Abstract
Despite remarkable performance in producing realistic samples, Generative Adversarial Networks (GANs) often produce low-quality samples near low-density regions of the data manifold, e.g., samples of minor groups. Many techniques have been developed to improve the quality of generated samples, either by postprocessing generated samples or by pre-processing the empirical data distribution, but at the cost of reduced diversity. To promote diversity in sample generation without degrading the overall quality, we propose a simple yet effective method to diagnose and emphasize underrepresented samples during training of a GAN. The main idea is to use the statistics of the discrepancy between the data distribution and the model distribution at each data instance. Based on the observation that the underrepresented samples have a high average discrepancy or high variability in discrepancy, we propose a method to emphasize those samples during training of a GAN. Our experimental results demonstrate that the proposed method improves GAN performance on various datasets, and it is especially effective in improving the quality and diversity of sample generation for minor groups.
Abstract
The generalization ability of most meta-reinforcement learning (meta-RL) methods is largely limited to test tasks that are sampled from the same distribution used to sample training tasks. To overcome the limitation, we propose Latent Dynamics Mixture (LDM) that trains a reinforcement learning agent with imaginary tasks generated from mixtures of learned latent dynamics. By training a policy on mixture tasks along with original training tasks, LDM allows the agent to prepare for unseen test tasks during training and prevents the agent from overfitting the training tasks. LDM significantly outperforms standard meta-RL methods in test returns on the gridworld navigation and MuJoCo tasks where we strictly separate the training task distribution and the test task distribution.
Abstract
In this paper, we propose a max-min entropy framework for reinforcement learning (RL) to overcome the limitation of the soft actor-critic (SAC) algorithm implementing the maximum entropy RL in model-free sample-based learning. Whereas the maximum entropy RL guides learning for policies to reach states with high entropy in the future, the proposed max-min entropy framework aims to learn to visit states with low entropy and maximize the entropy of these low-entropy states to promote better exploration. For general Markov decision processes (MDPs), an efficient algorithm is constructed under the proposed max-min entropy framework based on disentanglement of exploration and exploitation. Numerical results show that the proposed algorithm yields drastic performance improvement over the current state-of-the-art RL algorithms.
Abstract
Federated learning (FL) presents an appealing opportunity for individuals who are willing to make their private data available for building a communal model without revealing their data contents to anyone else. Of central issues that may limit a widespread adoption of FL is the significant communication resources required in the exchange of updated model parameters between the server and individual clients over many communication rounds. In this work, we focus on limiting the number of model exchange rounds in FL to some small fixed number, to control the communication burden. Following the spirit of meta-learning for few-shot learning, we take a meta-learning strategy to train the model so that once the meta-training phase is over, only rounds of FL would produce a model that will satisfy the needs of all participating clients. A key advantage of employing meta-training is that the main labeled dataset used in training could differ significantly (e.g., different classes of images) from the actual data sample presented at inference time. Compared to the meta-training approaches to optimize personalized local models at distributed devices, our method better handles the potential lack of data variability at individual nodes. Extensive experimental results indicate that meta-training geared to few-round learning provide large performance improvements compared to various baselines.
Abstract
While federated learning (FL) allows efficient model training with local data at edge devices, among major issues still to be resolved are: slow devices known as stragglers and malicious attacks launched by adversaries. While the presence of both of these issues raises serious concerns in practical FL systems, no known schemes or combinations of schemes effectively address them at the same time. We propose Sageflow, staleness-aware grouping with entropy-based filtering and loss-weighted averaging, to handle both stragglers and adversaries simultaneously. Model grouping and weighting according to staleness (arrival delay) provides robustness against stragglers, while entropy-based filtering and loss-weighted averaging, working in a highly complementary fashion at each grouping stage, counter a wide range of adversary attacks. A theoretical bound is established to provide key insights into the convergence behavior of Sageflow. Extensive experimental results show that Sageflow outperforms various existing methods aiming to handle stragglers/adversaries.
Abstract
We tackle a novel few-shot learning challenge, which we call few-shot semantic edge detection, aiming to localize crisp boundaries of novel categories using only a few labeled samples. We also present a Class-Agnostic Few-shot Edge detection Network (CAFENet) based on meta-learning strategy. CAFENet employs a semantic segmentation module in small-scale to compensate for lack of semantic information in edge labels. The predicted segmentation mask is used to generate an attention map to highlight the target object region, and make the decoder module concentrate on that region. We also propose a new regularization method based on multi-split matching. In meta-training, the metric-learning problem with high-dimensional vectors are divided into small subproblems with low-dimensional sub-vectors. Since there is no existing dataset for few-shot semantic edge detection, we construct two new datasets, FSE-1000 and SBD-5 i , and evaluate the performance of the proposed CAFENet on them. Extensive simulation results confirm the performance merits of the techniques adopted in CAFENet.
Abstract
This paper considers wireless communication system consisted of one transmitter, one legitimate receiver, and one eavesdropper. The transmitter transmits perturbation-added signal (i.e. adversarial example) with a certain modulation type, while the legitimate receiver and the eavesdropper adopt deep neural networks (DNN)-based classifier to recognize the modulation type of the received signal. Compared to the fact that the general goal of adversarial examples being a misclassification of all available classifiers, our objective is to design an adversarial example that lets the legitimate receiver classify accurately while the eavesdropper misclassifies. To this end, we propose two design approaches of the adversarial examples: (i) A priori; (ii) A posteriori, i.e. before and after learning steps of the receiver, respectively. Numerical results show that both approaches are effective for securing the communication link. The Korean Institute of Communications and Information Sciences (KICS). Publishing services by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Authors : Sangseok Yun, Jae-Mo Kang, Jeongseok Ha, Sangho Lee, Dong-Woo Ryu, Jihoe Kwon and Il-Min Kim
Journal : IEEE Transactions on Geoscience and Remote Sensing Letters (published: March 2021)
This letter studies deep learning-based efficient ground vibration monitoring systems. In this work, artificial intelligence (AI) techniques are adopted to effectively deal with practical issues of data collection and classification. Specifically, we develop a novel energy-efficient data collection scheme by adopting deep Q-network-based reinforcement learning. Also, we propose an enhanced joint recurrent neural network (RNN) and convolutional neural network (CNN) approach for ground vibration classification. The performance of the proposed scheme is evaluated using real-world ground vibration data. The experimental results show that the proposed classification scheme outperforms the best existing scheme with CNN by more than 13% in terms of classification accuracy. It is also shown that the proposed energy management scheme can improve the accuracy of the proposed ground vibration monitoring system by 7.6% over the comparable scheme using equal power allocation.
Fig. 1 Structure of the proposed ground vibration classification scheme.
Fig. 2. Performance of the proposed deep reinforcement learning-based energy management scheme.
Fig. 3. Performances of the proposed ground vibration classification scheme and the end-to-end ground vibration monitoring system.
Author: Ji Hyung Jung, Hye Won Chung, Ji Oon Lee
Conference and Year: ICML 2021
Keywords: Signal detection, Spiked Rectangular Model
We consider the problem of detecting signals in the rank-one signal-plus-noise data matrix models that generalize the spiked Wishart matrices. We show that the principal component analysis can be improved by pre-transforming the matrix entries if the noise is non-Gaussian. As an intermediate step, we prove a sharp phase transition of the largest eigenvalues of spiked rectangular matrices, which extends the BBP transition. We also propose a hypothesis test to detect the presence of signal with low computational complexity, based on the linear spectral statistics, which minimizes the sum of the Type-I and Type-II errors when the noise is Gaussian.
Figure1: We compare the reconstruction performance of the proposed PCA (top lines) and the standard PCA (bottom lines) for two FashionMNIST images, with the number of measurements N = [3136, 1568, 784, 588, 392] where the data dimension is M = 784. The left most column displays the original images for comparison.
Figure2: The histograms of the test statistic under null hypothesis H0 and alternative hypothesis H1, respectively, for the Gaussian noise with SNR ω = 0.35 and ω = 0.45. It can be shown that the difference of the means of the test statistic under H0 and H1 is larger for ω = 0.45.
Author: Doyeon Kim and Hye Won Chung
Conference and Year: IEEE International Symposium on Information Theory (ISIT), 2021.
We consider crowdsourced labeling under a d-type worker-task specialization model, where each worker and task is associated with one particular type among a finite set of types and a worker provides a more reliable answer to tasks of the matched type than to tasks of unmatched types. We design an inference algorithm that recovers binary task labels (up to any given recovery accuracy) by using worker clustering, worker skill estimation and weighted majority voting. The designed inference algorithm does not require any information about worker/task types, and achieves any targeted recovery accuracy with the best known performance (minimum number of queries per task).
