Recent studies on neural architecture search have shown that automatically designed networks perform as good expert-crafted architectures. While most existing works aim at finding architectures optimize the prediction accuracy, these may complexity and is therefore not suitable being deployed certain computing environment (e.g., with limited power budgets). We propose MONAS, a framework for Multi-Objective Neural Architectural Search employs reward functions considering both accuracy other...

In advanced IC manufacturing, as the gap increases between lithography optical wavelength and feature size, it becomes challenging to detect problematic layout patterns called hotspot. this paper, we propose a novel fuzzy matching model which extracts appropriate vectors of hotspot nonhotspot patterns. Our can dynamically tune regions around known hotspots. Based on develop fast algorithm for detection with high accuracy low probability false-alarm counts. addition, since higher dimensional...

Deep image classifiers often perform poorly when training data are heavily class-imbalanced. In this work, we propose a new regularization technique, Remix, that relaxes Mixup's formulation and enables the mixing factors of features labels to be disentangled. Specifically, two samples, while mixed in same fashion as Mixup, Remix assigns label favor minority class by providing disproportionately higher weight class. By doing so, classifier learns push decision boundaries towards majority...

This paper proposes an efficient algorithm for technology mapping targeting table look-up (TLU) blocks. It is capable of minimizing either the number TLUs used or depth produced circuit. Our approach consists two steps. First a network super nodes, created. Next Boolean function each node with appropriate don't care set decomposed into TLUs. To minimize circuit's depth, several rules are applied on critical portion mapped

Adversarial robustness has emerged as an important topic in deep learning carefully crafted attack samples can significantly disturb the performance of a model. Many recent methods have proposed to improve adversarial by utilizing training or model distillation, which adds additional procedures training. In this paper, we propose new paradigm called Guided Complement Entropy (GCE) that is capable achieving "adversarial defense for free," involves no process improving robustness. addition...

In advanced IC manufacturing, as the gap between lithography optical wavelength and feature size increases, it becomes challenging to detect problematic layout patterns called hotspot. this paper, we propose a novel fuzzy matching model which can dynamically tune appropriate regions around known hotspots. Based on model, develop fast algorithm for hotspot detection with very low chances of false-alarm. Our results are encouraging under 0.56 CPU-hrs/mm2 runtime.

Unreliable communication channels and limited computing resources at the edge end are two primary constraints of battery-powered movable devices, such as autonomous robots unmanned aerial vehicles (UAVs). The impact is especially severe for those performing deep neural network (DNN) computations. With increasing demand accuracy, trend in modern DNN designs use cascaded modularized layers. Implementing a increases computational workloads resource occupancy, leading to an increase battery...

In this study, the effect of plating current densities on self-annealing behaviors electroplated Cu films was found to be relevant polarization resistance electroplating systems. Porous with defects occurred at low density or resistance. contrast, dense small grains higher However, when more further supplied, aggregation and deposited became spongy dendritic. We suggest that both within porous underlying energy fine-grained deposits accelerated self-annealing. These two characteristics...

A compact tunable bandpass filter (BPF) with a broad stopband is presented. The proposed composed of two modified coupled lines and one varactor-loaded stub. In terms measurement, the BPF transmission zero within higher passband skirt has very wide tuning range, from 0.95 to 1.55 GHz (48%). Moreover, return loss > 10 dB, there controllable constant bandwidth 120 MHz. Furthermore, insertion 2.4 2.8 dB 20 dB.

Preconditioned Conjugate Gradient (PCG) method has been demonstrated to be effective in solving large-scale linear systems for sparse and symmetric positive definite matrices. One critical problem PCG is design a good preconditioner, which can significantly reduce the runtime while keeping memory usage efficient. Universal preconditioners are simple easy construct, but their effectiveness highly problem-dependent. On other hand, domain-specific that explore underlying physical meaning of...

Recent breakthroughs in Neural Architectural Search (NAS) have achieved state-of-the-art performance many tasks such as image classification and language understanding. However, most existing works only optimize for model accuracy largely ignore other important factors imposed by the underlying hardware devices, latency energy, when making inference. In this paper, we first introduce problem of NAS provide a survey on recent works. Then deep dive into two advancements extending...

Learning with a primary objective, such as softmax cross entropy for classification and sequence generation, has been the norm training deep neural networks years. Although being widely-adopted approach, using objective exploits mostly information from ground-truth class maximizing data likelihood, largely ignores complement (incorrect) classes. We argue that, in addition to also that leverages classes can be effective improving model performance. This motivates us study new paradigm...

Prediction of post-stroke functional outcomes is crucial for allocating medical resources. In this study, a total 577 patients were enrolled in the Post-Acute Care-Cerebrovascular Disease (PAC-CVD) program, and 77 predictors collected at admission. The outcome was whether patient could achieve Barthel Index (BI) score >60 upon discharge. Eight machine-learning (ML) methods applied, their results integrated by stacking method. area under curve (AUC) eight ML models ranged from 0.83 to...

In synchronous circuit designs, clock skew is difficult to minimize because a single physical layout of tree must satisfy multiple constraints in complicated power mode environment where certain modules may operate with different voltages. this paper, we use adjustable delay buffers (ADB) whose delays can be tuned or adjusted under modes. Assuming that the positions <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex...

In 3-D integrated circuits, through silicon via (TSV) is a critical enabling technique to provide vertical connections. However, it may suffer from many reliability issues such as undercut, misalignment, or random open defects. Various fault-tolerance mechanisms have been proposed in literature improve yield, at the cost of significant area overhead. this paper, we focus on structure that uses one spare TSV for group original TSVs, and study optimal assignment TSVs under yield timing...

We report the formation of arrays short In2O3 nanorods inside nanoscale channels mesoporous silica SBA-15. In(NO3)3 dissolved in methanol was incorporated into SBA-15 powder, followed by heat treatment under a nitrogen flow at 700 °C for 4 h to generate densely aligned nanorods. The have been characterized XRD patterns, TEM images, adsorption−desorption isotherm measurements, and optical spectroscopy. They possess cubic crystal structure. free-standing obtained after framework removal with...

Three-dimensional integrated circuit (3D IC) has become an emerging technology in view of its advantages packing density and flexibility heterogeneous integration. The multi-core processor (MCP), which is able to deliver equivalent performance with less power consumption, a candidate for 3D implementation. However, when maximizing the throughput MCP, due inherent heat removal limitation, thermal issues must be taken into consideration. Furthermore, since temperature core strongly depends on...

A tunable bandpass filter with a wide tuning range is presented in this letter. The proposed composed of two modified parallel-coupled lines and one varactor-loaded stub. Particularly, the line realized by varactor connected to end coupled asymmetric feeding an open stub added line's other end. Moreover, introduced control transmission zero around lower/higher passband skirt. measured results show that broad 41.7%, from 0.95 1.45 GHz, was achieved. Within range, insertion loss 2.4 2.9 dB...

Multiple string matching has a wide range of applications such as network intrusion detection systems, spam filters, information retrieval and bioinformatics. To accelerate multiple matching, many hardware approaches are proposed to matching. Among the approaches, memory architectures have been widely adopted because their flexibility scalability. A conventional architecture compiles patterns into state machine performs by traversing corresponding transition table. Due ever-increasing number...

Artificial intelligence models implemented in power-efficient Internet-of-Things (IoT) devices have accuracy degradation due to limited power consumption. To mitigate the loss on IoT devices, an edge-server joint inference system is introduced. On system, allocate more workloads server end can loss, but data transmission contributes consumption of edge device. Thus, this article, we present a novel two-stage method or maximize under constraint. In first stage, clusterwise threshold-based for...

Through elucidating the effects of current density, cupric ion concentration, bath temperature, and air agitation on plating uniformity filling capability copper electroplating, deposition in an acid electrolyte will be illustrated to scale down sub-0.13 μm features with uniform plating, which is required by chemical mechanical polishing damascene techniques. In order achieve defect-free vias trenches, must composed proper amounts ions, sulfuric acid, chloride wetting agent, promoter. The...

The authors address the problem of how to realize a given combinational circuit described by means Boolean equations using minimum number blocks target TLU table lookup architecture. Their decomposition scheme works directly on reduced ordered binary decision diagram (ROBDD) subject function, two techniques. first, referred as cutting, is an efficient implementation Roth-Karp decomposition. second technique substitution. idea replace subgraphs ROBDD new variables. substitution process...

In the VLSI design process, a implementation often needs to be corrected because of new specifications or constraint violations. This correction process is referred as engineering change (EC). Usually, an EC problem resolved by using spare cells, which have been inserted into unused spaces chip. this paper, we propose iterative method generate feasible mapping solutions for considering cells whose inputs may tied Vdd Gnd, called constant insertion. Applying insertion can increase cell's...

Gap-filling dynamics of several different species additives for copper electrodeposition was investigated by monitoring the cross section a partially filled profile on scanning electron microscopy photo. The filling ration Δy/Δx between “bottom-up” with “sidewall shift” found to be proportional power additives. adsorption-diffusion model combined cathode polarization and cyclic voltammetric stripping measurements employed explain attribution in superfilling phenomena. achieved under behavior...