Deep generative models parameterized by neural networks have recently achieved state-of-the-art performance in unsupervised and semi-supervised learning. We extend deep generative models with auxiliary variables which improves the variational approximation. The auxiliary variables leave the generative model unchanged b…
On-device research index
arXiv research
A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.
168,742 papers · 148 categories
Trend · papers per month
37 results for “MNIST-SVHN”
We provide the first experimental results on non-synthetic datasets for the quasi-diagonal Riemannian gradient descents for neural networks introduced in [Ollivier, 2015]. These include the MNIST, SVHN, and FACE datasets as well as a previously unpublished electroencephalogram dataset. The quasi-diagonal Riemannian alg…
The paper shows how integrating categorical semantics can enhance unsupervised domain translation.
problem Improving unsupervised domain translation between perceptually different domains.
method Learning invariant categorical semantic features in an unsupervised manner and conditioning them on the style encoder.
result Conditioning the style encoder on learned categorical semantics improves translation and stylization.
Improves machine learning model training with noisy labels using active learning and clustering.
problem Training machine learning models with noisy oracles and limited data.
method Uses clustering and importance sampling to select batches of samples, incorporates model uncertainty, and introduces a denoising layer.
result Significant improvement in model accuracy on benchmark datasets compared to existing methods.
We introduce a new framework for training deep generative models for high-dimensional conditional density estimation. The Bottleneck Conditional Density Estimator (BCDE) is a variant of the conditional variational autoencoder (CVAE) that employs layer(s) of stochastic variables as the bottleneck between the input a…
The paper introduces a method to quantify uncertainty in neural networks without parametric assumptions.
problem Uncertainty quantification for neural network predictions.
method Nonparametric estimation of conditional label distribution using Nadaraya-Watson kernel.
result The method effectively disentangles aleatoric and epistemic uncertainties.
One-shot neural architecture search limits depth search space and prunes networks for better performance and uncertainty.
problem Finding optimal depth in residual networks for efficient training and inference.
method Formulated a variational objective to approximate the depth distribution and pruned networks based on this distribution.
result Pruned networks achieve competitive accuracy with unpruned networks and better uncertainty calibration.
Improves sampling speed of ARM models without losing accuracy.
problem Slow sampling from ARMs.
method Predictive sampling algorithm using ARM fixed-point iteration and learned forecasting modules.
result Significant improvements in sampling speed and efficiency.
JRFs improve semi-supervised learning by balancing generation and classification.
problem Mode missing and mode covering issues in GANs and VAEs, and conflict between good classification and generation.
method Joint-stochastic-approximation random fields (JRFs) for deep undirected generative models.
result JRFs achieve good classification and generation results in SSL.
Develops a VAE model for datasets with missing data.
problem Applying VAEs to datasets with missing data.
method A novel latent variable model of a corruption process generating missing data, with a tractable ELBO.
result Improved marginal log-likelihood and better missing data imputation compared to existing approaches.
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.
Adversarial examples are typically constructed by perturbing an existing data point within a small matrix norm, and current defense methods are focused on guarding against this type of attack. In this paper, we propose unrestricted adversarial examples, a new threat model where the attackers are not restricted to small…
Improves clustering performance by mixing latent representations.
problem Finding well-defined clusters in data representations.
method Mixing Consistent Deep Clustering method that encourages realistic interpolations and semantic consistency.
result Improved clustering performance across various models and datasets.
Capsule networks are a recently proposed type of neural network shown to outperform alternatives in challenging shape recognition tasks. In capsule networks, scalar neurons are replaced with capsule vectors or matrices, whose entries represent different properties of objects. The relationships between objects and their…
Enhances neural networks' robustness against adversarial samples without sacrificing clean sample generalization.
problem Limited generalization and time complexity of adversarial training.
method Feature Pyramid Decoder (FPD) framework that integrates denoising and image restoration modules into CNNs and constrains the Lipschitz constant.
result FPD-enhanced CNNs achieve sufficient robustness against general adversarial samples on various datasets.
In this paper, we propose a novel unsupervised clustering approach exploiting the hidden information that is indirectly introduced through a pseudo classification objective. Specifically, we randomly assign a pseudo parent-class label to each observation which is then modified by applying the domain specific transforma…
We explore methods of producing adversarial examples on deep generative models such as the variational autoencoder (VAE) and the VAE-GAN. Deep learning architectures are known to be vulnerable to adversarial examples, but previous work has focused on the application of adversarial examples to classification tasks. Deep…
The design of neural network architectures for a new data set is a laborious task which requires human deep learning expertise. In order to make deep learning available for a broader audience, automated methods for finding a neural network architecture are vital. Recently proposed methods can already achieve human expe…
This paper proposes CodeX, an end-to-end framework that facilitates encoding, bitwidth customization, fine-tuning, and implementation of neural networks on FPGA platforms. CodeX incorporates nonlinear encoding to the computation flow of neural networks to save memory. The encoded features demand significantly lower sto…
The success of lottery ticket initializations (Frankle and Carbin, 2019) suggests that small, sparsified networks can be trained so long as the network is initialized appropriately. Unfortunately, finding these "winning ticket" initializations is computationally expensive. One potential solution is to reuse the same wi…
In an increasing number of domains it has been demonstrated that deep learning models can be trained using relatively large batch sizes without sacrificing data efficiency. However the limits of this massive data parallelism seem to differ from domain to domain, ranging from batches of tens of thousands in ImageNet to …
A new algorithm improves GAN training stability and performance.
problem Training GANs is difficult due to variance and rotational dynamics.
method Proposes Lookahead-Minmax algorithm for minmax optimization.
result Significant improvement in GAN performance and stability.
Expectation maximization (EM) algorithm is to find maximum likelihood solution for models having latent variables. A typical example is Gaussian Mixture Model (GMM) which requires Gaussian assumption, however, natural images are highly non-Gaussian so that GMM cannot be applied to perform clustering task on pixel space…
Building on the success of deep learning, two modern approaches to learn a probability model from the data are Generative Adversarial Networks (GANs) and Variational AutoEncoders (VAEs). VAEs consider an explicit probability model for the data and compute a generative distribution by maximizing a variational lower-boun…
New method lures adversaries to choose false directions in black-box attacks.
problem Improving robustness against black-box transfer attacks.
method Incorporates a removable neural network to induce a luring effect.
result Demonstrates effectiveness in thwarting state-of-the-art attacks.
dpVAEs improve VAEs by decoupling representation and generation.
problem VAEs struggle with both representation learning and sample generation.
method Introduce decoupled priors (dpVAEs) that separate representation and generation spaces.
result dpVAEs enable regularization without compromising sample generation.
PoET-BiN reduces power consumption in neural networks on embedded devices.
problem Power inefficiency in neural network implementations on embedded platforms.
method Look-Up Table based implementation with a modified Decision Tree approach.
result Near state-of-the-art results with up to 6 orders of magnitude energy reduction.
Enhances privacy in federated learning with Laplacian smoothing.
problem Protecting data privacy in federated learning while maintaining model accuracy.
method Laplacian smoothing for differentially private federated learning (DP-Fed-LS).
result Improves model accuracy with differential privacy guarantee and membership privacy.
Paper proposes OKGAN to improve GAN training, addressing mode collapse and cycling.
problem Challenges in GAN training, including mode collapse and cycling.
method Kernel-based non-parametric discriminator for online training.
result OKGAN mitigates training issues and performs better than other GAN formulations.
Recently, with convolutional neural networks gaining significant achievements in many challenging machine learning fields, hand-crafted neural networks no longer satisfy our requirements as designing a network will cost a lot, and automatically generating architectures has attracted increasingly more attention and focu…
SketchTransfer task tests deep networks' ability to generalize missing details.
problem Deep networks struggle with generalizing missing details in images.
method Proposes SketchTransfer dataset and task to study detail-invariance.
result State-of-the-art methods perform poorly on SketchTransfer task.
By injecting adversarial examples into training data, adversarial training is promising for improving the robustness of deep learning models. However, most existing adversarial training approaches are based on a specific type of adversarial attack. It may not provide sufficiently representative samples from the adversa…
Conventional application of convolutional neural networks (CNNs) for image classification and recognition is based on the assumption that all target classes are equal(i.e., no hierarchy) and exclusive of one another (i.e., no overlap). CNN-based image classifiers built on this assumption, therefore, cannot take into ac…
TMNs model brain memory systems for continual learning.
problem Catastrophic forgetting in neural networks.
method Triple Network architecture of GANs, incorporating brain-inspired algorithms.
result New state-of-the-art performance on class-incremental learning benchmarks.
For fast and energy-efficient deployment of trained deep neural networks on resource-constrained embedded hardware, each learned weight parameter should ideally be represented and stored using a single bit. Error-rates usually increase when this requirement is imposed. Here, we report large improvements in error rates …
This paper presents a method to train a public model with private data using GANs and differential privacy.
problem Privacy concerns in training deep learning models on sensitive data.
method A three-player learning framework with differential privacy protection.
result The proposed method achieves a balance between privacy and model accuracy.
A new method trains deep neural networks for open set domain adaptation without negative open set difference.
problem Training deep neural networks for open set domain adaptation without negative open set difference.
method Proposes a new upper bound of target-domain risk, including source-domain risk, ε-open set difference (), distributional discrepancy, and constant. Uses gradient descent for source-domain risk and , and adversarial training for distributional discrepancy. Trains DNNs via minimizing the new upper bound.
result Shows state-of-the-art performance on benchmark datasets.