Search Results for author:
Found 46 papers, 33 papers with code
Effective pruning of web-scale datasets based on complexity of concept clusters
Using a simple and intuitive complexity measure, we are able to reduce the training cost to a quarter of regular training.
An Interventional Perspective on Identifiability in Gaussian LTI Systems with Independent Component Analysis
We investigate the relationship between system identification and intervention design in dynamical systems.
Does CLIP's Generalization Performance Mainly Stem from High Train-Test Similarity?
Foundation models like CLIP are trained on hundreds of millions of samples and effortlessly generalize to new tasks and inputs.
Provable Compositional Generalization for Object-Centric Learning
Learning representations that generalize to novel compositions of known concepts is crucial for bridging the gap between human and machine perception.
Scale Alone Does not Improve Mechanistic Interpretability in Vision Models
We use a psychophysical paradigm to quantify one form of mechanistic interpretability for a diverse suite of nine models and find no scaling effect for interpretability - neither for model nor dataset size.
Don't trust your eyes: on the (un)reliability of feature visualizations
Today, visualization methods form the foundation of our knowledge about the internal workings of neural networks, as a type of mechanistic interpretability.
Provably Learning Object-Centric Representations
Under this generative process, we prove that the ground-truth object representations can be identified by an invertible and compositional inference model, even in the presence of dependencies between objects.
Increasing Confidence in Adversarial Robustness Evaluations
Hundreds of defenses have been proposed to make deep neural networks robust against minimal (adversarial) input perturbations.
Embrace the Gap: VAEs Perform Independent Mechanism Analysis
Leveraging self-consistency, we show that the ELBO converges to a regularized log-likelihood.
Visual Representation Learning Does Not Generalize Strongly Within the Same Domain
An important component for generalization in machine learning is to uncover underlying latent factors of variation as well as the mechanism through which each factor acts in the world.
How Well do Feature Visualizations Support Causal Understanding of CNN Activations?
A precise understanding of why units in an artificial network respond to certain stimuli would constitute a big step towards explainable artificial intelligence.
Partial success in closing the gap between human and machine vision
The longstanding distortion robustness gap between humans and CNNs is closing, with the best models now exceeding human feedforward performance on most of the investigated OOD datasets.
Self-Supervised Learning with Data Augmentations Provably Isolates Content from Style
A common practice is to perform data augmentation via hand-crafted transformations intended to leave the semantics of the data invariant.
Ranked #1 on
Image Classification
on Causal3DIdent
If your data distribution shifts, use self-learning
We demonstrate that self-learning techniques like entropy minimization and pseudo-labeling are simple and effective at improving performance of a deployed computer vision model under systematic domain shifts.
Ranked #1 on
Unsupervised Domain Adaptation
on ImageNet-A
(using extra training data)
Fast Minimum-norm Adversarial Attacks through Adaptive Norm Constraints
Evaluating adversarial robustness amounts to finding the minimum perturbation needed to have an input sample misclassified.
Contrastive Learning Inverts the Data Generating Process
Contrastive learning has recently seen tremendous success in self-supervised learning.
Ranked #1 on
Disentanglement
on KITTI-Masks
Exemplary natural images explain CNN activations better than synthetic feature visualizations
Using a well-controlled psychophysical paradigm, we compare the informativeness of synthetic images \citep{olah2017feature} with a simple baseline visualization, namely exemplary natural images that also strongly activate a specific feature map.
Exemplary Natural Images Explain CNN Activations Better than State-of-the-Art Feature Visualization
Even if only a single reference image is given, synthetic images provide less information than natural images ($65\pm5\%$ vs. $73\pm4\%$).
On the surprising similarities between supervised and self-supervised models
In the light of this recent breakthrough, we here compare self-supervised networks to supervised models and human behaviour.
Natural Images are More Informative for Interpreting CNN Activations than State-of-the-Art Synthetic Feature Visualizations
Using a well-controlled psychophysical paradigm, we compare the informativeness of synthetic images by Olah et al. [45] with a simple baseline visualization, namely natural images that also strongly activate a specific feature map.
EagerPy: Writing Code That Works Natively with PyTorch, TensorFlow, JAX, and NumPy
EagerPy is a Python framework that lets you write code that automatically works natively with PyTorch, TensorFlow, JAX, and NumPy.
Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding
We construct an unsupervised learning model that achieves nonlinear disentanglement of underlying factors of variation in naturalistic videos.
Ranked #1 on
Disentanglement
on Natural Sprites
Improving robustness against common corruptions by covariate shift adaptation
With the more robust DeepAugment+AugMix model, we improve the state of the art achieved by a ResNet50 model up to date from 53. 6% mCE to 45. 4% mCE.
Ranked #4 on
Unsupervised Domain Adaptation
on ImageNet-R
Local Convolutions Cause an Implicit Bias towards High Frequency Adversarial Examples
Inspired by theoretical work on linear full-width convolutional models, we hypothesize that the local (i. e. bounded-width) convolutional operations commonly used in current neural networks are implicitly biased to learn high frequency features, and that this is one of the root causes of high frequency adversarial examples.
Benchmarking Unsupervised Object Representations for Video Sequences
Perceiving the world in terms of objects and tracking them through time is a crucial prerequisite for reasoning and scene understanding.
Five Points to Check when Comparing Visual Perception in Humans and Machines
In the second case study, we highlight the difference between necessary and sufficient mechanisms in visual reasoning tasks.
Shortcut Learning in Deep Neural Networks
Deep learning has triggered the current rise of artificial intelligence and is the workhorse of today's machine intelligence.
On Adaptive Attacks to Adversarial Example Defenses
Adaptive attacks have (rightfully) become the de facto standard for evaluating defenses to adversarial examples.
A simple way to make neural networks robust against diverse image corruptions
The human visual system is remarkably robust against a wide range of naturally occurring variations and corruptions like rain or snow.
Learning From Brains How to Regularize Machines
We propose to regularize CNNs using large-scale neuroscience data to learn more robust neural features in terms of representational similarity.
Benchmarking Robustness in Object Detection: Autonomous Driving when Winter is Coming
The ability to detect objects regardless of image distortions or weather conditions is crucial for real-world applications of deep learning like autonomous driving.
Ranked #1 on
Robust Object Detection
on MS COCO
Accurate, reliable and fast robustness evaluation
We here develop a new set of gradient-based adversarial attacks which (a) are more reliable in the face of gradient-masking than other gradient-based attacks, (b) perform better and are more query efficient than current state-of-the-art gradient-based attacks, (c) can be flexibly adapted to a wide range of adversarial criteria and (d) require virtually no hyperparameter tuning.
Approximating CNNs with Bag-of-local-Features models works surprisingly well on ImageNet
Deep Neural Networks (DNNs) excel on many complex perceptual tasks but it has proven notoriously difficult to understand how they reach their decisions.
On Evaluating Adversarial Robustness
Correctly evaluating defenses against adversarial examples has proven to be extremely difficult.
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness
Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by learning increasingly complex representations of object shapes.
Ranked #1 on
Out-of-Distribution Generalization
on ImageNet-W
Adversarial Vision Challenge
The NIPS 2018 Adversarial Vision Challenge is a competition to facilitate measurable progress towards robust machine vision models and more generally applicable adversarial attacks.
One-shot Texture Segmentation
We introduce one-shot texture segmentation: the task of segmenting an input image containing multiple textures given a patch of a reference texture.
Towards the first adversarially robust neural network model on MNIST
Despite much effort, deep neural networks remain highly susceptible to tiny input perturbations and even for MNIST, one of the most common toy datasets in computer vision, no neural network model exists for which adversarial perturbations are large and make semantic sense to humans.
Trace your sources in large-scale data: one ring to find them all
An important preprocessing step in most data analysis pipelines aims to extract a small set of sources that explain most of the data.
Decision-Based Adversarial Attacks: Reliable Attacks Against Black-Box Machine Learning Models
Such decision-based attacks are (1) applicable to real-world black-box models such as autonomous cars, (2) need less knowledge and are easier to apply than transfer-based attacks and (3) are more robust to simple defences than gradient- or score-based attacks.
Foolbox: A Python toolbox to benchmark the robustness of machine learning models
Foolbox is a new Python package to generate such adversarial perturbations and to quantify and compare the robustness of machine learning models.
Comment on "Biologically inspired protection of deep networks from adversarial attacks"
A recent paper suggests that Deep Neural Networks can be protected from gradient-based adversarial perturbations by driving the network activations into a highly saturated regime.
Texture Synthesis Using Shallow Convolutional Networks with Random Filters
The current state of the art in parametric texture synthesis relies on the multi-layer feature space of deep CNNs that were trained on natural images.
Unsupervised learning of an efficient short-term memory network
Specifically, we show how these networks can learn to efficiently represent their present and past inputs, based on local learning rules only.
Demixed principal component analysis of population activity in higher cortical areas reveals independent representation of task parameters
Neurons in higher cortical areas, such as the prefrontal cortex, are known to be tuned to a variety of sensory and motor variables.
Demixed Principal Component Analysis
Standard dimensionality reduction techniques such as principal component analysis (PCA) can provide a succinct and complete description of the data, but the description is constructed independent of the relevant task variables and is often hard to interpret.
