arXiv subjects

Dazhi Yang

Publications and source records attributed to Dazhi Yang.

7 recordsLinked to original sources

arXiv2026

Activation Concentration: Characterizing Column-Level Output Sparsity Across Diffusion Model Architectures

Recent diffusion accelerators exploit activation sparsity by skipping near-zero GELU outputs, reporting 52--85% element-level sparsity. However, systolic-array hardware processes activations at column granularity, where a single non-zero element forces the entire column to be computed. We present the first systematic column-level sparsity characterization across seven diffusion workloads spanning three workload groups and four modalities. Our measurements reveal that element-level sparsity overstates hardware-exploitable sparsity by up to 78 percentage points and exposes a three-way taxonomy. UNet+transformer workloads exhibit activation concentration with workload-dependent cycle reductions up to 30.6%. Pure-transformer DiT shows dispersion, yielding 12.4%. Motion/dance transformer workloads range from modest reductions to 50.8% for MLD, driven by its extreme token dimension and expansion ratio. Cycle-level simulation on a GDDR6-based accelerator confirms that memory stalls account for up to 84--89% of total cycles and that layout sensitivity tracks the profiling-based taxonomy. A full accuracy sweep across five thresholds reveals that UNet+transformer workloads degrade gracefully, while motion models exhibit an accuracy cliff between the primary operating point and the next threshold. Our characterization shows that workload group and model dimensions jointly determine whether column-level memory layout optimization is beneficial, and element-level sparsity alone is insufficient for that prediction.

cs.AR↗

arXiv2025

A review of cultural heritage inspection: Toward terahertz from mid-infrared region

This review explores non-invasive imaging (NII) methods covering the mid- and far-infrared to the terahertz spectral regions (up to approximately 1000 um) for the detection and analysis of cultural heritage artifacts. In the thermal infrared domain, where radiation follows Planck's law, the self-emission of materials reveals intrinsic properties and internal degradation. By contrast, in the near-infrared range, external illumination enhances surface details and pigment differentiation. Far-infrared and terahertz techniques, operating in both transmission and reflection modes, provide complementary insights by penetrating surface layers to uncover subsurface structures and concealed features. Integrating visible and infrared imaging further enriches diagnostic capabilities by correlating conventional visual assessments with spectral information. Beyond reviewing the wide applications of these NII techniques in cultural heritage research, this work also summarizes recent advances in signal processing, encompassing both hardware and software developments. In particular, deep learning has revolutionized the field by enabling automated classification, feature extraction, defect detection, and super-resolution imaging. Through supervised and unsupervised learning strategies, neural networks can reliably identify subtle anomalies and material variations indicative of past restorations or early stages of deterioration. In conclusion, the convergence of advanced spectral imaging, sophisticated signal processing, and deep neural networks offers a transformative pathway toward more accurate, efficient, and data-driven cultural heritage analysis, ultimately supporting more informed conservation and restoration decisions.

physics.optics↗

arXiv2025

Post-processing of ensemble photovoltaic power forecasts with distributional and quantile regression methods

Accurate and reliable forecasting of photovoltaic (PV) power generation is crucial for grid operations, electricity markets, and energy planning, as solar systems now contribute a significant share of the electricity supply in many countries. PV power forecasts are often generated by converting forecasts of relevant weather variables to power predictions via a model chain. The use of ensemble simulations from numerical weather prediction models results in probabilistic PV forecasts in the form of a forecast ensemble. However, weather forecasts often exhibit systematic errors that propagate through the model chain, leading to biased and/or uncalibrated PV power predictions. These deficiencies can be mitigated by statistical post-processing. Using PV production data and corresponding short-term PV power ensemble forecasts at seven utility-scale PV plants in Hungary, we systematically evaluate and compare seven state-of-the-art methods for post-processing PV power forecasts. These include both parametric and non-parametric techniques, as well as statistical and machine learning-based approaches. Our results show that compared to the raw PV power ensemble, any form of statistical post-processing significantly improves the predictive performance. Non-parametric methods outperform parametric models, with advanced nonlinear quantile regression models showing the best results. Furthermore, machine learning-based approaches surpass their traditional statistical counterparts.

stat.AP↗

arXiv2024

Noise-Aware Bayesian Optimization Approach for Capacity Planning of the Distributed Energy Resources in an Active Distribution Network

The growing penetration of renewable energy sources (RESs) in active distribution networks (ADNs) leads to complex and uncertain operation scenarios, resulting in significant deviations and risks for the ADN operation. In this study, a collaborative capacity planning of the distributed energy resources in an ADN is proposed to enhance the RES accommodation capability. The variability of RESs, characteristics of adjustable demand response resources, ADN bi-directional power flow, and security operation limitations are considered in the proposed model. To address the noise term caused by the inevitable deviation between the operation simulation and real-world environments, an improved noise-aware Bayesian optimization algorithm with the probabilistic surrogate model is proposed to overcome the interference from the environmental noise and sample-efficiently optimize the capacity planning model under noisy circumstances. Numerical simulation results verify the superiority of the proposed approach in coping with environmental noise and achieving lower annual cost and higher computation efficiency.

cs.NE↗

arXiv2024

Potential root mean square error skill score

Consistency, in a narrow sense, denotes the alignment between the forecast-optimization strategy and the verification directive. The current recommended deterministic solar forecast verification practice is to report the skill score based on root mean square error (RMSE), which would violate the notion of consistency if the forecasts are optimized under another strategy such as minimizing the mean absolute error (MAE). This paper overcomes such difficulty by proposing a so-called "potential RMSE skill score," which depends only on: (1) the crosscorrelation between forecasts and observations, and (2) the autocorrelation of observations. While greatly simplifying the calculation, the new skill score does not discriminate inconsistent forecasts as much, e.g., even MAE-optimized forecasts can attain a high RMSE skill score.

stat.ME↗

arXiv2022

Site adaptation with machine learning for a Northern Europe gridded solar radiation product

Gridded global horizontal irradiance (GHI) databases are fundamental for analysing solar energy applications' technical and economic aspects, particularly photovoltaic applications. Today, there exist numerous gridded GHI databases whose quality has been thoroughly validated against ground-based irradiance measurements. Nonetheless, databases that generate data at latitudes above 65$^{\circ}$ are few, and those available gridded irradiance products, which are either reanalysis or based on polar orbiters, such as ERA5, COSMO-REA6, or CM SAF CLARA-A2, generally have lower quality or a coarser time resolution than those gridded irradiance products based on geostationary satellites. Among the high-latitude gridded GHI databases, the STRÅNG model developed by the Swedish Meteorological and Hydrological Institute (SMHI) is likely the most accurate one, providing data across Sweden. To further enhance the product quality, the calibration technique called "site adaptation" is herein used to improve the STRÅNG dataset, which seeks to adjust a long period of low-quality gridded irradiance estimates based on a short period of high-quality irradiance measurements. This study, differing from the conventional statistical approaches, adopts machine learning for site adaptation. Nine machine-learning algorithms have been analysed and compared with conventional statistical ones to identify Sweden's most favourable technique for site adaptation. Three weather stations of SMHI are used for training and validation. The results show that, due to the spatio-temporal heterogeneity in model performance, no universal model can be identified, which suggests that site adaptation is a location-dependent procedure.

physics.app-ph↗

arXiv2022

Benchmarks for Solar Radiation Time Series Forecasting

With an ever-increasing share of intermittent renewable energy in the world's energy mix,there is an increasing need for advanced solar power forecasting models to optimize the operation and control of solar power plants. In order to justify the need for more elaborate forecast modeling, one must compare the performance of advanced models with naive reference methods. On this point, a rigorous formalism using statistical tools, variational calculation and quantification of noise in the measurement is studied and five naive reference forecasting methods are considered, among which there is a newly proposed approach called ARTU (a particular autoregressive model of order two). These methods do not require any training phase nor demand any (or almost no) historical data. Additionally, motivated by the well-known benefits of ensemble forecasting, a combination of these models is considered, and then validated using data from multiple sites with diverse climatological characteristics, based on various error metrics, among which some are rarely used in the field of solar energy. The most appropriate benchmarking method depends on the salient features of the variable being forecast (e.g., seasonality, cyclicity, or conditional heteoroscedasity) as well as the forecast horizon. Hence, to ensure a fair benchmarking, forecasters should endeavor to discover the most appropriate naive reference method for their setup by testing all available options. Among the methods proposed in this paper, the combination and ARTU statistically offer the best results for the proposed study conditions.

stat.AP↗