Learning against label noise is a vital topic to guarantee a reliable performance for deep neural networks.Recent research usually refers to dynamic noise modeling with model output probabilities and loss values, and then separates clean and noisy samples.These methods have gained notable success. However, unlike cherry-picked data, existing approaches often cannot perform well when facing imbalanced datasets, a common scenario in the real world.We thoroughly investigate this phenomenon and point out two major issues that hinder the performance, i.e., inter-class loss distribution discrepancy and misleading predictions due to uncertainty.The first issue is that existing methods often perform class-agnostic noise modeling. However, loss distributions show a significant discrepancy among classes under class imbalance, and class-agnostic noise modeling can easily get confused with noisy samples and samples in minority classes.The second issue refers to that models may output misleading predictions due to epistemic uncertainty and aleatoric uncertainty, thus existing methods that rely solely on the output probabilities may fail to distinguish confident samples. Inspired by our observations, we propose an Uncertainty-aware Label Correction framework(ULC) to handle label noise on imbalanced datasets. First, we perform epistemic uncertainty-aware class-specific noise modeling to identify trustworthy clean samples and refine/discard highly confident true/corrupted labels.Then, we introduce aleatoric uncertainty in the subsequent learning process to prevent noise accumulation in the label noise modeling process. We conduct experiments on several synthetic and real-world datasets. The results demonstrate the effectiveness of the proposed method, especially on imbalanced datasets.

Microgrid (MG) is a promising component for future smart grid (SG) deployment. The balance of supply and demand of electric energy is one of the most important requirements of MG management. In this paper, we present a novel framework for smart energy management based on the concept of quality-of-service in electricity (QoSE). Specifically, the resident electricity demand is classified into basic usage and quality usage. The basic usage is always guaranteed by the MG, while the quality usage is controlled based on the MG state. The microgrid control center (MGCC) aims to minimize the MG operation cost and maintain the outage probability of quality usage, i.e., QoSE, below a target value, by scheduling electricity among renewable energy resources, energy storage systems, and macrogrid. The problem is formulated as a constrained stochastic programming problem. The Lyapunov optimization technique is then applied to derive an adaptive electricity scheduling algorithm by introducing the QoSE virtual queues and energy storage virtual queues. The proposed algorithm is an online algorithm since it does not require any statistics and future knowledge of the electricity supply, demand and price processes. We derive several "hard" performance bounds for the proposed algorithm, and evaluate its performance with trace-driven simulations. The simulation results demonstrate the efficacy of the proposed electricity scheduling algorithm.

ABSTRACT In this paper, we study the problem of power allocation for streaming multiple variable‐bit‐rate (VBR) videos in the downlink of a cellular network. We consider a deterministic model for VBR video traffic and finite playout buffer at the mobile users. The objective is to derive the optimal downlink power allocation for the VBR video sessions, such that the video data can be delivered in a timely fashion without causing playout buffer overflow and underflow. The formulated problem is a nonlinear nonconvex optimization problem. We analyze the convexity conditions for the formulated problem and propose a two‐step greedy approach to solve the problem. We also develop a distributed algorithm based on the dual decomposition technique, which can be incorporated into the two‐step solution procedure. The performance of the proposed algorithms is validated with simulations using VBR video traces under realistic scenarios. Copyright © 2012 John Wiley & Sons, Ltd.

We investigate the problem of downlink power control for streaming multiple variable bit rate (VBR) videos in a multicell wireless network, where downlink capacities are limited by inter-cell interference. We adopt a deterministic model for VBR video traffic that considers video frame sizes and playout buffers at the mobile users. The problem is to find the optimal transmit powers for the base stations, such that VBR video data can be delivered to mobile users without causing playout buffer underflow or overflow. We formulate a nonlinear nonconvex optimization problem and prove the condition for the existence of feasible solutions. A centralized branch-and-bound algorithm is then developed, which incorporates the Reformulation-Linearization Technique and can produce (1-ε)-optimal solutions. We also propose a low-complexity distributed algorithm with fast convergence as an alternative to the centralized algorithm. Through simulations with VBR video traces under fading channels, we find the distributed algorithm can achieve a performance very close to that of the centralized algorithm.

Microgrid (MG) is a key component for future smart grid (SG) deployment with high potentials. Balancing the supply and demand of energy is one of the most important goals of MG management. In this paper, we explore effective schemes for quality-of-usage (QoU) guarantees for local residents in an MG, under randomness in both electricity supply and demand. The microgrid control center (MGCC) aims to maintain the QoU blocking probability around a target value by serving or blocking QoU requests. The problem is formulated as a queue stability problem by introducing the concept of a QoU blocking virtual queue. The Lyapunov optimization technique is then applied to derive an adaptive QoU algorithm with complexity O(1). Furthermore, the proposed algorithm is an online algorithm since it does not require any future knowledge of the system. The stability of the proposed algorithm is proven, and its performance is evaluated with trace-driven simulations under random QoU requests. The simulation results demonstrate the efficacy and robustness of the proposed algorithm.

In this paper, we investigate the problem of optimal power control for multiuser variable bit rate (VBR) video streaming in a cellular network with orthogonal channels. We adopt a deterministic model for VBR video traffic that incorporates video frame and playout buffer characteristics, and formulate a constrained stochastic optimization problem. We then develop a majorization-based solution approach. For the case of a single VBR video session with relaxed peak power constraint, we develop a power optimal algorithm with low complexity. We prove the power optimality of the proposed algorithm and the uniqueness of the global optimum, and demonstrate that the proposed algorithm is also smoothness optimal. For the case of multiuser VBR video streaming, we develop a heuristic algorithm that selectively suspends some video sessions when the peak power constraint is violated. The proposed algorithms are evaluated with trace-driven simulations, and are shown to achieve considerable power savings and improved video quality over a conventional "lazy" scheme.

We investigate the problem of downlink power control for streaming multiple variable bit rate (VBR) videos in a multicell wireless network, where downlink capacities are limited by inter-cell interference. We adopt a deterministic model for VBR traffic that considers video frame sizes and playout buffers at the mobile users. The problem is to find the optimal transmit powers for the base stations, such that VBR video data can be delivered to mobile users without causing playout buffer underflow or overflow. We formulate a nonlinear nonconvex optimization problem and prove the condition for the existence of feasible solutions. We then develop a centralized branch-and-bound algorithm incorporating the Reformulation-Linearization Technique, which can produce (1-ε)-optimal solutions. We also propose a low-complexity distributed algorithm with fast convergence. Through simulations with VBR video traces under fading channels, we find the distributed algorithm can achieve a performance very close to that of the centralized algorithm.

In this paper, we study the problem of rate control for streaming videos by jointly considering encoder rate control and network congestion control. We adopt a control-theoretic approach that models video streaming as a feedback control system. Based on a properly chosen operating point, the model is linearized and a proportional-integral (PI) controller is designed to stabilize the streaming video quality. We derive the guidelines for choosing parameters for the proposed PI rate controller and prove its stability properties. We also show that the PI rate controller is highly robust to fluctuations in the bottleneck link capacity. Our simulation results verify the accuracy of the analysis and demonstrate the efficacy of the proposed control-theoretic approach.

As streaming videos are becoming increasingly popular, it is important to understand the end-to-end streaming system and to develop effective algorithms for quality control. In this paper, we address the problem of rate control for streaming videos with a control-theoretic approach. Among the various control knobs, video bit rate is one of the most effective in the sense that it has a direct impact on the interaction between the video coder and network system. While increasing rate reduces the coder-induced distortion, it may also cause congestion at a bottleneck link. The packet loss due to congestion will, then, increase the distortion of the decoded video. We model end-to-end video steaming as a feedback control system, taking into account video codec and sequence characteristics, rate control, active queue management, and receiver feedback. We then develop effective proportional (P) controllers to stabilize the received video quality as well as the bottleneck link queue, for both homogeneous and heterogeneous video systems. Simulation results are presented to demonstrate the efficacy of thePcontrollers and the viability of the proposed control-theoretic approach.