The rapid development of neural text-to-speech (TTS) systems enabled its usage in other areas of natural language processing such as automatic speech recognition (ASR) or spoken language translation (SLT). Due to the large number of different TTS architectures and their extensions, selecting which TTS systems to use for synthetic data creation is not an easy task. We use the comparison of five different TTS decoder architectures in the scope of synthetic data generation to show the impact on CTC-based speech recognition training. We compare the recognition results to computable metrics like NISQA MOS and intelligibility, finding that there are no clear relations to the ASR performance. We also observe that for data generation auto-regressive decoding performs better than non-autoregressive decoding, and propose an approach to quantify TTS generalization capabilities.
In several Machine Learning (ML) clustering and dimensionality reduction approaches, such as non-negative matrix factorization (NMF), RESCAL, and K-Means clustering, users must select a hyper-parameter k to define the number of clusters or components that yield an ideal separation of samples or clean clusters. This selection, while difficult, is crucial to avoid overfitting or underfitting the data. Several ML applications use scoring methods (e.g., Silhouette and Davies Boulding scores) to evaluate the cluster pattern stability for a specific k. The score is calculated for different trials over a range of k, and the ideal k is heuristically selected as the value before the model starts overfitting, indicated by a drop or increase in the score resembling an elbow curve plot. While the grid-search method can be used to accurately find a good k value, visiting a range of k can become time-consuming and computationally resource-intensive. In this paper, we introduce the Binary Bleed method based on binary search, which significantly reduces the k search space for these grid-search ML algorithms by truncating the target k values from the search space using a heuristic with thresholding over the scores. Binary Bleed is designed to work with single-node serial, single-node multi-processing, and distributed computing resources. In our experiments, we demonstrate the reduced search space gain over a naive sequential search of the ideal k and the accuracy of the Binary Bleed in identifying the correct k for NMFk, K-Means pyDNMFk, and pyDRESCALk with Silhouette and Davies Boulding scores. We make our implementation of Binary Bleed for the NMF algorithm available on GitHub.
The rapid evolution of multimedia and computer vision technologies requires adaptive visual model deployment strategies to effectively handle diverse tasks and varying environments. This work introduces AxiomVision, a novel framework that can guarantee accuracy by leveraging edge computing to dynamically select the most efficient visual models for video analytics under diverse scenarios. Utilizing a tiered edge-cloud architecture, AxiomVision enables the deployment of a broad spectrum of visual models, from lightweight to complex DNNs, that can be tailored to specific scenarios while considering camera source impacts. In addition, AxiomVision provides three core innovations: (1) a dynamic visual model selection mechanism utilizing continual online learning, (2) an efficient online method that efficiently takes into account the influence of the camera's perspective, and (3) a topology-driven grouping approach that accelerates the model selection process. With rigorous theoretical guarantees, these advancements provide a scalable and effective solution for visual tasks inherent to multimedia systems, such as object detection, classification, and counting. Empirically, AxiomVision achieves a 25.7\% improvement in accuracy.
In the field of Image-Text Retrieval (ITR), recent advancements have leveraged large-scale Vision-Language Pretraining (VLP) for Fine-Grained (FG) instance-level retrieval, achieving high accuracy at the cost of increased computational complexity. For Coarse-Grained (CG) category-level retrieval, prominent approaches employ Cross-Modal Hashing (CMH) to prioritise efficiency, albeit at the cost of retrieval performance. Due to differences in methodologies, FG and CG models are rarely compared directly within evaluations in the literature, resulting in a lack of empirical data quantifying the retrieval performance-efficiency tradeoffs between the two. This paper addresses this gap by introducing the \texttt{FiCo-ITR} library, which standardises evaluation methodologies for both FG and CG models, facilitating direct comparisons. We conduct empirical evaluations of representative models from both subfields, analysing precision, recall, and computational complexity across varying data scales. Our findings offer new insights into the performance-efficiency trade-offs between recent representative FG and CG models, highlighting their respective strengths and limitations. These findings provide the foundation necessary to make more informed decisions regarding model selection for specific retrieval tasks and highlight avenues for future research into hybrid systems that leverage the strengths of both FG and CG approaches.
Large Language Models (LLMs) are increasingly deployed in medicine. However, their utility in non-generative clinical prediction, often presumed inferior to specialized models, remains under-evaluated, leading to ongoing debate within the field and potential for misuse, misunderstanding, or over-reliance due to a lack of systematic benchmarking. Our ClinicRealm study addresses this by benchmarking 9 GPT-based LLMs, 5 BERT-based models, and 7 traditional methods on unstructured clinical notes and structured Electronic Health Records (EHR). Key findings reveal a significant shift: for clinical note predictions, leading LLMs (e.g., DeepSeek R1/V3, GPT o3-mini-high) in zero-shot settings now decisively outperform finetuned BERT models. On structured EHRs, while specialized models excel with ample data, advanced LLMs (e.g., GPT-4o, DeepSeek R1/V3) show potent zero-shot capabilities, often surpassing conventional models in data-scarce settings. Notably, leading open-source LLMs can match or exceed proprietary counterparts. These results establish modern LLMs as powerful non-generative clinical prediction tools, particularly with unstructured text and offering data-efficient structured data options, thus necessitating a re-evaluation of model selection strategies. This research should serve as an important insight for medical informaticists, AI developers, and clinical researchers, potentially prompting a reassessment of current assumptions and inspiring new approaches to LLM application in predictive healthcare.
This study introduces novel superior scoring rules called Penalized Brier Score (PBS) and Penalized Logarithmic Loss (PLL) to improve model evaluation for probabilistic classification. Traditional scoring rules like Brier Score and Logarithmic Loss sometimes assign better scores to misclassifications in comparison with correct classifications. This discrepancy from the actual preference for rewarding correct classifications can lead to suboptimal model selection. By integrating penalties for misclassifications, PBS and PLL modify traditional proper scoring rules to consistently assign better scores to correct predictions. Formal proofs demonstrate that PBS and PLL satisfy strictly proper scoring rule properties while also preferentially rewarding accurate classifications. Experiments showcase the benefits of using PBS and PLL for model selection, model checkpointing, and early stopping. PBS exhibits a higher negative correlation with the F1 score compared to the Brier Score during training. Thus, PBS more effectively identifies optimal checkpoints and early stopping points, leading to improved F1 scores. Comparative analysis verifies models selected by PBS and PLL achieve superior F1 scores. Therefore, PBS and PLL address the gap between uncertainty quantification and accuracy maximization by encapsulating both proper scoring principles and explicit preference for true classifications. The proposed metrics can enhance model evaluation and selection for reliable probabilistic classification.
We address the selection and evaluation of uncertain segmentation methods in medical imaging and present two case studies: prostate segmentation, illustrating that for minimal annotator variation simple deterministic models can suffice, and lung lesion segmentation, highlighting the limitations of the Generalized Energy Distance (GED) in model selection. Our findings lead to guidelines for accurately choosing and developing uncertain segmentation models, that integrate aleatoric and epistemic components. These guidelines are designed to aid researchers and practitioners in better developing, selecting, and evaluating uncertain segmentation methods, thereby facilitating enhanced adoption and effective application of segmentation uncertainty in practice.
This paper introduces Patched Round-Trip Correctness (Patched RTC), a novel evaluation technique for Large Language Models (LLMs) applied to diverse software development tasks, particularly focusing on "outer loop" activities such as bug fixing, code review, and documentation updates. Patched RTC extends the original Round-Trip Correctness method to work with any LLM and downstream task, offering a self-evaluating framework that measures consistency and robustness of model responses without human intervention. The study demonstrates a correlation between Patched RTC scores and task-specific accuracy metrics, presenting it as an alternative to the LLM-as-Judge paradigm for open-domain task evaluation. We implement Patched RTC in an open-source framework called patchwork, allowing for transparent evaluation during inference across various patchflows. Experiments comparing GPT-3.5 and GPT-4 models across different software development tasks reveal that Patched RTC effectively distinguishes model performance and task difficulty. The paper also explores the impact of consistency prompts on improving model accuracy, suggesting that Patched RTC can guide prompt refinement and model selection for complex software development workflows.
Statistical heterogeneity is a root cause of tension among accuracy, fairness, and robustness of federated learning (FL), and is key in paving a path forward. Personalized FL (PFL) is an approach that aims to reduce the impact of statistical heterogeneity by developing personalized models for individual users, while also inherently providing benefits in terms of fairness and robustness. However, existing PFL frameworks focus on improving the performance of personalized models while neglecting the global model. Moreover, these frameworks achieve sublinear convergence rates and rely on strong assumptions. In this paper, we propose FLAME, an optimization framework by utilizing the alternating direction method of multipliers (ADMM) to train personalized and global models. We propose a model selection strategy to improve performance in situations where clients have different types of heterogeneous data. Our theoretical analysis establishes the global convergence and two kinds of convergence rates for FLAME under mild assumptions. We theoretically demonstrate that FLAME is more robust and fair than the state-of-the-art methods on a class of linear problems. Our experimental findings show that FLAME outperforms state-of-the-art methods in convergence and accuracy, and it achieves higher test accuracy under various attacks and performs more uniformly across clients.
With the growing access to administrative health databases, retrospective studies have become crucial evidence for medical treatments. Yet, non-randomized studies frequently face selection biases, requiring mitigation strategies. Propensity score matching (PSM) addresses these biases by selecting comparable populations, allowing for analysis without further methodological constraints. However, PSM has several drawbacks. Different matching methods can produce significantly different Average Treatment Effects (ATE) for the same task, even when meeting all validation criteria. To prevent cherry-picking the best method, public authorities must involve field experts and engage in extensive discussions with researchers. To address this issue, we introduce a novel metric, A2A, to reduce the number of valid matches. A2A constructs artificial matching tasks that mirror the original ones but with known outcomes, assessing each matching method's performance comprehensively from propensity estimation to ATE estimation. When combined with Standardized Mean Difference, A2A enhances the precision of model selection, resulting in a reduction of up to 50% in ATE estimation errors across synthetic tasks and up to 90% in predicted ATE variability across both synthetic and real-world datasets. To our knowledge, A2A is the first metric capable of evaluating outcome correction accuracy using covariates not involved in selection. Computing A2A requires solving hundreds of PSMs, we therefore automate all manual steps of the PSM pipeline. We integrate PSM methods from Python and R, our automated pipeline, a new metric, and reproducible experiments into popmatch, our new Python package, to enhance reproducibility and accessibility to bias correction methods.
Animals flexibly change their behavior depending on context. It is reported that the hippocampus is one of the most prominent regions for contextual behaviors, and its sequential activity shows context dependency. However, how such context-dependent sequential activity is established through reorganization of neuronal activity (remapping) is unclear. To better understand the formation of hippocampal activity and its contribution to context-dependent flexible behavior, we present a novel biologically plausible reinforcement learning model. In this model, a context-selection module promotes the formation of context-dependent sequential activity and allows for flexible switching of behavior in multiple contexts. This model reproduces a variety of findings from neural activity, optogenetic inactivation, human fMRI, and clinical research. Furthermore, our model predicts that imbalances in the ratio between sensory and contextual inputs in the context-selection module account for schizophrenia (SZ) and autism spectrum disorder (ASD)-like behaviors.
Despite the proliferation of large vision-language foundation models, estimation of the learning and forgetting outcomes following fine-tuning of these models remains largely unexplored. Inspired by work highlighting the significance of the modality gap in contrastive dual-encoders, we propose the Inter-Intra Modal Measure (IIMM). Combining terms quantifying the similarity between image embeddings and the similarity between incorrect image and label embedding pairs, the IIMM functions as a strong predictor of performance changes with fine-tuning. Our extensive empirical analysis across four state-of-the-art vision-language models (CLIP, SigLIP, CoCa, EVA-02-CLIP) and five fine-tuning techniques (full fine-tuning, BitFit, attention-weight tuning, LoRA, CLIP-Adapter) demonstrates a strong, statistically significant linear relationship: fine-tuning on tasks with higher IIMM scores produces greater in-domain performance gains but also induces more severe out-of-domain performance degradation, with some parameter-efficient fine-tuning (PEFT) methods showing extreme forgetting. We compare our measure against transfer scores from state-of-the-art model selection methods and show that the IIMM is significantly more predictive of accuracy gains. With only a single forward pass of the target data, practitioners can leverage this key insight to heuristically evaluate the degree to which a model can be expected to improve following fine-tuning. Given additional knowledge about the model's performance on a few diverse tasks, this heuristic further evolves into a strong predictor of expected performance changes when training for new tasks.
The cost of errors related to machine learning classifiers, namely, false positives and false negatives, are not equal and are application dependent. For example, in cybersecurity applications, the cost of not detecting an attack is very different from marking a benign activity as an attack. Various design choices during machine learning model building, such as hyperparameter tuning and model selection, allow a data scientist to trade-off between these two errors. However, most of the commonly used metrics to evaluate model quality, such as $F_1$ score, which is defined in terms of model precision and recall, treat both these errors equally, making it difficult for users to optimize for the actual cost of these errors. In this paper, we propose a new cost-aware metric, $C_{score}$ based on precision and recall that can replace $F_1$ score for model evaluation and selection. It includes a cost ratio that takes into account the differing costs of handling false positives and false negatives. We derive and characterize the new cost metric, and compare it to $F_1$ score. Further, we use this metric for model thresholding for five cybersecurity related datasets for multiple cost ratios. The results show an average cost savings of 49%.
Test-Time Adaptation (TTA) has recently emerged as a promising strategy for tackling the problem of machine learning model robustness under distribution shifts by adapting the model during inference without access to any labels. Because of task difficulty, hyperparameters strongly influence the effectiveness of adaptation. However, the literature has provided little exploration into optimal hyperparameter selection. In this work, we tackle this problem by evaluating existing TTA methods using surrogate-based hp-selection strategies (which do not assume access to the test labels) to obtain a more realistic evaluation of their performance. We show that some of the recent state-of-the-art methods exhibit inferior performance compared to the previous algorithms when using our more realistic evaluation setup. Further, we show that forgetting is still a problem in TTA as the only method that is robust to hp-selection resets the model to the initial state at every step. We analyze different types of unsupervised selection strategies, and while they work reasonably well in most scenarios, the only strategies that work consistently well use some kind of supervision (either by a limited number of annotated test samples or by using pretraining data). Our findings underscore the need for further research with more rigorous benchmarking by explicitly stating model selection strategies, to facilitate which we open-source our code.
In the drug discovery process, where experiments can be costly and time-consuming, computational models that predict drug-target interactions are valuable tools to accelerate the development of new therapeutic agents. Estimating the uncertainty inherent in these neural network predictions provides valuable information that facilitates optimal decision-making when risk assessment is crucial. However, such models can be poorly calibrated, which results in unreliable uncertainty estimates that do not reflect the true predictive uncertainty. In this study, we compare different metrics, including accuracy and calibration scores, used for model hyperparameter tuning to investigate which model selection strategy achieves well-calibrated models. Furthermore, we propose to use a computationally efficient Bayesian uncertainty estimation method named Bayesian Linear Probing (BLP), which generates Hamiltonian Monte Carlo (HMC) trajectories to obtain samples for the parameters of a Bayesian Logistic Regression fitted to the hidden layer of the baseline neural network. We report that BLP improves model calibration and achieves the performance of common uncertainty quantification methods by combining the benefits of uncertainty estimation and probability calibration methods. Finally, we show that combining post hoc calibration method with well-performing uncertainty quantification approaches can boost model accuracy and calibration.
Model multiplicity, the phenomenon where multiple models achieve similar performance despite different underlying learned functions, introduces arbitrariness in model selection. While this arbitrariness may seem inconsequential in expectation, its impact on individuals can be severe. This paper explores various individual concerns stemming from multiplicity, including the effects of arbitrariness beyond final predictions, disparate arbitrariness for individuals belonging to protected groups, and the challenges associated with the arbitrariness of a single algorithmic system creating a monopoly across various contexts. It provides both an empirical examination of these concerns and a comprehensive analysis from the legal standpoint, addressing how these issues are perceived in the anti-discrimination law in Canada. We conclude the discussion with technical challenges in the current landscape of model multiplicity to meet legal requirements and the legal gap between current law and the implications of arbitrariness in model selection, highlighting relevant future research directions for both disciplines.
In financial applications, regulations or best practices often lead to specific requirements in machine learning relating to four key pillars: fairness, privacy, interpretability and greenhouse gas emissions. These all sit in the broader context of sustainability in AI, an emerging practical AI topic. However, although these pillars have been individually addressed by past literature, none of these works have considered all the pillars. There are inherent trade-offs between each of the pillars (for example, accuracy vs fairness or accuracy vs privacy), making it even more important to consider them together. This paper outlines a new framework for Sustainable Machine Learning and proposes FPIG, a general AI pipeline that allows for these critical topics to be considered simultaneously to learn the trade-offs between the pillars better. Based on the FPIG framework, we propose a meta-learning algorithm to estimate the four key pillars given a dataset summary, model architecture, and hyperparameters before model training. This algorithm allows users to select the optimal model architecture for a given dataset and a given set of user requirements on the pillars. We illustrate the trade-offs under the FPIG model on three classical datasets and demonstrate the meta-learning approach with an example of real-world datasets and models with different interpretability, showcasing how it can aid model selection.
Traditional single-task image restoration methods excel in handling specific degradation types but struggle with multiple degradations. To address this limitation, we propose Grouped Restoration with Image Degradation Similarity (GRIDS), a novel approach that harmonizes the competing objectives inherent in multiple-degradation restoration. We first introduce a quantitative method for assessing relationships between image degradations using statistical modeling of deep degradation representations. This analysis facilitates the strategic grouping of similar tasks, enhancing both the efficiency and effectiveness of the restoration process. Based on the degradation similarity, GRIDS divides restoration tasks into one of the optimal groups, where tasks within the same group are highly correlated. For instance, GRIDS effectively groups 11 degradation types into 4 cohesive groups. Trained models within each group show significant improvements, with an average improvement of 0.09dB over single-task upper bound models and 2.24dB over the mix-training baseline model. GRIDS incorporates an adaptive model selection mechanism for inference, automatically selecting the appropriate grouped-training model based on the input degradation. This mechanism is particularly useful for real-world scenarios with unknown degradations as it does not rely on explicit degradation classification modules. Furthermore, our method can predict model generalization ability without the need for network inference, providing valuable insights for practitioners.
Text-to-image generative models have recently attracted considerable interest, enabling the synthesis of high-quality images from textual prompts. However, these models often lack the capability to generate specific subjects from given reference images or to synthesize novel renditions under varying conditions. Methods like DreamBooth and Subject-driven Text-to-Image (SuTI) have made significant progress in this area. Yet, both approaches primarily focus on enhancing similarity to reference images and require expensive setups, often overlooking the need for efficient training and avoiding overfitting to the reference images. In this work, we present the $\lambda$-Harmonic reward function, which provides a reliable reward signal and enables early stopping for faster training and effective regularization. By combining the Bradley-Terry preference model, the $\lambda$-Harmonic reward function also provides preference labels for subject-driven generation tasks. We propose Reward Preference Optimization (RPO), which offers a simpler setup (requiring only $3\%$ of the negative samples used by DreamBooth) and fewer gradient steps for fine-tuning. Unlike most existing methods, our approach does not require training a text encoder or optimizing text embeddings and achieves text-image alignment by fine-tuning only the U-Net component. Empirically, $\lambda$-Harmonic proves to be a reliable approach for model selection in subject-driven generation tasks. Based on preference labels and early stopping validation from the $\lambda$-Harmonic reward function, our algorithm achieves a state-of-the-art CLIP-I score of 0.833 and a CLIP-T score of 0.314 on DreamBench.
Unsupervised Domain Adaptation (DA) consists of adapting a model trained on a labeled source domain to perform well on an unlabeled target domain with some data distribution shift. While many methods have been proposed in the literature, fair and realistic evaluation remains an open question, particularly due to methodological difficulties in selecting hyperparameters in the unsupervised setting. With SKADA-bench, we propose a framework to evaluate DA methods on diverse modalities, beyond computer vision task that have been largely explored in the literature. We present a complete and fair evaluation of existing shallow algorithms, including reweighting, mapping, and subspace alignment. Realistic hyperparameter selection is performed with nested cross-validation and various unsupervised model selection scores, on both simulated datasets with controlled shifts and real-world datasets across diverse modalities, such as images, text, biomedical, and tabular data. Our benchmark highlights the importance of realistic validation and provides practical guidance for real-life applications, with key insights into the choice and impact of model selection approaches. SKADA-bench is open-source, reproducible, and can be easily extended with novel DA methods, datasets, and model selection criteria without requiring re-evaluating competitors. SKADA-bench is available on Github at https://github.com/scikit-adaptation/skada-bench.