Do neural network models of vision learn brain-aligned representations because they share architectural constraints and task objectives with biological vision or because they learn universal features of natural image processing? We characterized the universality of hundreds of thousands of representational dimensions from visual neural networks with varied construction. We found that networks with varied architectures and task objectives learn to represent natural images using a shared set of latent dimensions, despite appearing highly distinct at a surface level. Next, by comparing these networks with human brain representations measured with fMRI, we found that the most brain-aligned representations in neural networks are those that are universal and independent of a network's specific characteristics. Remarkably, each network can be reduced to fewer than ten of its most universal dimensions with little impact on its representational similarity to the human brain. These results suggest that the underlying similarities between artificial and biological vision are primarily governed by a core set of universal image representations that are convergently learned by diverse systems.
相關內容
Networking：IFIP International Conferences on Networking。
Explanation：國際網絡會議。
Publisher：IFIP。
SIT：
We develop a new, unsupervised symmetry learning method that starts with raw data, and gives the minimal (discrete) generator of an underlying Lie group of symmetries, together with a symmetry equivariant representation of the data. The method is able to learn the pixel translation operator from a dataset with only an approximate translation symmetry, and can learn quite different types of symmetries which are not apparent to the naked eye, equally well. The method is based on the formulation of an information-theoretic loss function that measures both the degree to which the dataset is symmetric under a given candidate symmetry, and also, the degree of locality of the samples in the dataset with respect to this symmetry. We demonstrate that this coupling between symmetry and locality, together with a special optimization technique developed for entropy estimation, results in a highly stable system that gives reproducible results. The symmetry actions we consider are group representations, however, we believe the approach has the potential to be generalized to more general, nonlinear actions of non-commutative Lie groups.
With the increase in computational power for the available hardware, the demand for high-resolution data in computer graphics applications increases. Consequently, classical geometry processing techniques based on linear algebra solutions are starting to become obsolete. In this setting, we propose a novel approach for tackling mesh deformation tasks on high-resolution meshes. By reducing the input size with a fast remeshing technique and preserving a consistent representation of the original mesh with local reference frames, we provide a solution that is both scalable and robust in multiple applications, such as as-rigid-as-possible deformations, non-rigid isometric transformations, and pose transfer tasks. We extensively test our technique and compare it against state-of-the-art methods, proving that our approach can handle meshes with hundreds of thousands of vertices in tens of seconds while still achieving results comparable with the other solutions.
Nonparametric procedures are more powerful for detecting interaction in two-way ANOVA when the data are non-normal. In this paper, we compute null critical values for the aligned rank-based tests (APCSSA/APCSSM) where the levels of the factors are between 2 and 6. We compare the performance of these new procedures with the ANOVA F-test for interaction, the adjusted rank transform test (ART), Conover's rank transform procedure (RT), and a rank-based ANOVA test (raov) using Monte Carlo simulations. The new procedures APCSSA/APCSSM are comparable with existing competitors in all settings. Even though there is no single dominant test in detecting interaction effects for non-normal data, nonparametric procedure APCSSM is the most highly recommended procedure for Cauchy errors settings.
A Gaussian process is proposed as a model for the posterior distribution of the local predictive ability of a model or expert, conditional on a vector of covariates, from historical predictions in the form of log predictive scores. Assuming Gaussian expert predictions and a Gaussian data generating process, a linear transformation of the predictive score follows a noncentral chi-squared distribution with one degree of freedom. Motivated by this we develop a noncentral chi-squared Gaussian process regression to flexibly model local predictive ability, with the posterior distribution of the latent GP function and kernel hyperparameters sampled by Hamiltonian Monte Carlo. We show that a cube-root transformation of the log scores is approximately Gaussian with homoscedastic variance, making it possible to estimate the model much faster by marginalizing the latent GP function analytically. A multi-output Gaussian process regression is also introduced to model the dependence in predictive ability between experts, both for inference and prediction purposes. Linear pools based on learned local predictive ability are applied to predict daily bike usage in Washington DC.
The purpose of this work is to investigate the soundness and utility of a neural network-based approach as a framework for exploring the impact of image enhancement techniques on visual cortex activation. In a preliminary study, we prepare a set of state-of-the-art brain encoding models, selected among the top 10 methods that participated in The Algonauts Project 2023 Challenge [16]. We analyze their ability to make valid predictions about the effects of various image enhancement techniques on neural responses. Given the impossibility of acquiring the actual data due to the high costs associated with brain imaging procedures, our investigation builds up on a series of experiments. Specifically, we analyze the ability of brain encoders to estimate the cerebral reaction to various augmentations by evaluating the response to augmentations targeting objects (i.e., faces and words) with known impact on specific areas. Moreover, we study the predicted activation in response to objects unseen during training, exploring the impact of semantically out-of-distribution stimuli. We provide relevant evidence for the generalization ability of the models forming the proposed framework, which appears to be promising for the identification of the optimal visual augmentation filter for a given task, model-driven design strategies as well as for AR and VR applications.
A central task in knowledge compilation is to compile a CNF-SAT instance into a succinct representation format that allows efficient operations such as testing satisfiability, counting, or enumerating all solutions. Useful representation formats studied in this area range from ordered binary decision diagrams (OBDDs) to circuits in decomposable negation normal form (DNNFs). While it is known that there exist CNF formulas that require exponential size representations, the situation is less well studied for other types of constraints than Boolean disjunctive clauses. The constraint satisfaction problem (CSP) is a powerful framework that generalizes CNF-SAT by allowing arbitrary sets of constraints over any finite domain. The main goal of our work is to understand for which type of constraints (also called the constraint language) it is possible to efficiently compute representations of polynomial size. We answer this question completely and prove two tight characterizations of efficiently compilable constraint languages, depending on whether target format is structured. We first identify the combinatorial property of ``strong blockwise decomposability'' and show that if a constraint language has this property, we can compute DNNF representations of linear size. For all other constraint languages we construct families of CSP-instances that provably require DNNFs of exponential size. For a subclass of ``strong uniformly blockwise decomposable'' constraint languages we obtain a similar dichotomy for structured DNNFs. In fact, strong (uniform) blockwise decomposability even allows efficient compilation into multi-valued analogs of OBDDs and FBDDs, respectively. Thus, we get complete characterizations for all knowledge compilation classes between O(B)DDs and DNNFs.
We develop confidence sets which provide spatial uncertainty guarantees for the output of a black-box machine learning model designed for image segmentation. To do so we adapt conformal inference to the imaging setting, obtaining thresholds on a calibration dataset based on the distribution of the maximum of the transformed logit scores within and outside of the ground truth masks. We prove that these confidence sets, when applied to new predictions of the model, are guaranteed to contain the true unknown segmented mask with desired probability. We show that learning appropriate score transformations on a learning dataset before performing calibration is crucial for optimizing performance. We illustrate and validate our approach on a polpys tumor dataset. To do so we obtain the logit scores from a deep neural network trained for polpys segmentation and show that using distance transformed scores to obtain outer confidence sets and the original scores for inner confidence sets enables tight bounds on tumor location whilst controlling the false coverage rate.
Detecting specific structures in a network has been a very active theme of research in distributed computing for at least a decade. In this paper, we start the study of subgraph detection from the perspective of local certification. Remember that a local certification is a distributed mechanism enabling the nodes of a network to check the correctness of the current configuration, thanks to small pieces of information called certificates. Our main question is: For a given graph $H$, what is the minimum certificate size that allows checking that the network does not contain $H$ as a (possibly induced) subgraph? We show a variety of lower and upper bounds, uncovering an interesting interplay between the optimal certificate size, the size of the forbidden subgraph, and the locality of the verification. Along the way we introduce several new technical tools, in particular what we call the \emph{layered map}, which is not specific to forbidden subgraphs and that we expect to be useful for certifying many other properties.
Black-box runtime verification methods for cyber-physical systems can be used to discover errors in systems whose inputs and outputs are expressed as signals over time and their correctness requirements are specified in a temporal logic. Existing methods, such as requirement falsification, often focus on finding a single input that is a counterexample to system correctness. In this paper, we study how to create test generators that can produce multiple and diverse counterexamples for a single requirement. Several counterexamples expose system failures in varying input conditions and support the root cause analysis of the faults. We present the WOGAN algorithm to create such test generators automatically. The algorithm works by training iteratively a Wasserstein generative adversarial network that models the target distribution of the uniform distribution on the set of counterexamples. WOGAN is an algorithm that trains generative models that act as test generators for runtime verification. The training is performed online without the need for a previous model or dataset. We also propose criteria to evaluate such test generators. We evaluate the trained generators on several well-known problems including the ARCH-COMP falsification benchmarks. Our experimental results indicate that generators trained by the WOGAN algorithm are as effective as state-of-the-art requirement falsification algorithms while producing tests that are as diverse as a sample from uniform random sampling. We conclude that WOGAN is a viable method to produce test generators automatically and that these test generators can generate multiple and diverse counterexamples for the runtime verification of cyber-physical systems.
We introduce a novel neural network module that adeptly handles recursive data flow in neural network architectures. At its core, this module employs a self-consistent approach where a set of recursive equations is solved iteratively, halting when the difference between two consecutive iterations falls below a defined threshold. Leveraging this mechanism, we construct a new neural network architecture, an extension of the conformer transducer, which enriches automatic speech recognition systems with a stream of contextual information. Our method notably improves the accuracy of recognizing rare words without adversely affecting the word error rate for common vocabulary. We investigate the improvement in accuracy for these uncommon words using our novel model, both independently and in conjunction with shallow fusion with a context language model. Our findings reveal that the combination of both approaches can improve the accuracy of detecting rare words by as much as 4.5 times. Our proposed self-consistent recursive methodology is versatile and adaptable, compatible with many recently developed encoders, and has the potential to drive model improvements in speech recognition and beyond.
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