How do neural networks extract patterns from pixels? Feature visualizations attempt to answer this important question by visualizing highly activating patterns through optimization. Today, visualization methods form the foundation of our knowledge about the internal workings of neural networks, as a type of mechanistic interpretability. Here we ask: How reliable are feature visualizations? We start our investigation by developing network circuits that trick feature visualizations into showing arbitrary patterns that are completely disconnected from normal network behavior on natural input. We then provide evidence for a similar phenomenon occurring in standard, unmanipulated networks: feature visualizations are processed very differently from standard input, casting doubt on their ability to "explain" how neural networks process natural images. This can be used as a sanity check for feature visualizations. We underpin our empirical findings by theory proving that the set of functions that can be reliably understood by feature visualization is extremely small and does not include general black-box neural networks. Therefore, a promising way forward could be the development of networks that enforce certain structures in order to ensure more reliable feature visualizations.
相關內容
Networking：IFIP International Conferences on Networking。
Explanation：國際網絡會議。
Publisher：IFIP。
SIT：
We present USLR, a computational framework for longitudinal registration of brain MRI scans to estimate nonlinear image trajectories that are smooth across time, unbiased to any timepoint, and robust to imaging artefacts. It operates on the Lie algebra parameterisation of spatial transforms (which is compatible with rigid transforms and stationary velocity fields for nonlinear deformation) and takes advantage of log-domain properties to solve the problem using Bayesian inference. USRL estimates rigid and nonlinear registrations that: (i) bring all timepoints to an unbiased subject-specific space; and (i) compute a smooth trajectory across the imaging time-series. We capitalise on learning-based registration algorithms and closed-form expressions for fast inference. A use-case Alzheimer's disease study is used to showcase the benefits of the pipeline in multiple fronts, such as time-consistent image segmentation to reduce intra-subject variability, subject-specific prediction or population analysis using tensor-based morphometry. We demonstrate that such approach improves upon cross-sectional methods in identifying group differences, which can be helpful in detecting more subtle atrophy levels or in reducing sample sizes in clinical trials. The code is publicly available in //github.com/acasamitjana/uslr
Feedforward neural networks (FNNs) are typically viewed as pure prediction algorithms, and their strong predictive performance has led to their use in many machine-learning applications. However, their flexibility comes with an interpretability trade-off; thus, FNNs have been historically less popular among statisticians. Nevertheless, classical statistical theory, such as significance testing and uncertainty quantification, is still relevant. Supplementing FNNs with methods of statistical inference, and covariate-effect visualisations, can shift the focus away from black-box prediction and make FNNs more akin to traditional statistical models. This can allow for more inferential analysis, and, hence, make FNNs more accessible within the statistical-modelling context.
Augmented reality (AR) is emerging in visual search tasks for increasingly immersive interactions with virtual objects. We propose an AR approach providing visual and audio hints along with gaze-assisted instant post-task feedback for search tasks based on mobile head-mounted display (HMD). The target case was a book-searching task, in which we aimed to explore the effect of the hints together with the task feedback with two hypotheses. H1: Since visual and audio hints can positively affect AR search tasks, the combination outperforms the individuals. H2: The gaze-assisted instant post-task feedback can positively affect AR search tasks. The proof-of-concept was demonstrated by an AR app in HMD and a comprehensive user study (n=96) consisting of two sub-studies, Study I (n=48) without task feedback and Study II (n=48) with task feedback. Following quantitative and qualitative analysis, our results partially verified H1 and completely verified H2, enabling us to conclude that the synthesis of visual and audio hints conditionally improves the AR visual search task efficiency when coupled with task feedback.
This study addresses the significant challenge of developing efficient decoding algorithms for classifying steady-state visual evoked potentials (SSVEPs) in scenarios characterized by extreme scarcity of calibration data, where only one calibration is available for each stimulus target. To tackle this problem, we introduce a novel cross-subject dual-domain fusion network (CSDuDoFN) incorporating task-related and task-discriminant component analysis (TRCA and TDCA) for one-shot SSVEP classification. The CSDuDoFN framework is designed to comprehensively transfer information from source subjects, while TRCA and TDCA are employed to exploit the single available calibration of the target subject. Specifically, we develop multi-reference least-squares transformation (MLST) to map data from both source subjects and the target subject into the domain of sine-cosine templates, thereby mitigating inter-individual variability and benefiting transfer learning. Subsequently, the transformed data in the sine-cosine templates domain and the original domain data are separately utilized to train a convolutional neural network (CNN) model, with the adequate fusion of their feature maps occurring at distinct network layers. To further capitalize on the calibration of the target subject, source aliasing matrix estimation (SAME) data augmentation is incorporated into the training process of the ensemble TRCA (eTRCA) and TDCA models. Ultimately, the outputs of the CSDuDoFN, eTRCA, and TDCA are combined for SSVEP classification. The effectiveness of our proposed approach is comprehensively evaluated on three publicly available SSVEP datasets, achieving the best performance on two datasets and competitive performance on one. This underscores the potential for integrating brain-computer interface (BCI) into daily life.
Sketch-guided image editing aims to achieve local fine-tuning of the image based on the sketch information provided by the user, while maintaining the original status of the unedited areas. Due to the high cost of acquiring human sketches, previous works mostly relied on edge maps as a substitute for sketches, but sketches possess more rich structural information. In this paper, we propose a sketch generation scheme that can preserve the main contours of an image and closely adhere to the actual sketch style drawn by the user. Simultaneously, current image editing methods often face challenges such as image distortion, training cost, and loss of fine details in the sketch. To address these limitations, We propose a conditional diffusion model (SketchFFusion) based on the sketch structure vector. We evaluate the generative performance of our model and demonstrate that it outperforms existing methods.
A space-time-parameters structure of the parametric parabolic PDEs motivates the application of tensor methods to define reduced order models (ROMs). Within a tensor-based ROM framework, the matrix SVD -- a traditional dimension reduction technique -- yields to a low-rank tensor decomposition (LRTD). Such tensor extension of the Galerkin proper orthogonal decomposition ROMs (POD-ROMs) benefits both the practical efficiency of the ROM and its amenability for the rigorous error analysis when applied to parametric PDEs. The paper addresses the error analysis of the Galerkin LRTD-ROM for an abstract linear parabolic problem that depends on multiple physical parameters. An error estimate for the LRTD-ROM solution is proved, which is uniform with respect to problem parameters and extends to parameter values not in a sampling/training set. The estimate is given in terms of discretization and sampling mesh properties, and LRTD accuracy. The estimate depends on the smoothness rather than on the Kolmogorov n-widths of the parameterized manifold of solutions. Theoretical results are illustrated with several numerical experiments.
Visual inspection tasks often require humans to cooperate with AI-based image classifiers. To enhance this cooperation, explainable artificial intelligence (XAI) can highlight those image areas that have contributed to an AI decision. However, the literature on visual cueing suggests that such XAI support might come with costs of its own. To better understand how the benefits and cost of XAI depend on the accuracy of AI classifications and XAI highlights, we conducted two experiments that simulated visual quality control in a chocolate factory. Participants had to decide whether chocolate moulds contained faulty bars or not, and were always informed whether the AI had classified the mould as faulty or not. In half of the experiment, they saw additional XAI highlights that justified this classification. While XAI speeded up performance, its effects on error rates were highly dependent on (X)AI accuracy. XAI benefits were observed when the system correctly detected and highlighted the fault, but XAI costs were evident for misplaced highlights that marked an intact area while the actual fault was located elsewhere. Eye movement analyses indicated that participants spent less time searching the rest of the mould and thus looked at the fault less often. However, we also observed large interindividual differences. Taken together, the results suggest that despite its potentials, XAI can discourage people from investing effort into their own information analysis.
Topological signal processing (TSP) over simplicial complexes typically assumes observations associated with the simplicial complexes are real scalars. In this paper, we develop TSP theories for the case where observations belong to general abelian groups, including function spaces that are commonly used to represent time-varying signals. Our approach generalizes the Hodge decomposition and allows for signal processing tasks to be performed on these more complex observations. We propose a unified and flexible framework for TSP that expands its applicability to a wider range of signal processing applications. Numerical results demonstrate the effectiveness of this approach and provide a foundation for future research in this area.
Numerical simulations of kinetic problems can become prohibitively expensive due to their large memory footprint and computational costs. A method that has proven to successfully reduce these costs is the dynamical low-rank approximation (DLRA). One key question when using DLRA methods is the construction of robust time integrators that preserve the invariances and associated conservation laws of the original problem. In this work, we demonstrate that the augmented basis update & Galerkin integrator (BUG) preserves solution invariances and the associated conservation laws when using a conservative truncation step and an appropriate time and space discretization. We present numerical comparisons to existing conservative integrators and discuss advantages and disadvantages
This study compares the performance of (1) fine-tuned models and (2) extremely large language models on the task of check-worthy claim detection. For the purpose of the comparison we composed a multilingual and multi-topical dataset comprising texts of various sources and styles. Building on this, we performed a benchmark analysis to determine the most general multilingual and multi-topical claim detector. We chose three state-of-the-art models in the check-worthy claim detection task and fine-tuned them. Furthermore, we selected three state-of-the-art extremely large language models without any fine-tuning. We made modifications to the models to adapt them for multilingual settings and through extensive experimentation and evaluation. We assessed the performance of all the models in terms of accuracy, recall, and F1-score in in-domain and cross-domain scenarios. Our results demonstrate that despite the technological progress in the area of natural language processing, the models fine-tuned for the task of check-worthy claim detection still outperform the zero-shot approaches in a cross-domain settings.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191