Task2Vec creates task embeddings for meta-learning.
problem Creating a framework for selecting feature extractors for new tasks.
method Process images through a probe network to compute task embeddings based on Fisher information matrix.
result Task embeddings predict task similarities and feature extractor performance.
RTFE provides adversarial robustness to multiple models.
problem Adversarial examples can transfer to other models, compromising robustness.
method Proposes RTFE, a deep learning-based pre-processing mechanism.
result RTFE provides adversarial robustness to multiple independently trained classifiers.
Enhances ZSL models with biologically inspired feature enhancement.
problem Limited training data leads to poor feature extraction from pre-trained models.
method Dual-channel learning framework using auxiliary data sets.
result Improves ZSL model's generalization ability and achieves state-of-the-art results.
Headless attacks bypass classification heads to fool transfer learning models.
problem Adversarial attacks against transfer learning models without access to the classification head.
method Label-blind adversarial attacks that do not require class-label information.
result Transfer attack lowers ResNet18 accuracy on CIFAR10 by over 40%.
State-of-the-art speaker diarization systems utilize knowledge from external data, in the form of a pre-trained distance metric, to effectively determine relative speaker identities to unseen data. However, much of recent focus has been on choosing the appropriate feature extractor, ranging from pre-trained i−vectors…
SFSVI uses Gaussian mixtures to approximate neural network outputs for continual learning.
problem Learning new tasks without forgetting old ones in neural networks.
method Sequential function-space variational inference with Gaussian mixture approximation.
result Gaussian mixture SFSVI outperforms other methods in continual learning.
The study uses pre-trained neural networks to adjust for confounding in non-tabular data.
problem Neglecting non-tabular data sources can lead to biased ATE estimates.
method Leverages latent features from pre-trained neural networks to adjust for confounding.
result Neural networks can achieve fast convergence rates for ATE estimation with latent features.
Pre-trained model from healthy ADLs improves gait pattern classification for Parkinson's disease.
problem Limited training data for deep learning models in healthcare applications.
method Used convolutional autoencoder to extract features from healthy ADLs data and trained a multi-layer perceptron model for Parkinson's disease classification.
result Features extracted from healthy ADLs data can be used to train an effective classification model for Parkinson's disease.
ADNN uses prior knowledge to construct financial features.
problem Feature construction in financial trading.
method Tailored neural network structure with domain knowledge.
result ADNN constructs more informative features than genetic programming.
Discriminators can be good feature extractors despite their task focus.
problem Discriminators' features are often considered useless for downstream tasks.
method Theoretical analysis and feature space examination to understand discriminator's role.
result Discriminator features are robust and can prevent mode collapse, making them useful for transfer learning.
FOCA method prevents co-adaptation between feature extractor and classifier.
problem Co-adaptation between feature extractor and classifier degrades neural network performance.
method FOCA method uses randomly-generated, weak classifiers to optimize feature extractor without explicit co-adaptation.
result FOCA features form a point-like distribution within the same class under special conditions.
Enhances CNN feature extractors' separation capacity analysis.
problem Understanding the separation capacity of CNNs.
method Extending Cover's function-counting theory, analyzing scattering networks.
result Identifies factors affecting scattering networks' separation capacity.
Deep learning identifies smartphone users from motion sensor data.
problem Smartphone user identification using motion sensor signals.
method Transformed motion signals into images, trained CNN for classification, used SVM for few-shot identification.
result CNN achieved 89.75% multi-class user classification and 96.72% few-shot user identification accuracy.
System filters and ranks medical answers using pre-trained models.
problem Challenges in ranking and classifying medical answers due to input size and dataset limitations.
method Multi-task learning with pre-trained models as feature extractors.
result Achieved high performance on medical QA task (Spearman's Rho 0.338, MRR 0.9622).
Deep learning is still not a very common tool in speaker verification field. We study deep convolutional neural network performance in the text-prompted speaker verification task. The prompted passphrase is segmented into word states - i.e. digits -to test each digit utterance separately. We train a single high-level f…
TransMatch uses transfer learning to improve few-shot learning accuracy.
problem Building robust models with limited labeled data.
method Transfer-learning framework combining feature extraction, initialization, and semi-supervised learning.
result Significant improvement in few-shot learning accuracy.
SidAE combines autoencoders and Siamese networks for self-supervised feature extraction.
problem Efficiently extract meaningful features from unlabeled data.
method Proposes SidAE, a combination of autoencoders and Siamese networks for self-supervised image classification.
result SidAE outperforms self-supervised baselines across multiple datasets and scenarios, especially with limited labeled data.
Novel loss function improves few-shot classification performance.
problem Few-shot classification struggles with embedding space for new classes.
method Proposes self-compacting softmax loss (SSL) to prevent embedding space fullness.
result SSL leads to significant improvement in state-of-the-art performance.
Enhances fault tolerance in neural networks with a novel multi-criteria objective function.
problem Faults in neural networks, especially in deep learning accelerators, reduce classification accuracy.
method Modelled as two networks: Feature Extractor and Classifier. Proposed a novel multi-criteria objective function combining unsupervised and supervised training.
result The proposed approach achieves high accuracy with superior fault tolerance compared to existing regularizers.
Transfer learning boosts deep learning in source code modeling.
problem Deep learning models are problem-specific and data-hungry.
method Transfer learning to improve performance of deep learning models.
result The proposed approach outperforms state-of-the-art models in source code suggestion task.
Privacy-preserving distributed deep learning method for multiple classification.
problem Privacy issues in training deep learning models for various fields.
method Split learning into common extractor, cloud model, and local classifier.
result Average performance improvement of 2.63% over existing local training models.
Model predicts odor pleasantness using CNN, outperforming manual methods.
problem Predicting odor pleasantness for industry applications.
method Used a convolutional neural network to replace manual feature extraction.
result Shows 90% correlation with human ratings and 99.9% accuracy in categorizing odors.
Deep convolutional neural networks have led to breakthrough results in practical feature extraction applications. The mathematical analysis of these networks was pioneered by Mallat, 2012. Specifically, Mallat considered so-called scattering networks based on identical semi-discrete wavelet frames in each network layer…
We provide a methodology, resilient feature engineering, for creating adversarially resilient classifiers. According to existing work, adversarial attacks identify weakly correlated or non-predictive features learned by the classifier during training and design the adversarial noise to utilize these features. Therefore…
Improves deep learning training by matching mini-batch distributions.
problem Overfitting and noise in mini-batch training.
method ITDM, which matches the moments of mini-batch distributions to reduce overfitting.
result ITDM reduces overfitting and improves DNN training.
Meta-learning improves feature extraction for few-shot tasks.
problem Understanding why meta-learning models perform better on few-shot classification.
method Developed hypotheses and a regularizer to improve standard training routines.
result Meta-learned models outperform classical training routines in few-shot classification.
Improved conversion rate prediction in online advertising using self-supervised pre-training.
problem Data sparsity and calibration issues in predicting conversions given clicks.
method Self-supervised pre-training on all conversion events to enrich CVR prediction model without compromising calibration.
result Improvements in offline training and online A/B tests, with full deployment to Yahoo native advertising system.
Proposes a framework for extracting consistent physiological features across users.
problem Variability of biosignals across different users and tasks.
method Adversarial feature extractor for disentangled universal representations.
result Up to 8.8% improvement in average accuracy of classification.
Paper proposes a new method for robust speaker verification.
problem Improving robustness in speaker verification systems.
method Combines soft VAD and self-adaptive VAD with DNN-based VAD.
result Significant improvement in verification performance in real-world environments.
Hyperparameter optimization aims to find the optimal hyperparameter configuration of a machine learning model, which provides the best performance on a validation dataset. Manual search usually leads to get stuck in a local hyperparameter configuration, and heavily depends on human intuition and experience. A simple al…
Transfer learning using deep neural networks as feature extractors has become increasingly popular over the past few years. It allows to obtain state-of-the-art accuracy on datasets too small to train a deep neural network on its own, and it provides cutting edge descriptors that, combined with nonparametric learning m…
Adversarially robust transfer learning improves model robustness with minimal data.
problem Data scarcity and computational limitations in adversarially robust transfer learning.
method Train classifiers on top of robust feature extractors from source models, or fine-tune end-to-end in target domain using lifelong learning strategies.
result Accurate and robust models can be produced with little data and without adversarial training.
New method uses graph-based interpolation for few-shot classification.
problem Learning models with limited labeled examples.
method Graph-based feature vector interpolation for transductive learning.
result Significant gains in few-shot classification compared to other methods.
An unsupervised learning classification model is described. It achieves classification error probability competitive with that of popular supervised learning classifiers such as SVM or kNN. The model is based on the incremental execution of small step shift and rotation operations upon selected discriminative hyperplan…
Improves unsupervised domain adaptation by enforcing feature extractor to focus on task-relevant information.
problem Leveraging label information from source domain for accurate target domain models without labels.
method Variational Information Bottleneck (VBDA) method that explicitly enforces feature extractor to ignore irrelevant task factors.
result Significantly outperforms state-of-the-art methods across three domain adaptation benchmark datasets.
Improves speaker verification for variable-duration utterances using a feature pyramid module.
problem Improving robustness for variable-duration utterances in speaker verification.
method Integrates a feature pyramid module into multi-scale aggregation to enhance speaker-discriminative information from multiple layers.
result Improves performance for both short and long utterances compared to state-of-the-art approaches.
Pre-trains GNN for generic graph features, reducing data need.
problem Lack of labeled data and expressive features for GNN models.
method Synthetic graph tasks: link reconstruction, centrality ranking, cluster preservation.
result Pre-trained GNN achieves high performance with less labeled data.
A new model improves uncertainty estimation in deep learning.
problem Deep Kernel Learning (DKL) produces unreliable uncertainty estimates.
method Proposed a bi-Lipschitz constraint to preserve distances in feature space.
result DUE model outperforms previous DKL and other methods in uncertainty quality.
forgeNet uses a tree-based ensemble to learn feature graphs for deep learning in omics data.
problem Small sample size vs. large feature space in omics data.
method forgeNet integrates a forest feature graph extractor with a GEDFN architecture.
result ForgeNet achieves high classification accuracy on synthetic and real datasets.
We compare features for dynamic time warping (DTW) when used to bootstrap keyword spotting (KWS) in an almost zero-resource setting. Such quickly-deployable systems aim to support United Nations (UN) humanitarian relief efforts in parts of Africa with severely under-resourced languages. Our objective is to identify aco…
Paper presents CAD system for MI detection in smart cities.
problem Detecting myocardial infarction (MI) in urban healthcare settings.
method Employed transfer learning techniques to retrain VGG-Net for ECG signal classification.
result Achieved high accuracy, sensitivity, and specificity in MI detection.
The paper proposes reusable network components by making them compatible across tasks.
problem Training networks for different tasks independently leads to incompatible components.
method The paper splits a network into a features extractor and a target task head, and proposes various approaches to make them compatible.
result The proposed methods produce components that are directly compatible without compromising accuracy on original tasks.
Proposes PSCs for UQ in deep nets without retraining.
problem Estimating uncertainty in deep nets with a single pass.
method Identifies sensitive, smooth intermediate layer, fits probabilistic model.
result PSCs achieve UQ and OOD detection performance matching existing methods.
FixyNN improves energy efficiency of mobile computer vision tasks.
problem High energy consumption of state-of-the-art CNN models on mobile devices.
method Fixed-weight feature extractor and programmable CNN accelerator for transfer learning.
result Achieved up to 26.6 TOPS/W energy efficiency, nearly 2x more efficient than conventional accelerators.
This paper introduces a new transfer learning method for regression.
problem Boosting predictive power in scenarios with scarce data.
method Affine model transfer, following expected-square loss minimization.
result The affine model transfer method encompasses various existing methods and clarifies its theoretical properties.
In this article, we propose an approach that can make use of not only labeled EEG signals but also the unlabeled ones which is more accessible. We also suggest the use of data fusion to further improve the seizure prediction accuracy. Data fusion in our vision includes EEG signals, cardiogram signals, body temperature …
We propose a Variational Time Series Feature Extractor (VTSFE), inspired by the VAE-DMP model of Chen et al., to be used for action recognition and prediction. Our method is based on variational autoencoders. It improves VAE-DMP in that it has a better noise inference model, a simpler transition model constraining the …
The paper introduces metrics to objectively evaluate interpretability methods.
problem Lack of objective evaluation metrics for interpretability methods.
method Proposes a set of metrics to evaluate interpretability methods along simplicity and broadness.
result Validated metrics on different benchmark tasks and showed their utility in method selection.