Ehsan Abbasnejad
YOU?
Author Swipe
View article: Mysteries of the Deep: Role of Intermediate Representations in Out of Distribution Detection
Mysteries of the Deep: Role of Intermediate Representations in Out of Distribution Detection Open
Out-of-distribution (OOD) detection is essential for reliably deploying machine learning models in the wild. Yet, most methods treat large pre-trained models as monolithic encoders and rely solely on their final-layer representations for d…
View article: Beyond Imitation: Recovering Dense Rewards from Demonstrations
Beyond Imitation: Recovering Dense Rewards from Demonstrations Open
Conventionally, supervised fine-tuning (SFT) is treated as a simple imitation learning process that only trains a policy to imitate expert behavior on demonstration datasets. In this work, we challenge this view by establishing a fundament…
View article: Parameter-Efficient Action Planning with Large Language Models for Vision-and-Language Navigation
Parameter-Efficient Action Planning with Large Language Models for Vision-and-Language Navigation Open
View article: Bayesian Low-Rank Learning (Bella): A Practical Approach to Bayesian Neural Networks
Bayesian Low-Rank Learning (Bella): A Practical Approach to Bayesian Neural Networks Open
Computational complexity of Bayesian learning is impeding its adoption in practical, large-scale tasks, despite demonstrations of significant merits such as improved robustness and resilience to unseen or out-of-distribution inputs over th…
View article: Do We Always Need the Simplicity Bias? Looking for Optimal Inductive Biases in the Wild
Do We Always Need the Simplicity Bias? Looking for Optimal Inductive Biases in the Wild Open
Neural architectures tend to fit their data with relatively simple functions. This "simplicity bias" is widely regarded as key to their success. This paper explores the limits of this principle. Building on recent findings that the simplic…
View article: RandLoRA: Full-rank parameter-efficient fine-tuning of large models
RandLoRA: Full-rank parameter-efficient fine-tuning of large models Open
Low-Rank Adaptation (LoRA) and its variants have shown impressive results in reducing the number of trainable parameters and memory requirements of large transformer networks while maintaining fine-tuning performance. The low-rank nature o…
View article: Learning to Reason and Navigate: Parameter Efficient Action Planning with Large Language Models
Learning to Reason and Navigate: Parameter Efficient Action Planning with Large Language Models Open
View article: Modelling individual variation in human walking gait across populations and walking conditions via gait recognition
Modelling individual variation in human walking gait across populations and walking conditions via gait recognition Open
Human walking gait is a personal story written by the body, a tool for understanding biological identity in healthcare and security. Gait analysis methods traditionally diverged between these domains but are now merging their complementary…
View article: ETAGE: Enhanced Test Time Adaptation with Integrated Entropy and Gradient Norms for Robust Model Performance
ETAGE: Enhanced Test Time Adaptation with Integrated Entropy and Gradient Norms for Robust Model Performance Open
Test time adaptation (TTA) equips deep learning models to handle unseen test data that deviates from the training distribution, even when source data is inaccessible. While traditional TTA methods often rely on entropy as a confidence metr…
View article: InvariantStock: Learning Invariant Features for Mastering the Shifting Market
InvariantStock: Learning Invariant Features for Mastering the Shifting Market Open
Accurately predicting stock returns is crucial for effective portfolio management. However, existing methods often overlook a fundamental issue in the market, namely, distribution shifts, making them less practical for predicting future ma…
View article: Rethinking State Disentanglement in Causal Reinforcement Learning
Rethinking State Disentanglement in Causal Reinforcement Learning Open
One of the significant challenges in reinforcement learning (RL) when dealing with noise is estimating latent states from observations. Causality provides rigorous theoretical support for ensuring that the underlying states can be uniquely…
View article: On the Credibility of Backdoor Attacks Against Object Detectors in the Physical World
On the Credibility of Backdoor Attacks Against Object Detectors in the Physical World Open
Object detectors are vulnerable to backdoor attacks. In contrast to classifiers, detectors possess unique characteristics, architecturally and in task execution; often operating in challenging conditions, for instance, detecting traffic si…
View article: Bayesian Low-Rank LeArning (Bella): A Practical Approach to Bayesian Neural Networks
Bayesian Low-Rank LeArning (Bella): A Practical Approach to Bayesian Neural Networks Open
Computational complexity of Bayesian learning is impeding its adoption in practical, large-scale tasks. Despite demonstrations of significant merits such as improved robustness and resilience to unseen or out-of-distribution inputs over th…
View article: Knowledge Composition using Task Vectors with Learned Anisotropic Scaling
Knowledge Composition using Task Vectors with Learned Anisotropic Scaling Open
Pre-trained models produce strong generic representations that can be adapted via fine-tuning. The learned weight difference relative to the pre-trained model, known as a task vector, characterises the direction and stride of fine-tuning. …
View article: Synergy and Diversity in CLIP: Enhancing Performance Through Adaptive Backbone Ensembling
Synergy and Diversity in CLIP: Enhancing Performance Through Adaptive Backbone Ensembling Open
Contrastive Language-Image Pretraining (CLIP) stands out as a prominent method for image representation learning. Various architectures, from vision transformers (ViTs) to convolutional networks (ResNets) have been trained with CLIP to ser…
View article: BruSLeAttack: A Query-Efficient Score-Based Black-Box Sparse Adversarial Attack
BruSLeAttack: A Query-Efficient Score-Based Black-Box Sparse Adversarial Attack Open
We study the unique, less-well understood problem of generating sparse adversarial samples simply by observing the score-based replies to model queries. Sparse attacks aim to discover a minimum number-the l0 bounded-perturbations to model …
View article: Bayesian Learned Models Can Detect Adversarial Malware For Free
Bayesian Learned Models Can Detect Adversarial Malware For Free Open
The vulnerability of machine learning-based malware detectors to adversarial attacks has prompted the need for robust solutions. Adversarial training is an effective method but is computationally expensive to scale up to large datasets and…
View article: Premonition: Using Generative Models to Preempt Future Data Changes in Continual Learning
Premonition: Using Generative Models to Preempt Future Data Changes in Continual Learning Open
Continual learning requires a model to adapt to ongoing changes in the data distribution, and often to the set of tasks to be performed. It is rare, however, that the data and task changes are completely unpredictable. Given a description …
View article: Do Deep Neural Network Solutions Form a Star Domain?
Do Deep Neural Network Solutions Form a Star Domain? Open
It has recently been conjectured that neural network solution sets reachable via stochastic gradient descent (SGD) are convex, considering permutation invariances (Entezari et al., 2022). This means that a linear path can connect two indep…
View article: Neural Redshift: Random Networks are not Random Functions
Neural Redshift: Random Networks are not Random Functions Open
Our understanding of the generalization capabilities of neural networks (NNs) is still incomplete. Prevailing explanations are based on implicit biases of gradient descent (GD) but they cannot account for the capabilities of models from gr…
View article: Invariant Representation Learning for Generalizable Imitation
Invariant Representation Learning for Generalizable Imitation Open
International audience
View article: Unveiling Backbone Effects in CLIP: Exploring Representational Synergies and Variances
Unveiling Backbone Effects in CLIP: Exploring Representational Synergies and Variances Open
Contrastive Language-Image Pretraining (CLIP) stands out as a prominent method for image representation learning. Various neural architectures, spanning Transformer-based models like Vision Transformers (ViTs) to Convolutional Networks (Co…
View article: Zero-shot Retrieval: Augmenting Pre-trained Models with Search Engines
Zero-shot Retrieval: Augmenting Pre-trained Models with Search Engines Open
Large pre-trained models can dramatically reduce the amount of task-specific data required to solve a problem, but they often fail to capture domain-specific nuances out of the box. The Web likely contains the information necessary to exce…
View article: SCONE-GAN: Semantic Contrastive learning-based Generative Adversarial Network for an end-to-end image translation
SCONE-GAN: Semantic Contrastive learning-based Generative Adversarial Network for an end-to-end image translation Open
SCONE-GAN presents an end-to-end image translation, which is shown to be effective for learning to generate realistic and diverse scenery images. Most current image-to-image translation approaches are devised as two mappings: a translation…
View article: Progressive Feature Adjustment for Semi-supervised Learning from Pretrained Models
Progressive Feature Adjustment for Semi-supervised Learning from Pretrained Models Open
As an effective way to alleviate the burden of data annotation, semi-supervised learning (SSL) provides an attractive solution due to its ability to leverage both labeled and unlabeled data to build a predictive model. While significant pr…
View article: RanPAC: Random Projections and Pre-trained Models for Continual Learning
RanPAC: Random Projections and Pre-trained Models for Continual Learning Open
Continual learning (CL) aims to incrementally learn different tasks (such as classification) in a non-stationary data stream without forgetting old ones. Most CL works focus on tackling catastrophic forgetting under a learning-from-scratch…
View article: Feature-Space Bayesian Adversarial Learning Improved Malware Detector Robustness
Feature-Space Bayesian Adversarial Learning Improved Malware Detector Robustness Open
We present a new algorithm to train a robust malware detector. Malware is a prolific problem and malware detectors are a front-line defense. Modern detectors rely on machine learning algorithms. Now, the adversarial objective is to devise …
View article: Semantic Role Labeling Guided Out-of-distribution Detection
Semantic Role Labeling Guided Out-of-distribution Detection Open
Identifying unexpected domain-shifted instances in natural language processing is crucial in real-world applications. Previous works identify the out-of-distribution (OOD) instance by leveraging a single global feature embedding to represe…
View article: Selective Mixup Helps with Distribution Shifts, But Not (Only) because of Mixup
Selective Mixup Helps with Distribution Shifts, But Not (Only) because of Mixup Open
Mixup is a highly successful technique to improve generalization of neural networks by augmenting the training data with combinations of random pairs. Selective mixup is a family of methods that apply mixup to specific pairs, e.g. only com…
View article: Deep Metric Learning for Scalable Gait-Based Person Re-Identification Using Force Platform Data
Deep Metric Learning for Scalable Gait-Based Person Re-Identification Using Force Platform Data Open
Walking gait data acquired with force platforms may be used for person re-identification (re-ID) in various authentication, surveillance, and forensics applications. Current force platform-based re-ID systems classify a fixed set of identi…