Currently, the neural network architecture design is mostly guided by the indirect metric of computation complexity, i.e., FLOPs. However, the direct metric, e.g., speed, also depends on the other factors such as memory access cost and platform characterics. Thus, this work proposes to evaluate the direct metric on the target platform, beyond only considering FLOPs. Based on a series of controlled experiments, this work derives several practical guidelines for efficient network design. Accordingly, a new architecture is presented, called ShuffleNet V2.Comprehensive ablation experiments verify that our model is the stateof-the-art in terms of speed and accuracy tradeoff.

Lack of transparency in deep neural networks (DNNs) make them susceptible to backdoor attacks, where hidden associations or triggers override normal classification to produce unexpected results. For example, a model with a backdoor always identifies a face as Bill Gates if a specific symbol is present in the input. Backdoors can stay hidden indefinitely until activated by an input, and present a serious security risk to many security or safety related applications, e.g., biometric authentication systems or self-driving cars.We present the first robust and generalizable detection and mitigation system for DNN backdoor attacks. Our techniques identify backdoors and reconstruct possible triggers. We identify multiple mitigation techniques via input filters, neuron pruning and unlearning. We demonstrate their efficacy via extensive experiments on a variety of DNNs, against two types of backdoor injection methods identified by prior work. Our techniques also prove robust against a number of variants of the backdoor attack.

Abstract-The Open Spectrum approach to spectrum access can achieve near-optimal utilization by allowing devices to sense and utilize available spectrum opportunistically. However, a naive distributed spectrum assignment can lead to significant interference between devices. In this paper, we define a general framework that defines the spectrum access problem for several definitions of overall system utility. By reducing the allocation problem to a variant of the graph coloring problem, we show that the global optimization problem is NP-hard, and provide a general approximation methodology through vertex labeling. We examine both a centralized strategy, where a central server calculates an allocation assignment based on global knowledge, and a distributed approach, where devices collaborate to negotiate local channel assignments towards global optimization. Our experimental results show that our allocation algorithms can dramatically reduce interference and improve throughput (as much as 12-fold). Further simulations show that our distributed algorithms generate allocation assignments similar in quality to our centralized algorithms using global knowledge, while incurring substantially less computational complexity in the process.

No abstract