Measuring diversity accurately is important for many scientific fields, including machine learning (ML), ecology, and chemistry. The Vendi Score was introduced as a generic similarity-based diversity metric that extends the Hill number of order q=1 by leveraging ideas from quantum statistical mechanics. Contrary to many diversity metrics in ecology, the Vendi Score accounts for similarity and does not require knowledge of the prevalence of the categories in the collection to be evaluated for diversity. However, the Vendi Score treats each item in a given collection with a level of sensitivity proportional to the item's prevalence. This is undesirable in settings where there is a significant imbalance in item prevalence. In this paper, we extend the other Hill numbers using similarity to provide flexibility in allocating sensitivity to rare or common items. This leads to a family of diversity metrics -- Vendi scores with different levels of sensitivity -- that can be used in a variety of applications. We study the properties of the scores in a synthetic controlled setting where the ground truth diversity is known. We then test their utility in improving molecular simulations via Vendi Sampling. Finally, we use the Vendi scores to better understand the behavior of image generative models in terms of memorization, duplication, diversity, and sample quality.
Sensitivity analyses reveal the influence of various modeling choices on the outcomes of statistical analyses. While theoretically appealing, they are overwhelmingly inefficient for complex Bayesian models. In this work, we propose sensitivity-aware amortized Bayesian inference (SA-ABI), a multifaceted approach to efficiently integrate sensitivity analyses into simulation-based inference with neural networks. First, we utilize weight sharing to encode the structural similarities between alternative likelihood and prior specifications in the training process with minimal computational overhead. Second, we leverage the rapid inference of neural networks to assess sensitivity to data perturbations and preprocessing steps. In contrast to most other Bayesian approaches, both steps circumvent the costly bottleneck of refitting the model for each choice of likelihood, prior, or data set. Finally, we propose to use deep ensembles to detect sensitivity arising from unreliable approximation (e.g., due to model misspecification). We demonstrate the effectiveness of our method in applied modeling problems, ranging from disease outbreak dynamics and global warming thresholds to human decision-making. Our results support sensitivity-aware inference as a default choice for amortized Bayesian workflows, automatically providing modelers with insights into otherwise hidden dimensions.
As large language models (LLMs) are adopted as a fundamental component of language technologies, it is crucial to accurately characterize their performance. Because choices in prompt design can strongly influence model behavior, this design process is critical in effectively using any modern pre-trained generative language model. In this work, we focus on LLM sensitivity to a quintessential class of meaning-preserving design choices: prompt formatting. We find that several widely used open-source LLMs are extremely sensitive to subtle changes in prompt formatting in few-shot settings, with performance differences of up to 76 accuracy points when evaluated using LLaMA-2-13B. Sensitivity remains even when increasing model size, the number of few-shot examples, or performing instruction tuning. Our analysis suggests that work evaluating LLMs with prompting-based methods would benefit from reporting a range of performance across plausible prompt formats, instead of the currently-standard practice of reporting performance on a single format. We also show that format performance only weakly correlates between models, which puts into question the methodological validity of comparing models with an arbitrarily chosen, fixed prompt format. To facilitate systematic analysis we propose FormatSpread, an algorithm that rapidly evaluates a sampled set of plausible prompt formats for a given task, and reports the interval of expected performance without accessing model weights. Furthermore, we present a suite of analyses that characterize the nature of this sensitivity, including exploring the influence of particular atomic perturbations and the internal representation of particular formats.
Various Large Language Models~(LLMs) from the Generative Pretrained Transformer(GPT) family have achieved outstanding performances in a wide range of text generation tasks. However, the enormous model sizes have hindered their practical use in real-world applications due to high inference latency. Therefore, improving the efficiencies of LLMs through quantization, pruning, and other means has been a key issue in LLM studies. In this work, we propose a method based on Hessian sensitivity-aware mixed sparsity pruning to prune LLMs to at least 50% sparsity without the need of any retraining. It allocates sparsity adaptively based on sensitivity, allowing us to reduce pruning-induced error while maintaining the overall sparsity level. The advantages of the proposed method exhibit even more when the sparsity is extremely high. Furthermore, our method is compatible with quantization, enabling further compression of LLMs. We have released the available code.
We present a novel learned image reconstruction method for accelerated cardiac MRI with multiple receiver coils based on deep convolutional neural networks (CNNs) and algorithm unrolling. In contrast to many existing learned MR image reconstruction techniques that necessitate coil-sensitivity map (CSM) estimation as a distinct network component, our proposed approach avoids explicit CSM estimation. Instead, it implicitly captures and learns to exploit the inter-coil relationships of the images. Our method consists of a series of novel learned image and k-space blocks with shared latent information and adaptation to the acquisition parameters by feature-wise modulation (FiLM), as well as coil-wise data-consistency (DC) blocks. Our method achieved PSNR values of 34.89 and 35.56 and SSIM values of 0.920 and 0.942 in the cine track and mapping track validation leaderboard of the MICCAI STACOM CMRxRecon Challenge, respectively, ranking 4th among different teams at the time of writing. Code will be made available at https://github.com/fzimmermann89/CMRxRecon
We study policy evaluation of offline contextual bandits subject to unobserved confounders. Sensitivity analysis methods are commonly used to estimate the policy value under the worst-case confounding over a given uncertainty set. However, existing work often resorts to some coarse relaxation of the uncertainty set for the sake of tractability, leading to overly conservative estimation of the policy value. In this paper, we propose a general estimator that provides a sharp lower bound of the policy value using convex programming. The generality of our estimator enables various extensions such as sensitivity analysis with f-divergence, model selection with cross validation and information criterion, and robust policy learning with the sharp lower bound. Furthermore, our estimation method can be reformulated as an empirical risk minimization problem thanks to the strong duality, which enables us to provide strong theoretical guarantees of the proposed estimator using techniques of the M-estimation.
Personalized federated learning (PFL) reduces the impact of non-independent and identically distributed (non-IID) data among clients by allowing each client to train a personalized model when collaborating with others. A key question in PFL is to decide which parameters of a client should be localized or shared with others. In current mainstream approaches, all layers that are sensitive to non-IID data (such as classifier layers) are generally personalized. The reasoning behind this approach is understandable, as localizing parameters that are easily influenced by non-IID data can prevent the potential negative effect of collaboration. However, we believe that this approach is too conservative for collaboration. For example, for a certain client, even if its parameters are easily influenced by non-IID data, it can still benefit by sharing these parameters with clients having similar data distribution. This observation emphasizes the importance of considering not only the sensitivity to non-IID data but also the similarity of data distribution when determining which parameters should be localized in PFL. This paper introduces a novel guideline for client collaboration in PFL. Unlike existing approaches that prohibit all collaboration of sensitive parameters, our guideline allows clients to share more parameters with others, leading to improved model performance. Additionally, we propose a new PFL method named FedCAC, which employs a quantitative metric to evaluate each parameter's sensitivity to non-IID data and carefully selects collaborators based on this evaluation. Experimental results demonstrate that FedCAC enables clients to share more parameters with others, resulting in superior performance compared to state-of-the-art methods, particularly in scenarios where clients have diverse distributions.
A central question for neuroscience is how to characterize brain representations of perceptual and cognitive content. An ideal characterization should distinguish different functional regions with robustness to noise and idiosyncrasies of individual brains that do not correspond to computational differences. Previous studies have characterized brain representations by their representational geometry, which is defined by the representational dissimilarity matrix (RDM), a summary statistic that abstracts from the roles of individual neurons (or responses channels) and characterizes the discriminability of stimuli. Here we explore a further step of abstraction: from the geometry to the topology of brain representations. We propose topological representational similarity analysis (tRSA), an extension of representational similarity analysis (RSA) that uses a family of geo-topological summary statistics that generalizes the RDM to characterize the topology while de-emphasizing the geometry. We evaluate this new family of statistics in terms of the sensitivity and specificity for model selection using both simulations and fMRI data. In the simulations, the ground truth is a data-generating layer representation in a neural network model and the models are the same and other layers in different model instances (trained from different random seeds). In fMRI, the ground truth is a visual area and the models are the same and other areas measured in different subjects. Results show that topology-sensitive characterizations of population codes are robust to noise and interindividual variability and maintain excellent sensitivity to the unique representational signatures of different neural network layers and brain regions. These methods enable researchers to calibrate comparisons among representations in brains and models to be sensitive to the geometry, the topology, or a combination of both.
We present a set of metrics that utilize vision priors to effectively assess the performance of saliency methods on image classification tasks. To understand behavior in deep learning models, many methods provide visual saliency maps emphasizing image regions that most contribute to a model prediction. However, there is limited work on analyzing the reliability of saliency methods in explaining model decisions. We propose the metric COnsistency-SEnsitivity (COSE) that quantifies the equivariant and invariant properties of visual model explanations using simple data augmentations. Through our metrics, we show that although saliency methods are thought to be architecture-independent, most methods could better explain transformer-based models over convolutional-based models. In addition, GradCAM was found to outperform other methods in terms of COSE but was shown to have limitations such as lack of variability for fine-grained datasets. The duality between consistency and sensitivity allow the analysis of saliency methods from different angles. Ultimately, we find that it is important to balance these two metrics for a saliency map to faithfully show model behavior.
Random forests (RFs) are well suited for prediction modeling and variable selection in high-dimensional omics studies. The effect of hyperparameters of the RF algorithm on prediction performance and variable importance estimation have previously been investigated. However, how hyperparameters impact RF-based variable selection remains unclear. We evaluate the effects on the Vita and the Boruta variable selection procedures based on two simulation studies utilizing theoretical distributions and empirical gene expression data. We assess the ability of the procedures to select important variables (sensitivity) while controlling the false discovery rate (FDR). Our results show that the proportion of splitting candidate variables and the sample fraction for the training dataset influence the selection procedures more than the drawing strategy of the training datasets and the minimal terminal node size. A suitable setting of the RF hyperparameters depends on the correlation structure in the data. For weakly correlated predictor variables, the default value of the number of splitting variables is optimal, but smaller values of the sample fraction result in larger sensitivity. In contrast, the difference in sensitivity of the optimal compared to the default value of sample fraction is negligible for strongly correlated predictor variables, whereas smaller values than the default are better in the other settings. In conclusion, the default values of the hyperparameters will not always be suitable for identifying important variables. Thus, adequate values differ depending on whether the aim of the study is optimizing prediction performance or variable selection.
Annotators' sociodemographic backgrounds (i.e., the individual compositions of their gender, age, educational background, etc.) have a strong impact on their decisions when working on subjective NLP tasks, such as toxic language detection. Often, heterogeneous backgrounds result in high disagreements. To model this variation, recent work has explored sociodemographic prompting, a technique, which steers the output of prompt-based models towards answers that humans with specific sociodemographic profiles would give. However, the available NLP literature disagrees on the efficacy of this technique - it remains unclear for which tasks and scenarios it can help, and the role of the individual factors in sociodemographic prompting is still unexplored. We address this research gap by presenting the largest and most comprehensive study of sociodemographic prompting today. We analyze its influence on model sensitivity, performance and robustness across seven datasets and six instruction-tuned model families. We show that sociodemographic information affects model predictions and can be beneficial for improving zero-shot learning in subjective NLP tasks. However, its outcomes largely vary for different model types, sizes, and datasets, and are subject to large variance with regards to prompt formulations. Most importantly, our results show that sociodemographic prompting should be used with care for sensitive applications, such as toxicity annotation or when studying LLM alignment. Code and data: https://github.com/UKPLab/arxiv2023-sociodemographic-prompting
The rapid development of Large Language Models (LLMs) and the emergence of novel abilities with scale have necessitated the construction of holistic, diverse and challenging benchmarks such as HELM and BIG-bench. However, at the moment, most of these benchmarks focus only on performance in English and evaluations that include Southeast Asian (SEA) languages are few in number. We therefore propose BHASA, a holistic linguistic and cultural evaluation suite for LLMs in SEA languages. It comprises three components: (1) a NLP benchmark covering eight tasks across Natural Language Understanding (NLU), Generation (NLG) and Reasoning (NLR) tasks, (2) LINDSEA, a linguistic diagnostic toolkit that spans the gamut of linguistic phenomena including syntax, semantics and pragmatics, and (3) a cultural diagnostics dataset that probes for both cultural representation and sensitivity. For this preliminary effort, we implement the NLP benchmark only for Indonesian, Vietnamese, Thai and Tamil, and we only include Indonesian and Tamil for LINDSEA and the cultural diagnostics dataset. As GPT-4 is purportedly one of the best-performing multilingual LLMs at the moment, we use it as a yardstick to gauge the capabilities of LLMs in the context of SEA languages. Our initial experiments on GPT-4 with BHASA find it lacking in various aspects of linguistic capabilities, cultural representation and sensitivity in the targeted SEA languages. BHASA is a work in progress and will continue to be improved and expanded in the future. The repository for this paper can be found at: https://github.com/aisingapore/BHASA
Domain shift is a common problem in clinical applications, where the training images (source domain) and the test images (target domain) are under different distributions. Unsupervised Domain Adaptation (UDA) techniques have been proposed to adapt models trained in the source domain to the target domain. However, those methods require a large number of images from the target domain for model training. In this paper, we propose a novel method for Few-Shot Unsupervised Domain Adaptation (FSUDA), where only a limited number of unlabeled target domain samples are available for training. To accomplish this challenging task, first, a spectral sensitivity map is introduced to characterize the generalization weaknesses of models in the frequency domain. We then developed a Sensitivity-guided Spectral Adversarial MixUp (SAMix) method to generate target-style images to effectively suppresses the model sensitivity, which leads to improved model generalizability in the target domain. We demonstrated the proposed method and rigorously evaluated its performance on multiple tasks using several public datasets.
Biochemical reaction models describing subcellular processes generally come with a large uncertainty. To be able to account for this during the modeling process, we have developed the R-package UQSA, performing uncertainty quantification and sensitivity analysis in an integrated fashion. UQSA is designed for fast sampling of complicated multi-dimensional parameter distributions, using efficient Markov chain Monte Carlo (MCMC) sampling techniques and Vine-copulas to model complicated joint distributions. We perform MCMC sampling both from stochastic and deterministic models, in either likelihood-free or likelihood-based settings. In the likelihood-free case, we use Approximate Bayesian Computation (ABC), while for likelihood-based sampling we provide different algorithms, including the fast geometry-informed algorithm SMMALA (Simplified Manifold Metropolis-Adjusted Langevin Algorithm). The uncertainty quantification can be followed by a variance decomposition-based global sensitivity analysis. We are aiming for biochemical models, but UQSA can be used for any type of reaction networks. The use of Vine-copulas allows us to describe, evaluate, and sample from complicated parameter distributions, as well as adding new datasets in a sequential manner without redoing the previous parameter fit. The code is written in R, with C as a backend to improve speed. We use the SBtab table format for Systems Biology projects for the model description as well as the experimental data. An event system allows the user to model complicated transient input, common within, e.g., neuroscience. UQSA has an extensive documentation with several examples describing different types of models and data. The code has been tested on up to 2000 cores on several nodes on a computing cluster, but we also include smaller examples that can be run on a laptop. Source code: https://github.com/icpm-kth/uqsa
In deep learning research, many melody extraction models rely on redesigning neural network architectures to improve performance. In this paper, we propose an input feature modification and a training objective modification based on two assumptions. First, harmonics in the spectrograms of audio data decay rapidly along the frequency axis. To enhance the model's sensitivity on the trailing harmonics, we modify the Combined Frequency and Periodicity (CFP) representation using discrete z-transform. Second, the vocal and non-vocal segments with extremely short duration are uncommon. To ensure a more stable melody contour, we design a differentiable loss function that prevents the model from predicting such segments. We apply these modifications to several models, including MSNet, FTANet, and a newly introduced model, PianoNet, modified from a piano transcription network. Our experimental results demonstrate that the proposed modifications are empirically effective for singing melody extraction.
Background: We aimed at improving image quality (IQ) of sparse-view computed tomography (CT) images using a U-Net for lung metastasis detection and determining the best tradeoff between number of views, IQ, and diagnostic confidence. Methods: CT images from 41 subjects aged 62.8 $\pm$ 10.6 years (mean $\pm$ standard deviation), 23 men, 34 with lung metastasis, 7 healthy, were retrospectively selected (2016-2018) and forward projected onto 2,048-view sinograms. Six corresponding sparse-view CT data subsets at varying levels of undersampling were reconstructed from sinograms using filtered backprojection with 16, 32, 64, 128, 256, and 512 views. A dual-frame U-Net was trained and evaluated for each subsampling level on 8,658 images from 22 diseased subjects. A representative image per scan was selected from 19 subjects (12 diseased, 7 healthy) for a single-blinded multireader study. These slices, for all levels of subsampling, with and without U-Net postprocessing, were presented to three readers. IQ and diagnostic confidence were ranked using predefined scales. Subjective nodule segmentation was evaluated using sensitivity and Dice similarity coefficient (DSC); clustered Wilcoxon signed-rank test was used. Results: The 64-projection sparse-view images resulted in 0.89 sensitivity and 0.81 DSC, while their counterparts, postprocessed with the U-Net, had improved metrics (0.94 sensitivity and 0.85 DSC) (p = 0.400). Fewer views led to insufficient IQ for diagnosis. For increased views, no substantial discrepancies were noted between sparse-view and postprocessed images. Conclusions: Projection views can be reduced from 2,048 to 64 while maintaining IQ and the confidence of the radiologists on a satisfactory level.
Background: Mental illness can lead to adverse outcomes such as homelessness and police interaction and understanding of the events leading up to these adverse outcomes is important. Predictive models may help identify individuals at risk of such adverse outcomes. Using a fixed observation window cohort with logistic regression (LR) or machine learning (ML) models can result in lower performance when compared with adaptive and parcellated windows. Method: An administrative healthcare dataset was used, comprising of 240,219 individuals in Calgary, Alberta, Canada who were diagnosed with addiction or mental health (AMH) between April 1, 2013, and March 31, 2018. The cohort was followed for 2 years to identify factors associated with homelessness and police interactions. To understand the benefit of flexible windows to predictive models, an alternative cohort was created. Then LR and ML models, including random forests (RF), and extreme gradient boosting (XGBoost) were compared in the two cohorts. Results: Among 237,602 individuals, 0.8% (1,800) experienced first homelessness, while 0.32% (759) reported initial police interaction among 237,141 individuals. Male sex (AORs: H=1.51, P=2.52), substance disorder (AORs: H=3.70, P=2.83), psychiatrist visits (AORs: H=1.44, P=1.49), and drug abuse (AORs: H=2.67, P=1.83) were associated with initial homelessness (H) and police interaction (P). XGBoost showed superior performance using the flexible method (sensitivity =91%, AUC =90% for initial homelessness, and sensitivity =90%, AUC=89% for initial police interaction) Conclusion: This study identified key features associated with initial homelessness and police interaction and demonstrated that flexible windows can improve predictive modeling.
Existing customization methods require access to multiple reference examples to align pre-trained diffusion probabilistic models (DPMs) with user-provided concepts. This paper aims to address the challenge of DPM customization when the only available supervision is a differentiable metric defined on the generated contents. Since the sampling procedure of DPMs involves recursive calls to the denoising UNet, na\"ive gradient backpropagation requires storing the intermediate states of all iterations, resulting in extremely high memory consumption. To overcome this issue, we propose a novel method AdjointDPM, which first generates new samples from diffusion models by solving the corresponding probability-flow ODEs. It then uses the adjoint sensitivity method to backpropagate the gradients of the loss to the models' parameters (including conditioning signals, network weights, and initial noises) by solving another augmented ODE. To reduce numerical errors in both the forward generation and gradient backpropagation processes, we further reparameterize the probability-flow ODE and augmented ODE as simple non-stiff ODEs using exponential integration. Finally, we demonstrate the effectiveness of AdjointDPM on three interesting tasks: converting visual effects into identification text embeddings, finetuning DPMs for specific types of stylization, and optimizing initial noise to generate adversarial samples for security auditing.
Quantization is a widely used technique to compress neural networks. Assigning uniform bit-widths across all layers can result in significant accuracy degradation at low precision and inefficiency at high precision. Mixed-precision quantization (MPQ) addresses this by assigning varied bit-widths to layers, optimizing the accuracy-efficiency trade-off. Existing sensitivity-based methods for MPQ assume that quantization errors across layers are independent, which leads to suboptimal choices. We introduce CLADO, a practical sensitivity-based MPQ algorithm that captures cross-layer dependency of quantization error. CLADO approximates pairwise cross-layer errors using linear equations on a small data subset. Layerwise bit-widths are assigned by optimizing a new MPQ formulation based on cross-layer quantization errors using an Integer Quadratic Program. Experiments with CNN and vision transformer models on ImageNet demonstrate that CLADO achieves state-of-the-art mixed-precision quantization performance. Code repository available here: https://github.com/JamesTuna/CLADO_MPQ
Emotions are experienced and expressed differently across the world. In order to use Large Language Models (LMs) for multilingual tasks that require emotional sensitivity, LMs must reflect this cultural variation in emotion. In this study, we investigate whether the widely-used multilingual LMs in 2023 reflect differences in emotional expressions across cultures and languages. We find that embeddings obtained from LMs (e.g., XLM-RoBERTa) are Anglocentric, and generative LMs (e.g., ChatGPT) reflect Western norms, even when responding to prompts in other languages. Our results show that multilingual LMs do not successfully learn the culturally appropriate nuances of emotion and we highlight possible research directions towards correcting this.