IKA approximates kernels with linear combinations of chosen functions, outperforming Nyström method.
problem Efficient kernel approximation for large datasets.
method IKA method approximates kernels as a linear combination of user-defined functions.
result IKA consistently outperformed Nyström method on the STL-10 dataset.
HybridNet combines unsupervised and supervised learning for image classification.
problem Improving image classifier performance with limited labeled data.
method Two-branch architecture: one branch for supervised learning, the other for unsupervised reconstruction.
result HybridNet outperforms state-of-the-art models on various semi-supervised datasets.
Scale-equivariant CNNs handle scale changes for improved performance.
problem Translation equivariance is not sufficient for handling scale changes in CNNs.
method Developed scale-equivariant convolutional networks with steerable filters.
result Demonstrated state-of-the-art results on MNIST-scale and STL-10 datasets.
BetaDataWeighter learns weights for unlabelled data to improve self-supervised learning accuracy.
problem Improving unsupervised representations with domain shift between unlabelled and target data.
method Learning Bayesian instance weights for unlabelled data to prioritize useful instances.
result BetaDataWeighter achieves highest average accuracy and prunes up to 78% of images without significant loss in accuracy.
Soft Truncation improves diffusion model performance by balancing loss scales across diffusion times.
problem Inverse correlation between density estimation and sample generation in diffusion models.
method Introduces Soft Truncation, a training technique that softens the truncation hyperparameter.
result Soft Truncation achieves state-of-the-art performance on various datasets.
A new mutual information optimization method using self-supervised binary contrastive learning.
problem Improving self-supervised contrastive learning for better model performance.
method Proposes a novel loss function for contrastive learning that optimizes mutual information in positive and negative pairs.
result The proposed method outperforms state-of-the-art self-supervised contrastive frameworks on various benchmark datasets.
Deep neural networks are a powerful tool for feature learning and extraction given their ability to model high-level abstractions in highly complex data. One area worth exploring in feature learning and extraction using deep neural networks is efficient neural connectivity formation for faster feature learning and extr…
One of the challenges in the study of generative adversarial networks is the instability of its training. In this paper, we propose a novel weight normalization technique called spectral normalization to stabilize the training of the discriminator. Our new normalization technique is computationally light and easy to in…
Most existing GANs architectures that generate images use transposed convolution or resize-convolution as their upsampling algorithm from lower to higher resolution feature maps in the generator. We argue that this kind of fixed operation is problematic for GANs to model objects that have very different visual appearan…
Generative Adversarial Networks (GANs) produce systematically better quality samples when class label information is provided., i.e. in the conditional GAN setup. This is still observed for the recently proposed Wasserstein GAN formulation which stabilized adversarial training and allows considering high capacity netwo…
AGAN automates GAN design, outperforming human-designed models.
problem Designing effective GAN architectures requires human expertise and trial-and-error.
method Automated neural architecture search (AGAN) for deep generative models.
result AGAN finds architectures that outperform state-of-the-art models in unsupervised and supervised image generation tasks.
Learning invariant representations is an important problem in machine learning and pattern recognition. In this paper, we present a novel framework of transformation-invariant feature learning by incorporating linear transformations into the feature learning algorithms. For example, we present the transformation-invari…
An important goal in visual recognition is to devise image representations that are invariant to particular transformations. In this paper, we address this goal with a new type of convolutional neural network (CNN) whose invariance is encoded by a reproducing kernel. Unlike traditional approaches where neural networks …
FedMCC learns from distributed data to cluster and extract features.
problem Learning from distributed data and clustering.
method Federated Momentum Contrastive Clustering (FedMCC) framework.
result FedMCC outperforms existing methods in linear evaluation and semi-supervised learning.
MixMatch combines unlabeled data with labeled data to improve semi-supervised learning.
problem Improving semi-supervised learning with limited labeled data.
method MixMatch uses MixUp to blend labeled and unlabeled data, guessing low-entropy labels for augmented examples.
result MixMatch achieves state-of-the-art results across various datasets and labeled data amounts.
We consider the problem of using a factor model we call {\em spike-and-slab sparse coding} (S3C) to learn features for a classification task. The S3C model resembles both the spike-and-slab RBM and sparse coding. Since exact inference in this model is intractable, we derive a structured variational inference procedure …
Meta-Semi learns to optimize SSL with minimal hyper-parameter tuning.
problem Limited labeled data in SSL makes it impractical to tune many hyper-parameters.
method Meta-learning approach that dynamically reweights unlabeled data loss.
result Meta-Semi achieves competitive performance on various SSL tasks.
New method reduces version space for CNNs, improving active learning performance.
problem Sampling bias in active learning hinders optimal hypothesis finding in neural networks.
method Version space reduction through prior mass reduction and diameter reduction, proposing a new Gibbs-vote disagreement method.
result Diameter-based querying method reduces version space more effectively than prior mass reduction and other methods.
We examine two different techniques for parameter averaging in GAN training. Moving Average (MA) computes the time-average of parameters, whereas Exponential Moving Average (EMA) computes an exponentially discounted sum. Whilst MA is known to lead to convergence in bilinear settings, we provide the -- to our knowledge …
New framework improves GAN training by controlling weight spectra.
problem Training GANs is challenging due to instability and poor generalization.
method Proposes a new reparameterization approach for the discriminator's weight matrices to control spectra.
result Spectrum control enhances GANs' generalization ability and image quality.
Improves GAN training with a repulsive loss function.
problem Discourages learning of fine details in data.
method Proposes a repulsive loss function and a bounded Gaussian kernel.
result Significantly improves GAN performance without additional computational cost.
GraN-GAN normalizes gradients for better GAN performance.
problem Improving image generation in GANs with piecewise linear discriminators.
method Piecewise Gradient Normalization (GraN) for input-dependent normalization.
result Significant performance gains in image generation across various datasets.
Kernel VICReg improves SSL in RKHS, capturing nonlinear structures.
problem Limited ability of existing SSL methods to handle nonlinear dependencies.
method Kernel VICReg framework in RKHS, kernelizing VICReg objectives.
result Kernel VICReg mitigates representational collapse and improves performance.
Discriminator optimizes to approximate optimal transport for better image generation.
problem Improving the quality of generated images using GANs.
method Trains discriminator to optimize a lower bound of Wasserstein distance, approximating optimal transport.
result Trained discriminator improves inception score and FID metrics.
This paper introduces GLT for better input data representation in BNN and proposes a compact topology with block pruning.
problem Improving input data representation for Binary Neural Networks (BNN).
method Generic Learned Thermometer (GLT) for encoding, block pruning and Knowledge Distillation for compact topology.
result Significant accuracy gains and lightweight fully-binarized models with limited accuracy degradation.
Proposes ContSup to boost local learning by supplying context between isolated modules.
problem Local learning's performance degrades with more isolated modules.
method Theoretical analysis and ContSup scheme to supply context between modules.
result Significant performance improvement with minimal overhead.
Simplified image clustering achieves competitive results without text-based embeddings.
problem Complexity and resource requirements of state-of-the-art clustering methods.
method SCP: trains a small cluster head using pre-trained vision model features and positive data pairs.
result SCP achieves highly competitive performance on various benchmark datasets.
Improves GANs' generalization by promoting local robustness.
problem Stability and generalization issues in GANs.
method Designs a robust optimization framework where generator and discriminator compete in worst-case setting within a small Wasserstein ball.
result Proves tighter generalization upper bound for RGAN than traditional GANs under mild assumptions.
Scattering networks improve image representation learning without deep learning.
problem Improving image representation learning without deep learning.
method Scattering networks as generic representations in scattering space.
result Scattering networks achieve competitive results in supervised and unsupervised learning.
A new method transfers adversarial robustness from teacher to student using feature distillation.
problem Adversarial robustness transfer across different models and tasks.
method Guided Adversarial Contrastive Distillation (GACD) with contrastive learning and sample reweighted estimation.
result GACD effectively transfers adversarial robustness from teacher to student, achieving comparable or better results.
Motivated by an important insight from neural science, we propose a new framework for understanding the success of the recently proposed "maxout" networks. The framework is based on encoding information on sparse pathways and recognizing the correct pathway at inference time. Elaborating further on this insight, we pro…
We propose in this paper a novel approach to tackle the problem of mode collapse encountered in generative adversarial network (GAN). Our idea is intuitive but proven to be very effective, especially in addressing some key limitations of GAN. In essence, it combines the Kullback-Leibler (KL) and reverse KL divergences …
This paper proposes an alternative to E2E training for deep networks, reducing memory footprint.
problem High GPUs memory footprint in end-to-end training of deep networks.
method Locally supervised learning with information propagation loss to avoid information collapse.
result The proposed method achieves competitive performance with less than 40% memory footprint compared to E2E training.
Recently, deep residual networks have been successfully applied in many computer vision and natural language processing tasks, pushing the state-of-the-art performance with deeper and wider architectures. In this work, we interpret deep residual networks as ordinary differential equations (ODEs), which have long been s…
We propose a new approach to train the Generative Adversarial Nets (GANs) with a mixture of generators to overcome the mode collapsing problem. The main intuition is to employ multiple generators, instead of using a single one as in the original GAN. The idea is simple, yet proven to be extremely effective at covering …
New method proves non-contrastive self-supervised learning learns useful features.
problem Understanding how non-contrastive self-supervised learning (NS-SL) learns useful features.
method Proved in a linear network, NS-SL learns a desirable projection matrix and reduces sample complexity. Suggested weight decay acts as an implicit threshold.
result DirectCopy, a simpler and more efficient algorithm, outperforms DirectPred on various datasets.
VJE learns latent representations without contrastive learning, providing probabilistic semantics.
problem Learning latent representations without contrastive signals.
method VJE maximizes a symmetric conditional evidence lower bound (ELBO) on paired encoder embeddings, using a Student-t distribution on a polar representation.
result VJE outperforms standard non-contrastive baselines in ImageNet-1K, CIFAR-10/100, and STL-10.
Deep neural networks trained using a softmax layer at the top and the cross-entropy loss are ubiquitous tools for image classification. Yet, this does not naturally enforce intra-class similarity nor inter-class margin of the learned deep representations. To simultaneously achieve these two goals, different solutions h…
CDL index improves clustering validation for non-convex data.
problem Selecting clustering algorithms and hyperparameters without labeled data.
method CDL uses compactness, centers, and covariances to compute a probabilistic description length bound.
result CDL outperforms conventional CVIs on synthetic and image benchmarks.
Dependent MMD coresets help compare multiple related datasets.
problem Comparing multiple related datasets for insights into model generalization.
method Dependent MMD coresets for collections of datasets.
result Dependent MMD coresets facilitate comparison and understanding of multiple related datasets.
Synthetic dataset for deep learning with known Gaussian distribution.
problem Lack of datasets with known distribution for deep learning verification.
method Proposes a method to generate a synthetic dataset with Gaussian distribution.
result Synthetic dataset mimics MNIST and can be used with DNNs.
Proposes a method to analyze distributed datasets without sharing original data.
problem Difficulty in centralizing large, distributed datasets due to size and privacy concerns.
method Centralizes intermediate representations instead of original datasets.
result Achieves higher prediction performance compared to individual analyses.
SCARY dataset generates complex causal scenarios for causality research.
problem Lack of complexity in existing causal datasets.
method Synthetic dataset with 40 scenarios, three seeds, and two data generation mechanisms.
result Provides a valuable resource for realistic causal discovery.
MusPy is a toolkit for symbolic music generation, providing tools for dataset management and analysis.
problem Facilitating the creation and analysis of symbolic music datasets.
method Development of an open-source Python library (MusPy) with features for dataset management, data I/O, preprocessing, and model evaluation. Demonstrated through statistical analysis and cross-dataset generalizability experiments.
result MusPy's dataset analysis reveals varying degrees of cross-genre representation across different music datasets.
New handwritten digits dataset for Kannada script.
problem Lack of datasets for Kannada numeral digits.
method Developed Kannada-MNIST and Dig-MNIST datasets.
result Initial CNN accuracy is lower than MNIST, indicating a challenge in generalization.
StyleDiff compares unlabeled datasets using disentangled image spaces.
problem Mismatches between development and real-world datasets lead to inaccurate predictions.
method Uses disentangled image spaces and focuses on attributes to compare datasets.
result Accurately detects and presents differences between datasets.
New framework assesses graph-learning datasets for better evaluation.
problem Insufficient evaluation of graph-learning datasets and methods.
method Introduces Rings framework for dataset ablations and proposes performance separability and mode complementarity measures.
result Demonstrates utility of Rings framework for graph-learning dataset evaluation.
Improved dataset distillation for images and texts boosts model accuracy.
problem Reducing dataset size for faster and more energy-efficient model training.
method Simultaneous distillation of images and soft labels, extending to text datasets.
result 2-4% increase in accuracy for image classification tasks, 20% reduction in distilled samples.