arXiv subjects

Yu Yang

Publications and source records attributed to Yu Yang.

At least 271 records · Page 15Linked to original sources

arXiv2019

ReLeTA: Reinforcement Learning for Thermal-Aware Task Allocation on Multicore

In this paper, we propose \textit{ReLeTA}: Reinforcement Learning based Task Allocation for temperature minimization. We design a new reward function and use a new state model to facilitate optimization of reinforcement learning algorithm. By means of the new reward function and state model, \releta is able to effectively reduce the system peak temperature without compromising the application performance. We implement and evaluate \releta on a real platform in comparison with the state-of-the-art approaches. Experimental results show \releta can reduce the average peak temperature by 4 $^{\circ}$C and the maximum difference is up to 13 $^{\circ}$C.

eess.SY↗

arXiv2019

eBrainII: A 3 kW Realtime Custom 3D DRAM integrated ASIC implementation of a Biologically Plausible Model of a Human Scale Cortex

The Artificial Neural Networks (ANNs) like CNN/DNN and LSTM are not biologically plausible and in spite of their initial success, they cannot attain the cognitive capabilities enabled by the dynamic hierarchical associative memory systems of biological brains. The biologically plausible spiking brain models, for e.g. cortex, basal ganglia and amygdala have a greater potential to achieve biological brain like cognitive capabilities. Bayesian Confidence Propagation Neural Network (BCPNN) is a biologically plausible spiking model of cortex. A human scale model of BCPNN in real time requires 162 TFlops/s, 50 TBs of synaptic weight storage to be accessed with a bandwidth of 200 TBs. The spiking bandwidth is relatively modest at 250 GBs/s. A hand optimized implementation of rodent scale BCPNN has been implemented on Tesla K80 GPUs require 3 kW, we extrapolate from that a human scale network will require 3 MW. These power numbers rule out such implementations for field deployment as advanced cognition engines in embedded systems. The key innovation that this paper reports is that it is feasible and affordable to implement real time BCPNN as a custom tiled ASIC in 28 nm technology with custom 3D DRAM - eBrain II - that consumes 3 kWs for human scale and 12 W for rodent scale cortex model. Such implementations eminently fulfill the demands for field deployment.

cs.DC↗

arXiv2019

The complexity of total edge domination and some related results on trees

For a graph $G = (V, E)$ with vertex set $V$ and edge set $E$, a subset $F$ of $E$ is called an $\emph{edge dominating set}$ (resp. a $\emph{total edge dominating set}$) if every edge in $E\backslash F$ (resp. in $E$) is adjacent to at least one edge in $F$, the minimum cardinality of an edge dominating set (resp. a total edge dominating set) of $G$ is the {\em edge domination number} (resp. {\em total edge domination number}) of $G$, denoted by $γ^{'}(G)$ (resp. $γ_t^{'}(G)$). In the present paper, we prove that the total edge domination problem is NP-complete for bipartite graphs with maximum degree 3. We also design a linear-time algorithm for solving this problem for trees. Finally, for a graph $G$, we give the inequality $γ^{'}(G)\leqslant γ^{'}_{t}(G)\leqslant 2γ^{'}(G)$ and characterize the trees $T$ which obtain the upper or lower bounds in the inequality.

math.CO↗

arXiv2019

Finding Theme Communities from Database Networks

Given a database network where each vertex is associated with a transaction database, we are interested in finding theme communities. Here, a theme community is a cohesive subgraph such that a common pattern is frequent in all transaction databases associated with the vertices in the subgraph. Finding all theme communities from a database network enjoys many novel applications. However, it is challenging since even counting the number of all theme communities in a database network is #P-hard. Inspired by the observation that a theme community shrinks when the length of the pattern increases, we investigate several properties of theme communities and develop TCFI, a scalable algorithm that uses these properties to effectively prune the patterns that cannot form any theme community. We also design TC-Tree, a scalable algorithm that decomposes and indexes theme communities efficiently. Retrieving a ranked list of theme communities from a TC-Tree of hundreds of millions of theme communities takes less than 1 second. Extensive experiments and a case study demonstrate the effectiveness and scalability of TCFI and TC-Tree in discovering and querying meaningful theme communities from large database networks.

cs.DB↗

arXiv2019

Dynamic Filtering with Large Sampling Field for ConvNets

We propose a dynamic filtering strategy with large sampling field for ConvNets (LS-DFN), where the position-specific kernels learn from not only the identical position but also multiple sampled neighbor regions. During sampling, residual learning is introduced to ease training and an attention mechanism is applied to fuse features from different samples. Such multiple samples enlarge the kernels' receptive fields significantly without requiring more parameters. While LS-DFN inherits the advantages of DFN, namely avoiding feature map blurring by position-wise kernels while keeping translation invariance, it also efficiently alleviates the overfitting issue caused by much more parameters than normal CNNs. Our model is efficient and can be trained end-to-end via standard back-propagation. We demonstrate the merits of our LS-DFN on both sparse and dense prediction tasks involving object detection, semantic segmentation, and flow estimation. Our results show LS-DFN enjoys stronger recognition abilities in object detection and semantic segmentation tasks on VOC benchmark and sharper responses in flow estimation on FlyingChairs dataset compared to strong baselines.

cs.CV↗

arXiv2019

The geometry properties of parity and time reversal operators in two dimensional spaces

The parity operator $\cal P$ and time reversal operator $\cal T$ are two important operators in the quantum theory, in particular, in the $\cal PT$-symmetric quantum theory. By using the concrete forms of $\cal P$ and $\cal T$, we discuss their geometrical properties in two dimensional spaces. It is showed that if $\cal T$ is given, then all $\cal P$ links with the quadric surfaces; if $\cal P$ is given, then all $\cal T$ links with the quadric curves. Moreover, we give out the generalized unbroken $\cal PT$-symmetric condition of an operator. The unbroken $\cal PT$-symmetry of a Hermitian operator is also showed in this way.

quant-ph↗

arXiv2019

Accelerated Stochastic Algorithms for Nonconvex Finite-sum and Multi-block Optimization

In this paper, we present new stochastic methods for solving two important classes of nonconvex optimization problems. We first introduce a randomized accelerated proximal gradient (RapGrad) method for solving a class of nonconvex optimization problems consisting of the sum of $m$ component functions, and show that it can significantly reduce the number of gradient computations especially when the condition number $L/μ$ (i.e., the ratio between the Lipschitz constant and negative curvature) is large. More specifically, RapGrad can save up to ${\cal O}(\sqrt{m})$ gradient computations than existing deterministic nonconvex accelerated gradient methods. Moreover, the number of gradient computations required by RapGrad can be ${\cal O}(m^\frac{1}{6} L^\frac{1}{2} / μ^\frac{1}{2})$ (at least ${\cal O}(m^\frac{2}{3})$) times smaller than the best-known randomized nonconvex gradient methods when $L/μ\ge m$. Inspired by RapGrad, we also develop a new randomized accelerated proximal dual (RapDual) method for solving a class of multi-block nonconvex optimization problems coupled with linear constraints. We demonstrate that RapDual can also save up to a factor of ${\cal O}(\sqrt{m})$ projection subproblems than its deterministic counterpart, where $m$ denotes the number of blocks. To the best of our knowledge, all these complexity results associated with RapGrad and RapDual seem to be new in the literature. We also illustrate potential advantages of these algorithms through our preliminary numerical experiments.

math.OC↗

arXiv2019

Two-parameter Hong-Ou-Mandel dip

A modification of the standard Hong-Ou-Mandel interferometer is proposed which allows one to replicate the celebrated coincidence dip in the case of two-independent delay parameters. In the ideal case where such delays are sufficiently stable with respect to the mean wavelength of the pump source, properly symmetrized input bi-photon states allow one to pinpoint their values through the identification of a zero in the coincidence counts, a feature that cannot be simulated by semiclassical inputs having the same spectral properties. Besides, in the presence of fluctuating parameters the zero in the coincidences is washed away: still the bi-photon state permits to recover the values of parameters with a visibility which is higher than the one allowed by semiclassical sources. The detrimental role of loss and dispersion is also analyzed and an application in the context of quantum positioning is presented.

quant-ph↗

arXiv2019

Exclusive Hong-Ou-Mandel zero-coincidence point

A generalized multi-parameter Hong-Ou-Mandel interferometer is presented which extends the conventional "Mandel dip" configuration to the case where a symmetric biphoton source is used to monitor the contemporary presence of k independent time-delays. Our construction results in a two-input/two-output setup, obtained by concatenating 50:50 beam splitters with a collection of adjustable achromatic wave-plates. For k=1,2 and k=4 explicit examples can be exhibited that prove the possibility of uniquely linking the zero value of the coincidence counts registered at the output of the interferometer, with the contemporary absence of all the time-delays. Interestingly enough the same result cannot be extended to k=3. Besides, the sensitivity of the interferometer is analyzed when the time-delays are affected by the fluctuations over time-scales that are larger than the inverse of the frequency of the pump used to generate the biphoton state.

quant-ph↗

arXiv2019

HVAC Energy Cost Optimization for a Multi-zone Building via a Decentralized Approach

It has been well acknowledged that buildings account for a large proportion of the world's energy consumption. However, the energy use of buildings, especially the heating, ventilation and air-conditioning (HVAC), is far from being efficient. There still exists a dramatic potential to save energy through improving building energy efficiency. Therefore, this paper studies the control of HVAC system for multi-zone buildings with the objective to reduce energy consumption cost while satisfying thermal comfort. In particular, the thermal couplings due to the heat transfer between the adjacent zones are incorporated in the optimization. Considering that a centralized method is generally computationally prohibitive for large buildings, an efficient decentralized approach is developed, based on the Accelerated Distributed Augmented Lagrangian (ADAL) method [1]. To evaluate the performance of the proposed method, we first compare it with a centralized method, in which the optimal solution of a small-scale problem can be obtained. We find that this decentralized approach can almost approach the optimal solution of the problem. Further, this decentralized approach is compared with the Distributed Token-Based Scheduling Strategy (DTBSS) [2]. The numeric results reveal that when the number of zones is relatively small (less than 20), the two decentralized methods can achieve a comparable performance regarding the cost of the HVAC system. However, with an increase of the number of zones in the building, the proposed decentralized approach demonstrates better performance with a considerable reduction of the total cost. Moreover, the decentralized approach proposed in this paper demonstrate better scalability with less average computation required.

eess.SY↗

arXiv2019

A Generative Model for Sampling High-Performance and Diverse Weights for Neural Networks

Recent work on mode connectivity in the loss landscape of deep neural networks has demonstrated that the locus of (sub-)optimal weight vectors lies on continuous paths. In this work, we train a neural network that serves as a hypernetwork, mapping a latent vector into high-performance (low-loss) weight vectors, generalizing recent findings of mode connectivity to higher dimensional manifolds. We formulate the training objective as a compromise between accuracy and diversity, where the diversity takes into account trivial symmetry transformations of the target network. We demonstrate how to reduce the number of parameters in the hypernetwork by parameter sharing. Once learned, the hypernetwork allows for a computationally efficient, ancestral sampling of neural network weights, which we recruit to form large ensembles. The improvement in classification accuracy obtained by this ensembling indicates that the generated manifold extends in dimensions other than directions implied by trivial symmetries. For computational efficiency, we distill an ensemble into a single classifier while retaining generalization.

stat.ML↗

arXiv2019

Interpreting CNNs via Decision Trees

This paper aims to quantitatively explain rationales of each prediction that is made by a pre-trained convolutional neural network (CNN). We propose to learn a decision tree, which clarifies the specific reason for each prediction made by the CNN at the semantic level. I.e., the decision tree decomposes feature representations in high conv-layers of the CNN into elementary concepts of object parts. In this way, the decision tree tells people which object parts activate which filters for the prediction and how much they contribute to the prediction score. Such semantic and quantitative explanations for CNN predictions have specific values beyond the traditional pixel-level analysis of CNNs. More specifically, our method mines all potential decision modes of the CNN, where each mode represents a common case of how the CNN uses object parts for prediction. The decision tree organizes all potential decision modes in a coarse-to-fine manner to explain CNN predictions at different fine-grained levels. Experiments have demonstrated the effectiveness of the proposed method.

cs.CV↗

arXiv2019

Raman spectrum of 1T'-WTe2 under tensile strain: A first-principles prediction

Monolayer WTe2 attracts rapidly growing interests for its large-gap quantum spin Hall effect,which enables promising apllications in flexible logic devices. Due to one-dimensional W-W chains,1T'-WTe2 exhibits unique anisotropic structure and promising properties, which can be modified by simply applying strains. Based on the first-principles simulations, we show that phonon branch undergoes soft down to negative frequency at special q points under different critical strains, i.e., epsilon_a = 11.55 percent along a-axis (with W-W chains) direction, epsilon_b = 7.0 percent along b-axis direction and epsilon_ab = 8.44 percent along biaxial direction. Before each critical strain, the Raman-shift of A1_g, A3_g, and A4_g modes, corresponding to the main peaks in Raman spectra of 1T'-WTe2 , shows anisotropic response to uniaxial strain but most sensitive to biaxial strain. Interestingly, we find that the frequency shift of A3_g mode show parabolic characters of strained 1T'-WTe2, then we split it into two parts and it shows a Raman-shift transition at about 5 percent strains. While for the A1_g and A4_g modes,the frequencies change linearly. Through careful structure and vibration analysis, we try to explain these Raman irregularity in strained 1T'-WTe2.

cond-mat.mtrl-sci↗

arXiv2019

Network Transplanting (extended abstract)

This paper focuses on a new task, i.e., transplanting a category-and-task-specific neural network to a generic, modular network without strong supervision. We design a functionally interpretable structure for the generic network. Like building LEGO blocks, we teach the generic network a new category by directly transplanting the module corresponding to the category from a pre-trained network with a few or even without sample annotations. Our method incrementally adds new categories to the generic network but does not affect representations of existing categories. In this way, our method breaks the typical bottleneck of learning a net for massive tasks and categories, i.e., the requirement of collecting samples for all tasks and categories at the same time before the learning begins. Thus, we use a new distillation algorithm, namely back-distillation, to overcome specific challenges of network transplanting. Our method without training samples even outperformed the baseline with 100 training samples.

cs.LG↗

arXiv2019

Unsupervised Learning of Neural Networks to Explain Neural Networks (extended abstract)

This paper presents an unsupervised method to learn a neural network, namely an explainer, to interpret a pre-trained convolutional neural network (CNN), i.e., the explainer uses interpretable visual concepts to explain features in middle conv-layers of a CNN. Given feature maps of a conv-layer of the CNN, the explainer performs like an auto-encoder, which decomposes the feature maps into object-part features. The object-part features are learned to reconstruct CNN features without much loss of information. We can consider the disentangled representations of object parts a paraphrase of CNN features, which help people understand the knowledge encoded by the CNN. More crucially, we learn the explainer via knowledge distillation without using any annotations of object parts or textures for supervision. In experiments, our method was widely used to interpret features of different benchmark CNNs, and explainers significantly boosted the feature interpretability without hurting the discrimination power of the CNNs.

cs.LG↗

arXiv2019

Explaining AlphaGo: Interpreting Contextual Effects in Neural Networks

In this paper, we propose to disentangle and interpret contextual effects that are encoded in a pre-trained deep neural network. We use our method to explain the gaming strategy of the alphaGo Zero model. Unlike previous studies that visualized image appearances corresponding to the network output or a neural activation only from a global perspective, our research aims to clarify how a certain input unit (dimension) collaborates with other units (dimensions) to constitute inference patterns of the neural network and thus contribute to the network output. The analysis of local contextual effects w.r.t. certain input units is of special values in real applications. Explaining the logic of the alphaGo Zero model is a typical application. In experiments, our method successfully disentangled the rationale of each move during the Go game.

cs.CV↗

arXiv2018

VeMo: Enabling Transparent Vehicular Mobility Modeling at Individual Levels with Full Penetration

Understanding and predicting real-time vehicle mobility patterns on highways are essential to address traffic congestion and respond to the emergency. However, almost all existing works (e.g., based on cellphones, onboard devices, or traffic cameras) suffer from high costs, low penetration rates, or only aggregate results. To address these drawbacks, we utilize Electric Toll Collection systems (ETC) as a large-scale sensor network and design a system called VeMo to transparently model and predict vehicle mobility at the individual level with a full penetration rate. Our novelty is how we address uncertainty issues (i.e., unknown routes and speeds) due to sparse implicit ETC data based on a key data-driven insight, i.e., individual driving behaviors are strongly correlated with crowds of drivers under certain spatiotemporal contexts and can be predicted by combining both personal habits and context information. More importantly, we evaluate VeMo with (i) a large-scale ETC system with tracking devices at 773 highway entrances and exits capturing more than 2 million vehicles every day; (ii) a fleet consisting of 114 thousand vehicles with GPS data as ground truth. We compared VeMo with state-of-the-art benchmark mobility models, and the experimental results show that VeMo outperforms them by average 10% in terms of accuracy.

cs.NI↗

arXiv2018

Network Transplanting

This paper focuses on a new task, i.e., transplanting a category-and-task-specific neural network to a generic, modular network without strong supervision. We design an functionally interpretable structure for the generic network. Like building LEGO blocks, we teach the generic network a new category by directly transplanting the module corresponding to the category from a pre-trained network with a few or even without sample annotations. Our method incrementally adds new categories to the generic network but does not affect representations of existing categories. In this way, our method breaks the typical bottleneck of learning a net for massive tasks and categories, i.e. the requirement of collecting samples for all tasks and categories at the same time before the learning begins. Thus, we use a new distillation algorithm, namely back-distillation, to overcome specific challenges of network transplanting. Our method without training samples even outperformed the baseline with 100 training samples.

cs.LG↗