arXiv subjects

Li Li

Publications and source records attributed to Li Li.

At least 595 records · Page 33Linked to original sources

arXiv2020

Backpressure Control with Estimated Queue Lengths for Urban Network Traffic

Backpressure (BP) control was originally used for packet routing in communications networks. Since its first application to network traffic control, it has undergone different modifications to tailor it to traffic problems with promising results. Most of these BP variants are based on an assumption of perfect knowledge of traffic conditions throughout the network at all times, specifically the queue lengths (more accurately, the traffic volumes). However, it has been well established that accurate queue length information at signalized intersections is never available except in fully connected environments. Although connected vehicle technologies are developing quickly, we are still far from a fully connected environment in the real world. This paper test the effectiveness of BP control when incomplete or imperfect knowledge about traffic conditions is available. We combine BP control with a speed/density field estimation module suitable for a partially connected environment. We refer to the proposed system as a BP with estimated queue lengths (BP-EQ). We test the robustness of BP-EQ to varying levels of connected vehicle penetration, and we compared BP-EQ with the original BP (i.e., assuming accurate knowledge of traffic conditions), a real-world adaptive signal controller, and optimized fixed timing control using microscopic traffic simulation with field calibrated data. Our results show that with a connected vehicle penetration rate as little as 10%, BP-EQ can outperform the adaptive controller and the fixed timing controller in terms of average delay, throughput, and maximum stopped queue lengths under high demand scenarios.

eess.SY↗

arXiv2020

Kohn-Sham equations as regularizer: building prior knowledge into machine-learned physics

Including prior knowledge is important for effective machine learning models in physics, and is usually achieved by explicitly adding loss terms or constraints on model architectures. Prior knowledge embedded in the physics computation itself rarely draws attention. We show that solving the Kohn-Sham equations when training neural networks for the exchange-correlation functional provides an implicit regularization that greatly improves generalization. Two separations suffice for learning the entire one-dimensional H$_2$ dissociation curve within chemical accuracy, including the strongly correlated region. Our models also generalize to unseen types of molecules and overcome self-interaction error.

physics.comp-ph↗

arXiv2020

DLFusion: An Auto-Tuning Compiler for Layer Fusion on Deep Neural Network Accelerator

Many hardware vendors have introduced specialized deep neural networks (DNN) accelerators owing to their superior performance and efficiency. As such, how to generate and optimize the code for the hardware accelerator becomes an important yet less explored problem. In this paper, we perform the compiler-stage optimization study using a novel and representative Cambricon DNN accelerator and demonstrate that the code optimization knobs play an important role in unleashing the potential of hardware computational horsepower. However, even only two studied code optimization knobs, namely the number of cores and layer fusion scheme, present an enormous search space that prevents the naive brute-force search. This work introduces a joint, auto-tuning optimization framework to address this challenge. We first use a set of synthesized DNN layers to study the interplay between the hardware performance and layer characteristics. Based on the insights, we extract the operation count and feature map channel size as each layer's characteristics and derive a joint optimization strategy to decide the performance-optimal core number and fusion scheme. We evaluate the performance of the proposed approach using a set of representative DNN models and show that it achieves the minimal of 3.6x and the maximal of 7.9x performance speedup compared to no optimization baseline. We also show that the achieved speedup is close to the oracle case that is based on a reduced brute-force search but with much less search time.

cs.DC↗

arXiv2020

Learnability and Complexity of Quantum Samples

Given a quantum circuit, a quantum computer can sample the output distribution exponentially faster in the number of bits than classical computers. A similar exponential separation has yet to be established in generative models through quantum sample learning: given samples from an n-qubit computation, can we learn the underlying quantum distribution using models with training parameters that scale polynomial in n under a fixed training time? We study four kinds of generative models: Deep Boltzmann machine (DBM), Generative Adversarial Networks (GANs), Long Short-Term Memory (LSTM) and Autoregressive GAN, on learning quantum data set generated by deep random circuits. We demonstrate the leading performance of LSTM in learning quantum samples, and thus the autoregressive structure present in the underlying quantum distribution from random quantum circuits. Both numerical experiments and a theoretical proof in the case of the DBM show exponentially growing complexity of learning-agent parameters required for achieving a fixed accuracy as n increases. Finally, we establish a connection between learnability and the complexity of generative models by benchmarking learnability against different sets of samples drawn from probability distributions of variable degrees of complexities in their quantum and classical representations.

quant-ph↗

arXiv2020

Learning Data-adaptive Nonparametric Kernels

In this paper, we propose a data-adaptive non-parametric kernel learning framework in margin based kernel methods. In model formulation, given an initial kernel matrix, a data-adaptive matrix with two constraints is imposed in an entry-wise scheme. Learning this data-adaptive matrix in a formulation-free strategy enlarges the margin between classes and thus improves the model flexibility. The introduced two constraints are imposed either exactly (on small data sets) or approximately (on large data sets) in our model, which provides a controllable trade-off between model flexibility and complexity with theoretical demonstration. In algorithm optimization, the objective function of our learning framework is proven to be gradient-Lipschitz continuous. Thereby, kernel and classifier/regressor learning can be efficiently optimized in a unified framework via Nesterov's acceleration. For the scalability issue, we study a decomposition-based approach to our model in the large sample case. The effectiveness of this approximation is illustrated by both empirical studies and theoretical guarantees. Experimental results on various classification and regression benchmark data sets demonstrate that our non-parametric kernel learning framework achieves good performance when compared with other representative kernel learning based algorithms.

cs.LG↗

arXiv2020

Comparison of Cooperative Driving Strategies for CAVs at Signal-Free Intersections

The properties of cooperative driving strategies for planning and controlling Connected and Automated Vehicles (CAVs) at intersections range from some that achieve highly efficient coordination performance to others whose implementation is computationally fast. This paper comprehensively compares the performance of four representative strategies in terms of travel time, energy consumption, computation time, and fairness under different conditions, including the geometric configuration of intersections, asymmetry in traffic arrival rates, and the relative magnitude of these rates. Our simulation-based study has led to the following conclusions: 1) the Monte Carlo Tree Search (MCTS)-based strategy achieves the best traffic efficiency, whereas the Dynamic Resequencing (DR)-based strategy is energy-optimal; both strategies perform well in all metrics of interest. If the computation budget is adequate, the MCTS strategy is recommended; otherwise, the DR strategy is preferable; 2) An asymmetric intersection has a noticeable impact on the strategies, whereas the influence of the arrival rates can be neglected. When the geometric shape is asymmetrical, the modified First-In-First-Out (FIFO) strategy significantly outperforms the FIFO strategy and works well when the traffic demand is moderate, but their performances are similar in other situations; and 3) Improving traffic efficiency sometimes comes at the cost of fairness, but the DR and MCTS strategies can be adjusted to realize a better trade-off between various performance metrics by appropriately designing their objective functions.

eess.SY↗

arXiv2020

A General Framework for Decentralized Safe Optimal Control of Connected and Automated Vehicles in Multi-Lane Intersections

We address the problem of optimally controlling Connected and Automated Vehicles (CAVs) arriving from four multi-lane roads at an intersection where they conflict in terms of safely crossing (including turns) with no collision. The objective is to jointly minimize the travel time and energy consumption of each CAV while ensuring safety. This problem was solved in prior work for single-lane roads. A direct extension to multiple lanes on each road is limited by the computational complexity required to obtain an explicit optimal control solution. Instead, we propose a general framework that first converts a multi-lane intersection problem into a decentralized optimal control problem for each CAV with less conservative safety constraints than prior work. We then employ a method combining optimal control and control barrier functions, which has been shown to efficiently track tractable unconstrained optimal CAV trajectories while also guaranteeing the satisfaction of all constraints. Simulation examples are included to show the effectiveness of the proposed framework under symmetric and asymmetric intersection geometries and different schemes for CAV sequencing.

eess.SY↗

arXiv2020

Cloning of Quantum Entanglement

Quantum no-cloning, the impossibility of perfectly cloning an arbitrary unknown quantum state, is one of the most fundamental limitations due to the laws of quantum mechanics, which underpin the physical security of quantum key distribution. Quantum physics does allow, however, approximate cloning with either imperfect state fidelity and/or probabilistic success. Whereas approximate quantum cloning of single-particle states has been tested previously, experimental cloning of quantum entanglement -- a highly non-classical correlation -- remained unexplored. Based on a multiphoton linear optics platform, we demonstrate quantum cloning of two photon entangled states for the first time. Remarkably our results show that one maximally entangled photon pair can be broadcast into two entangled pairs, both with state fidelities above 50\%. Our results are a key step towards cloning of complex quantum systems, and are likely to provide new insights into quantum entanglement.

quant-ph↗

arXiv2020

Open-destination measurement-device-independent quantum key distribution network

Quantum key distribution (QKD) networks hold promise for sharing secure randomness over multi-partities. Most existing QKD network schemes and demonstrations are based on trusted relays or limited to point-to-point scenario. Here, we propose a flexible and extensible scheme named as open-destination measurement-device-independent QKD network. The scheme enjoys security against untrusted relays and all detector side-channel attacks. Particularly, any users can accomplish key distribution under assistance of others in the network. As an illustration, we show in detail a four-user network where two users establish secure communication and present realistic simulations by taking into account imperfections of both sources and detectors.

quant-ph↗

arXiv2020

Referenceless Rate-Distortion Modeling with Learning from Bitstream and Pixel Features

Generally, adaptive bitrates for variable Internet bandwidths can be obtained through multi-pass coding. Referenceless prediction-based methods show practical benefits compared with multi-pass coding to avoid excessive computational resource consumption, especially in low-latency circumstances. However, most of them fail to predict precisely due to the complex inner structure of modern codecs. Therefore, to improve the fidelity of prediction, we propose a referenceless prediction-based R-QP modeling (PmR-QP) method to estimate bitrate by leveraging a deep learning algorithm with only one-pass coding. It refines the global rate-control paradigm in modern codecs on flexibility and applicability with few adjustments as possible. By exploring the potentials of bitstream and pixel features from the prerequisite of one-pass coding, it can reach the expectation of bitrate estimation in terms of precision. To be more specific, we first describe the R-QP relationship curve as a robust quadratic R-QP modeling function derived from the Cauchy-based distribution. Second, we simplify the modeling function by fastening one operational point of the relationship curve received from the coding process. Third, we learn the model parameters from bitstream and pixel features, named them hybrid referenceless features, comprising texture information, hierarchical coding structure, and selected modes in intra-prediction. Extensive experiments demonstrate the proposed method significantly decreases the proportion of samples' bitrate estimation error within 10% by 24.60% on average over the state-of-the-art.

eess.IV↗

arXiv2020

Investigating Quantum Approximate Optimization Algorithms under Bang-bang Protocols

The quantum approximate optimization algorithm (QAOA) is widely seen as a possible usage of noisy intermediate-scale quantum (NISQ) devices. We analyze the algorithm as a bang-bang protocol with fixed total time and a randomized greedy optimization scheme. We investigate the performance of bang-bang QAOA on MAX-2-SAT, finding the appearance of phase transitions with respect to the total time. As the total time increases, the optimal bang-bang protocol experiences a number of jumps and plateaus in performance, which match up with an increasing number of switches in the standard QAOA formulation. At large times, it becomes more difficult to find a globally optimal bang-bang protocol and performances suffer. We also investigate the effects of changing the initial conditions of the randomized optimization algorithm and see that better local optima can be found by using an adiabatic initialization.

quant-ph↗

arXiv2020

Quantum teleportation of physical qubits into logical code-spaces

Quantum error correction is an essential tool for reliably performing tasks for processing quantum information on a large scale. However, integration into quantum circuits to achieve these tasks is problematic when one realizes that non-transverse operations, which are essential for universal quantum computation, lead to the spread of errors. Quantum gate teleportation has been proposed as an elegant solution for this. Here, one replaces these fragile, non-transverse inline gates with the generation of specific, highly entangled offline resource states that can be teleported into the circuit to implement the non-transverse gate. As the first important step, we create a maximally entangled state between a physical and an error-correctable logical qubit and use it as a teleportation resource. We then demonstrate the teleportation of quantum information encoded on the physical qubit into the error-corrected logical qubit with fidelities up to 0.786. Our scheme can be designed to be fully fault-tolerant so that it can be used in future large-scale quantum technologies.

quant-ph↗

arXiv2020

Constraining Non-Relativistic RG Flows with Holography

We examine non-relativistic holographic RG flows by working with Einstein-Maxwell-scalar theories which support geometries that break Lorentz invariance at some energy scale. We adopt the superpotential formalism, which helps us characterize the radial flow in this setup and bring to light a number of generic features. In particular, we identify several quantities that behave monotonically under RG flow. As an example, we show that the index of refraction is generically monotonic. We also construct a combination of the superpotentials that flows monotonically in Einstein-scalar theories supporting non-relativistic solutions, and which reduces to the known c-function in the relativistic limit. Interestingly, such quantity also exhibits monotonicity in a variety of black hole solutions to the full Einstein-Maxwell-scalar theory, hinting at a deeper structure. Finally, we comment on the breakdown of such monotonicity conditions and on the relation to a candidate c-function obtained previously from entanglement entropy.

hep-th↗

arXiv2020

Magnetotransport and Complexity of Holographic Metal-Insulator Transitions

We study the magnetotransport in a minimal holographic setup of a metal-insulator transition in two spatial dimensions. Some generic features are obtained without referring to the non-linear details of the holographic theory. The temperature dependence of resistivity is found to be well scaled with a single parameter $T_0$, which approaches zero at some critical charge density $ρ_c$, and increases as a power law $T_0\sim|ρ-ρ_c|^{1/2}$ both in metallic $(ρ>ρ_c)$ and insulating $(ρ<ρ_c)$ regions in the vicinity of the transition. Similar features also happen by changing the disorder strength as well as magnetic field. By requiring a positive definite longitudinal conductivity in the presence of an applied magnetic field restricts the allowed parameter space of theory parameters. We explicitly check the consistency of parameter range for two representative models, and compute the optical conductivities for both metallic and insulating phases, from which a disorder-induced transfer of spectral weight from low to high energies is manifest. We construct the phase diagram in terms of temperature and disorder strength. The complexity during the transition is studied and is found to be not a good probe to the metal-insulator transition.

hep-th↗

arXiv2020

Newton polytopes of rank 3 cluster variables

We characterize the cluster variables of skew-symmetrizable cluster algebras of rank 3 by their Newton polytopes. The Newton polytope of the cluster variable $z$ is the convex hull of the set of all $\mathbf{p}\in\mathbb{Z}^3$ such that the Laurent monomial ${\bf x}^{\mathbf{p}}$ appears with nonzero coefficient in the Laurent expansion of $z$ in the cluster ${\bf x}$. We give an explicit construction of the Newton polytope in terms of the exchange matrix and the denominator vector of the cluster variable. Along the way, we give a new proof of the fact that denominator vectors of non-initial cluster variables are non-negative in a cluster algebra of arbitrary rank.

math.AC↗

arXiv2020

Explainable Recommender Systems via Resolving Learning Representations

Recommender systems play a fundamental role in web applications in filtering massive information and matching user interests. While many efforts have been devoted to developing more effective models in various scenarios, the exploration on the explainability of recommender systems is running behind. Explanations could help improve user experience and discover system defects. In this paper, after formally introducing the elements that are related to model explainability, we propose a novel explainable recommendation model through improving the transparency of the representation learning process. Specifically, to overcome the representation entangling problem in traditional models, we revise traditional graph convolution to discriminate information from different layers. Also, each representation vector is factorized into several segments, where each segment relates to one semantic aspect in data. Different from previous work, in our model, factor discovery and representation learning are simultaneously conducted, and we are able to handle extra attribute information and knowledge. In this way, the proposed model can learn interpretable and meaningful representations for users and items. Unlike traditional methods that need to make a trade-off between explainability and effectiveness, the performance of our proposed explainable model is not negatively affected after considering explainability. Finally, comprehensive experiments are conducted to validate the performance of our model as well as explanation faithfulness.

cs.LG↗

arXiv2020

A Framework for Behavioral Biometric Authentication using Deep Metric Learning on Mobile Devices

Mobile authentication using behavioral biometrics has been an active area of research. Existing research relies on building machine learning classifiers to recognize an individual's unique patterns. However, these classifiers are not powerful enough to learn the discriminative features. When implemented on the mobile devices, they face new challenges from the behavioral dynamics, data privacy and side-channel leaks. To address these challenges, we present a new framework to incorporate training on battery-powered mobile devices, so private data never leaves the device and training can be flexibly scheduled to adapt the behavioral patterns at runtime. We re-formulate the classification problem into deep metric learning to improve the discriminative power and design an effective countermeasure to thwart side-channel leaks by embedding a noise signature in the sensing signals without sacrificing too much usability. The experiments demonstrate authentication accuracy over 95% on three public datasets, a sheer 15% gain from multi-class classification with less data and robustness against brute-force and side-channel attacks with 99% and 90% success, respectively. We show the feasibility of training with mobile CPUs, where training 100 epochs takes less than 10 mins and can be boosted 3-5 times with feature transfer. Finally, we profile memory, energy and computational overhead. Our results indicate that training consumes lower energy than watching videos and slightly higher energy than playing games.

cs.LG↗

arXiv2020

Defending Adversarial Examples via DNN Bottleneck Reinforcement

This paper presents a DNN bottleneck reinforcement scheme to alleviate the vulnerability of Deep Neural Networks (DNN) against adversarial attacks. Typical DNN classifiers encode the input image into a compressed latent representation more suitable for inference. This information bottleneck makes a trade-off between the image-specific structure and class-specific information in an image. By reinforcing the former while maintaining the latter, any redundant information, be it adversarial or not, should be removed from the latent representation. Hence, this paper proposes to jointly train an auto-encoder (AE) sharing the same encoding weights with the visual classifier. In order to reinforce the information bottleneck, we introduce the multi-scale low-pass objective and multi-scale high-frequency communication for better frequency steering in the network. Unlike existing approaches, our scheme is the first reforming defense per se which keeps the classifier structure untouched without appending any pre-processing head and is trained with clean images only. Extensive experiments on MNIST, CIFAR-10 and ImageNet demonstrate the strong defense of our method against various adversarial attacks.

cs.CV↗