Real-world optimization problems often involve complex physical or industrial process constraints that must be met to ensure feasible solutions. Constrained Bayesian Optimization (CBO) provides an efficient framework for such optimization tasks that relies on uncertainty-aware surrogate models to approximate both the objective function and the constraints. Existing approaches for modeling the constraints often assume homoscedastic Gaussian noise and utilize Gaussian processes (GPs), which can lead to miscalibrated uncertainty estimates, particularly in the presence of model mismatch. To address these challenges, we propose a novel framework that improves both the calibration and adaptivity of constraint modeling. First, we use conformal prediction (CP) to construct prediction intervals with guaranteed calibration, independent of the noise distribution. Next, we improve local adaptivity by modeling residuals with a nonparametric kernel density estimator, enabling the intervals to adjust dynamically to heteroscedastic and non-Gaussian noise. Finally, we use a distance-based uncertainty heuristic to detect distribution shifts, improving robustness in regions with sparse or out-of-distribution samples. Our framework is model-agnostic and can be applied to any prediction model. We validate our method on a synthetic CBO task and on real-world data by modeling constraints for carbite-free bainitic steel optimization [1]. Our results show that standard GP models often produce miscalibrated uncertainty estimates, while our method yields improved calibration, and reduced prediction intervals.
Phase retrieval is a fundamental inverse problem that arises in many scientific and engineering disciplines, where the goal is to reconstruct a signal from intensity-only measurements. In this work, we develop prNet, a novel phase retrieval framework that stochastically refines initial reconstructions with learned denoising and model-based updates. Our framework combines Langevin dynamics-based posterior sampling, adaptive noise schedule learning, warm-start initialization from classical solvers, and a progressive training strategy inspired by algorithm unrolling. By considering the perception-distortion tradeoff, our method also mitigates the over-smoothing effects commonly observed in prior approaches and enables reconstructions with fine details while minimizing artifacts. Extensive experiments demonstrate that our approach achieves state-of-the-art performance in both efficiency and reconstruction quality.
Over the last decade, deep learning models have been widely used for automatic feature extraction and classification in various Brain-Computer Interface (BCI) tasks. However, their performance and generalization capabilities are often not adequately assessed, as these models are frequently trained and tested under flawed setups and / or influenced by spurious correlations. Recently, these limitations have also been observed in the training and evaluation of Large Brainwave Foundation Models (LBMs). In this work, we employ causal reasoning and careful consideration for task-discriminative artifacts in various EEG datasets covering diverse BCI paradigms and propose a benchmarking protocol to properly evaluate the decoding performance and generalization capabilities of LBMs. Utilising a subject-independent cross-validation approach for each curated benchmark dataset, we showcase that LBMs achieve marginal performance gains over conventional deep learning baselines.
Exploiting large-scale pre-trained models has proven effective in many domains, enabling limited data downstream tasks to be solved more effectively by leveraging upstream representations. However, transfer learning remains underexplored in passive sonar, where models are often trained from scratch using vision-oriented architectures on engineered features. We present the first systematic benchmark comparing transformer and CNN models across multiple passive sonar classification datasets. We evaluate six models across four sonar datasets, employing various fine-tuning strategies, including parameter-efficient Low-Rank Adaptation. Our findings demonstrate that full fine-tuning with audio pre-trained transformers performs most consistently. Fine-tuning on just 10 % of the data can outperform training from scratch on the full dataset. Importantly, we introduce standardised training and evaluation protocols across multiple datasets, making conclusions more robust and replicable compared to prior work. This will inform both research and practice in passive sonar recognition.
Image restoration seeks to reconstruct high-quality images from low-quality observations, with tasks such as superresolution, deblurring, and inpainting. Recently, diffusion models have emerged as powerful tools for image restoration, generating diverse and high-quality samples. However, their high computational demands, including multiple denoising steps, hinder deployment on resource-constrained devices like smartphones. While efforts to optimize sampling trajectories have been made, the large parameters of DMs remain a challenge. Model quantization, a common solution for reducing latency and memory usage, faces channel-wise variance when applied to DMs. Thus, an incoherence processing is needed for the outlier suppression to produce approximately Gaussian distribution. Futhermore, the uniform bitwidth allocation also turns into the bottleneck of quantization due to the ignorance of layer-wise sensitivity. To address the two main challenges, we propose MPRDiff, a mixed-precision quantization method that uses a computationally efficient randomized Hadamard transform technique to eliminate channel-wise incoherency and enable W3.8A4 layer-wise quantization of the DMs. Additionally, we use integer linear programming for optimal mixed bitwidth allocation and incorporate sampling step correction techniques. MPRDiff is evaluated on CelebA and ImageNet, demonstrating a minimal increase in FID while significantly reducing model size.
