baselines, with zero thermal violations.
Chlor-alkali production is a large-scale industrial process whose operating conditions and equipment states evolve over time. Its process optimization requires ongoing trade-offs among conflicting objectives such as product yield, energy consumption, and equipment life. Existing optimization approaches are typically static and must be re-optimized after environmental changes, limiting their real-world applicability. In this work, we model the problem as a dynamic multi-objective sequential decision-making problem that continuously tracks a time-varying Pareto set under changing conditions. We propose MORL-CA, a multi-objective reinforcement learning framework that integrates offline pretraining on historical data with constrained online policy refinement. MORL-CA introduces a state-aware adaptive objective weighting mechanism within a multi-critic actor-critic architecture, enabling localized Pareto-improving policy updates while satisfying operational and safety constraints. Extensive experiments in an environment conducted from real chlor-alkali data demonstrate that MORL-CA achieves superior Pareto solution quality and smoother adaptation to dynamics compared with state-of-the-art multi-objective optimizers and MORL baselines.
Dynamically Reconfigurable Battery (DRB) systems employ power electronic switches to create dynamic topologies. They enable effective management of cell difference through real-time adjustment of cell connections. However, existing DRB control methods struggle to learn effective strategies due to sparse rewards, which arise from blind exploration in large topological action spaces and complex operational constraints. This leads to insufficient policy learning, making safety and balancing performance difficult to ensure in practical applications. To this end, we propose a Subgraph-Augmented Hierarchical Reinforcement Learning (SAHRL) framework. By combining hierarchical policies with topological structural knowledge, SAHRL effectively accelerates policy exploration and mitigates reward sparsity. Specifically, the high-level policy determines the strategic direction, while the subgraph-augmented low-level policy refines actions to meet operational constraints. The topological structural knowledge, extracted in the form of subgraphs and incorporated as an inductive bias, helps the agent focus on meaningful action patterns and reduce invalid exploration in the large action space. Extensive simulations and real-world experiments show that SAHRL achieves safe and efficient balancing. Notably, it increases the energy release by 10.56% compared to conventional methods in real-world applications.
With the rapid expansion of photovoltaic (PV) power generation worldwide, PV systems have become key to global energy construction. Accurate PV forecasting is essential for safe grid operation and renewable energy integration. However, most existing models rely heavily on site-specific historical data and perform poorly when deployed in cold-start scenarios of newly built power plants. We propose PhysTrans, a physics-aware transferable framework for cold-start PV forecasting. Firstly, we design a physics-constrained residual network that utilizes a clear-sky module for better physical consistency. Furthermore, we propose a dynamic cloud cropping method to obtain the cloud information of shaded PV stations by fitting the angle of the sun offsets. To fuse the asymmetric data, a query-based asymmetric fusion mechanism is introduced to achieve high-precision alignment of multi-modal data. We conduct experiments on global datasets, and the results show that the PhysTrans outperforms state-of-the-art models with a 13.2% decrease in MAE in the single-site task, and also outperforms existing migration models with an average decrease in MAE of 12.7% in the cross-sites task. Our work advances reliable and transferable PV forecasting for early-stage grid integration and contributes to SDG 7 (Affordable and Clean Energy) and SDG 13 (Climate Action), in line with the Leave No One Behind principle.
Along with the rapid development of new urban mobility options like ride-sharing over the past decade, on-demand micro-transit services stand out as a middle ground, bridging the gap between fixed-line mass transit and single-request ride-hailing, balancing ridership maximization and travel time minimization. Micro-transit adoption can have significant social impact. It improves urban sustainability, through lower energy consumption and reduced emissions, while enhancing equitable mobility access for disadvantaged communities, thanks to its lower vehicle miles per passenger, flexible schedules, and affordable pricing. However, effective operation of micro-transit services requires planning geo-fenced zones in advance, which involves solving a challenging combinatorial optimization problem. Existing approaches enumerate candidate zones first and selects a fixed number of optimal zones in the second step. In this paper, we generalize the Micro-Transit Zoning Problem (MZP) to allow a global budget rather than imposing a size limit for candidate zones. We also design a Column Generation (CG) framework to solve the problem exactly and several pricing heuristics to accelerate computation. Extensive numerical experiments across major U.S. cities demonstrate that our approach produces higher-quality solutions more efficiently and scales better in the generalized setting.
Inertial Confinement Fusion (ICF) holds transformative promise for sustainable, near-limitless clean energy, yet remains constrained by prohibitively high costs and limited experimental opportunities. This paper presents Human-in-the-Loop Meta Bayesian Optimization (HL-MBO), a framework that integrates expert knowledge with few-shot, uncertainty-aware machine learning to accelerate discovery in data-scarce, high-stakes scientific domains. HL-MBO introduces a meta-learned surrogate model with an expert-informed acquisition function to recommend candidate experiments. To foster trust and enable informed decisions, HL-MBO also provides interpretable explanations of its suggestions. We show that HL-MBO outperforms current BO methods for ICF energy yield optimization, as well as benchmarks in molecular optimization and critical-temperature maximization for superconducting materials.
The robustness of autonomous vehicles such as drones and Unmanned Surface Vehicles (USV) is crucial when facing unknown and complex marine environments, especially when heteroscedastic observational noise poses significant challenges to sensor-based navigation tasks. Recently, Distributional Reinforcement Learning (DistRL) has shown promising results in some challenging autonomous navigation tasks without prior environmental information. However, these methods overlook situations where noise patterns vary across different environmental conditions, hindering safe navigation and disrupting the learning of value functions. To address the problem, we propose DRIQN to integrate Distributionally Robust Optimization (DRO) with implicit quantile networks to optimize worst-case performance under natural environmental conditions. Leveraging explicit subgroup modeling in the replay buffer, DRIQN incorporates heterogeneous noise sources and target robustness-critical scenarios. Experimental results based on the risk-sensitive environment demonstrate that DRIQN significantly outperforms state-of-the-art meth- ods, achieving +13.51% success rate, -12.28% collision rate and +35.46% for time saving, +27.99% for energy saving, compared with the runner-up.
Contrastive vision–language pretraining models such as CLIP align images and text in a shared embedding space but do not explicitly model or evaluate the hierarchical semantics common in medical image interpretation. We propose HCE-CLIP (Hierarchical Conditional Energy CLIP), a vision–language pretraining framework that formulates medical image–text alignment as a hierarchical label-conditional energy modeling problem. HCE-CLIP encodes an image series using transformer-based aggregation and aligns it with free-text reports and structured label state prompts across multiple semantic levels. At each level, conditional energy functions favor clinically consistent label states while suppressing contradictory alternatives, enabling uncertainty-aware inference. To assess semantic coherence, we introduce a hierarchical contradiction-based metric that quantifies logical inconsistencies between fine-grained disease predictions and higher-level clinical summaries. Experiments on MIMIC-CXR and other public benchmarks show that HCE-CLIP outperforms existing medical vision–language pretraining methods in seen-label, zero-shot and linear-probe settings, while producing substantially fewer hierarchical contradictions.
Active learning aims to maximize model performance with minimal annotation costs by selecting the most informative samples from large unlabeled pools, which often face a budget dilemma: uncertainty-based methods induce redundancy under low budgets, while representativeness-based methods struggle to mine challenging samples under high budgets. Although some heuristic parameter interpolation schemes attempt to bridge this gap, such strategies suffer from a misalignment between theoretical assumptions and real-world distributions, failing to achieve a good exploration-exploitation balance across all scenarios. In this paper, we propose a general active learning framework based on Distribution-Aware Energy Minimization, which reformulates sample selection as minimizing the energy function for the distributional discrepancy between the selected subset and the global uncertainty field. This physical-inspired perspective naturally derives a Hamiltonian comprising attractive terms and repulsive terms, mathematically achieving an intrinsic and dynamic balance between uncertainty and representativeness. Furthermore, we transform the optimization objective to the ground-state search of an Ising Model, enabling efficient solutions via Coherent Ising Machine. Extensive numerical experiments show that our method outperforms prior state-of-the-art methods across multi-budget regimes on several benchmark datasets. Validation on a real quantum hardware also demonstrates the potential for quantum computer in future large-scale selection tasks.
The deployment of edge servers plays a crucial role in supporting large-scale edge computing systems, where multiple conflicting objectives—such as latency, energy consumption, load balancing, and service reliability—must be jointly optimized in complex, dynamic environments. Existing solutions often struggle to scale effectively or to balance these objectives in a unified learning framework. In this paper, we propose GMM-TDQN, a two-stage multi-objective reinforcement learning framework for large-scale edge server deployment. The first stage employs a Gaussian Mixture Model (GMM) to capture spatial and workload heterogeneity, enabling an efficient reduction of the deployment search space. Building upon this structured initialization, the second stage formulates the deployment problem as a sequential decision-making task and adopts a Transformer-enhanced Deep Q-Network (TDQN) to learn adaptive deployment policies that balance multiple objectives. Extensive experiments on real-world datasets demonstrate that GMM-TDQN consistently outperforms state-of-the-art methods, achieving reductions of 29.18% in average latency and 17.55% in energy consumption, while improving load balancing by 27.50% and system reliability by 32.55%. These results validate the effectiveness and scalability of the proposed framework for multi-objective edge server deployment.
Time series forecasting plays a vital role in supporting decision-making across a wide range of critical applications, including energy, healthcare, and finance. Despite recent advances, forecasting accuracy remains limited due to the challenge of integrating historical numerical sequences with contextual features, which often comprise unstructured textual data. To address this challenge, we propose TokenCast, a large language model (LLM) driven framework that leverages language-based symbolic representations as a unified intermediary for context-aware time series forecasting. Specifically, TokenCast employs a discrete tokenizer to transform continuous numerical sequences into temporal tokens, enabling structural alignment with language-based inputs. To effectively bridge the semantic gap between modalities, both temporal and contextual tokens are embedded into a shared representation space via a pre-trained LLM, further optimized with generative objectives. Building upon this unified semantic space, the aligned LLM is subsequently fine-tuned in a supervised manner to predict future temporal tokens, which are then decoded back into the original numerical space. Extensive experiments on real-world datasets demonstrate the effectiveness of our framework and highlight its potential as a generative framework for context-aware time series forecasting. The code is available at https://github.com/Xiaoyu-Tao/TokenCast.
Parallel tempering, or the replica exchange method, is a Markov Chain Monte Carlo (MCMC) sampling technique for finding low-energy states in complex landscapes by executing multiple processes at different temperatures and allowing for states to migrate between parallel processes based on the Metropolis criterion. Despite a growing interest in the technique as a randomized search heuristic, it is not clear when and why parallel tempering is effective. We conduct a first runtime analysis of the parallel tempering algorithm on a simple quadratic unconstrained binary optimization problem that induces a rugged energy landscape with tunable parameters. We prove that a single-temperature Metropolis process fails to efficiently optimize the problem at any temperature. In sharp contrast, a two-state parallel tempering approach using a low and a high temperature typically solves the problem in O(n^2 log n) iterations. The low-temperature process ensures stability, as it is unlikely to accept worsenings, whereas the high-temperature process facilitates exploration by traversing energy barriers. Improvements are discovered through exploration in the high-temperature process and transferred to the low-temperature process, where they are permanently retained. This effective interplay between processes demonstrates the power of parallel tempering and lends credence to its design.
Large language models (LLMs) dominate both everyday and specialized applications, but their high computational demand, energy consumption, and privacy risks are increasingly critiqued. Small language models (SLMs) mitigate these drawbacks and are gaining momentum in scenarios where full LLM capabilities are not required, such as agents, industrial systems, or edge devices. Nevertheless, a systematic comparison of model capabilities, energy usage, and scaling behavior has not been conducted yet. We evaluate 70+ SLMs from 2023–2025 on five task-specific benchmarks and compare them with two popular LLMs, revealing key trade-offs between energy, performance, and model selection. Our findings challenge common assumptions: First, smaller models are not automatically more efficient, and energy increases do not guarantee performance gains. Second, newer SLMs show clear improvements in performance–energy trade-offs, though the progress begins to plateau. Last, the efficiency landscape forms a clear Pareto frontier: initial energy increases yield substantial gains, but the last percentage points of performance need orders of magnitude more energy. These results highlight diminishing returns of scaling and emphasize the need for informed, task-aware model selection rather than size-driven choices.
Deep convolutional neural networks achieve remarkable performance by exhaustively processing dense spatial feature maps, yet this brute-force strategy introduces significant computational redundancy and encourages reliance on spurious background correlations. As a result, modern vision models remain brittle and difficult to interpret. We propose Energy-Regularized Spatial Masking (ERSM), a novel framework that reformulates feature selection as a differentiable energy minimization problem. By embedding a lightweight Energy-Mask Layer inside standard convolutional backbones, each visual token is assigned a scalar energy composed of two competing forces: an intrinsic Unary importance cost and a Pairwise spatial coherence penalty. Unlike prior pruning methods that enforce rigid sparsity budgets or rely on heuristic importance scores, ERSM allows the network to autonomously discover an optimal information-density equilibrium tailored to each input. We validate ERSM on convolutional architectures and demonstrate that it produces emergent sparsity, improved robustness to structured occlusion, and highly interpretable spatial masks, while preserving classification accuracy. Furthermore, we show that the learned energy ranking significantly outperforms magnitude-based pruning in deletion-based robustness tests, revealing ERSM as an intrinsic denoising mechanism that isolates semantic object regions without pixel-level supervision. Code is available at https://github.com/Tom-Dvk/ERSM.
4DVarGen: A 4D Variational-Inspired Generative Model for Eddy-Resolving Surface Ocean Reconstruction
Sea surface variable reconstruction from sparse observations is a key ocean-science challenge. Traditional methods, such as the four-dimensional variational (4DVar) approach, rely on numerical models for background information, leading to high computational costs. Deep learning methods are more efficient but often fail to capture eddy dynamics, resulting in limited effective resolution. We propose 4DVarGen, a 4DVar-inspired generative framework for reconstructing sea surface variable fields at eddy-resolving scales from sparse remote-sensing observations. 4DVarGen establishes a mathematical equivalence between 4DVar and an observation-guided denoising process. Its key innovation is injecting the observation-likelihood gradient into denoising iterations, driving the generated trajectories to evolve in the direction of minimizing the 4DVar objective function toward a maximum a posteriori solution. Spatiotemporal priors learned by a diffusion model serve as background information, reducing computational costs and mitigating the adverse effects of Gaussian assumptions. Experiments show that 4DVarGen effectively leverages the temporal evolution patterns of sea surface temperature (SST) and sea surface height (SSH), as well as their dynamical mappings learned by the diffusion model, leading to improved reconstruction accuracy and effective resolution. Our model, pretrained on GLORYS12V1 reanalysis data, generates sea surface variable fields guided by real observations, achieving accuracy and effective resolution improvements of 18% and 58%, respectively, compared to GLORYS12V1. This study offers a novel framework for reconstructing Earth system states from sparse observations.
Agents in the real world must often balance multiple objectives, such as speed, stability, and energy efficiency in continuous control. To account for changing conditions and preferences, an agent must ideally learn a Pareto frontier of policies representing multiple optimal trade-offs. Recent advances in multi-policy multi-objective reinforcement learning (MORL) enable learning a Pareto front directly, but require full multi-objective consideration from the start of training. In practice, multi-objective preferences may arise after a policy has already been trained on a single specialised objective. Existing MORL methods cannot leverage such a pre-trained ‘specialist’ to learn Pareto fronts and avoid incurring the sample costs of retraining. We introduce Mixed Advantage Pareto Extraction (MAPEX), an offline MORL method that constructs a frontier of policies by reusing pre-trained specialist policies, critics, and replay buffers. MAPEX combines evaluations from specialist critics into a mixed advantage signal, and weights a behaviour cloning loss with it to train new policies that balance multiple objectives. MAPEX’s post hoc Pareto front extraction preserves the simplicity of single-objective off-policy RL, and avoids retrofitting these algorithms into complex MORL frameworks. We formally describe the MAPEX procedure and evaluate MAPEX on five multi-objective MuJoCo environments. Given the same starting policies, MAPEX produces comparable fronts at 0.001% the sample cost of established baselines.
Designing enzymes with substrate-binding pockets is a critical challenge in protein engineering, as catalytic activity depends on the precise interaction between pockets and substrates. Currently, generative models dominate functional protein design but cannot model pocket-substrate interactions, which limits enzyme generation with precise catalytic environments. To address this issue, we propose EnzyPGM, a unified framework that jointly generates enzymes and substrate-binding pockets conditioned on functional priors and substrates, with a particular focus on learning accurate pocket–substrate interactions. At its core, EnzyPGM includes two main modules: a Residue-atom Bi-scale Attention (RBA) that jointly models intra-residue dependencies and fine-grained interactions between pocket residues and substrate atoms, and a Residue Function Fusion (RFF) that incorporates enzyme function priors into residue representations. Also, we curate EnzyPock, an enzyme–pocket dataset comprising 84,336 enzyme–substrate pairs across 1,036 four-level enzyme families. Extensive experiments demonstrate that EnzyPGM achieves state-of-the-art performance on EnzyPock. Notably, EnzyPGM reduces the average binding energy by 0.47 kcal/mol over EnzyGen, showing its superior performance on substrate-specific enzyme design. The code is available at https://github.com/John-Lin98/EnzyPGM.
Time series forecasting (TSF) plays a vital role across various domains such as finance, energy, healthcare, and meteorology. Currently, most deep learning based TSF methods typically operate with a fixed lookback window. This approach comes from the high compute and memory costs of long contexts, as well as the standard practice of using sliding windows. This creates a trade-off. Making the window larger reduces the number of training samples, which can harm stability and generalization. However, keeping the window small prevents the model from using long history during inference. We propose an inference-only streaming autoregressive framework that replaces repeated full-context recomputation with a one-time context warmup and incremental decoding, enabling efficient long-history forecasting without retraining. While straightforward caching attentions is brittle for time series due to distribution shifts and noisy or redundant histories, we address these issues with cache-consistent normalization and selective memory under a fixed cache budget. Across diverse benchmarks, our approach substantially reduces inference latency with no or marginal accuracy loss, and often improves performance when longer lookbacks are beneficial.