arXiv subjects

Hyunwook Park

Publications and source records attributed to Hyunwook Park.

5 recordsLinked to original sources

arXiv2022

Transformer Network-based Reinforcement Learning Method for Power Distribution Network (PDN) Optimization of High Bandwidth Memory (HBM)

In this article, for the first time, we propose a transformer network-based reinforcement learning (RL) method for power distribution network (PDN) optimization of high bandwidth memory (HBM). The proposed method can provide an optimal decoupling capacitor (decap) design to maximize the reduction of PDN self- and transfer impedance seen at multiple ports. An attention-based transformer network is implemented to directly parameterize decap optimization policy. The optimality performance is significantly improved since the attention mechanism has powerful expression to explore massive combinatorial space for decap assignments. Moreover, it can capture sequential relationships between the decap assignments. The computing time for optimization is dramatically reduced due to the reusable network on positions of probing ports and decap assignment candidates. This is because the transformer network has a context embedding process to capture meta-features including probing ports positions. In addition, the network is trained with randomly generated data sets. Therefore, without additional training, the trained network can solve new decap optimization problems. The computing time for training and data cost are critically decreased due to the scalability of the network. Thanks to its shared weight property, the network can adapt to a larger scale of problems without additional training. For verification, we compare the results with conventional genetic algorithm (GA), random search (RS), and all the previous RL-based methods. As a result, the proposed method outperforms in all the following aspects: optimality performance, computing time, and data efficiency.

cs.LG↗

arXiv2022

Confidence Guided Depth Completion Network

The paper proposes an image-guided depth completion method to estimate accurate dense depth maps with fast computation time. The proposed network has two-stage structure. The first stage predicts a first depth map. Then, the second stage further refines the first depth map using the confidence maps. The second stage consists of two layers, each of which focuses on different regions and generates a refined depth map and a confidence map. The final depth map is obtained by combining two depth maps from the second stage using the corresponding confidence maps. Compared with the top-ranked models on the KITTI depth completion online leaderboard, the proposed model shows much faster computation time and competitive performance.

cs.CV↗

arXiv2021

Unsupervised Anomaly Detection in MR Images using Multi-Contrast Information

Anomaly detection in medical imaging is to distinguish the relevant biomarkers of diseases from those of normal tissues. Deep supervised learning methods have shown potentials in various detection tasks, but its performances would be limited in medical imaging fields where collecting annotated anomaly data is limited and labor-intensive. Therefore, unsupervised anomaly detection can be an effective tool for clinical practices, which uses only unlabeled normal images as training data. In this paper, we developed an unsupervised learning framework for pixel-wise anomaly detection in multi-contrast magnetic resonance imaging (MRI). The framework has two steps of feature generation and density estimation with Gaussian mixture model (GMM). A feature is derived through the learning of contrast-to-contrast translation that effectively captures the normal tissue characteristics in multi-contrast MRI. The feature is collaboratively used with another feature that is the low-dimensional representation of multi-contrast images. In density estimation using GMM, a simple but efficient way is introduced to handle the singularity problem which interrupts the joint learning process. The proposed method outperforms previous anomaly detection approaches. Quantitative and qualitative analyses demonstrate the effectiveness of the proposed method in anomaly detection for multi-contrast MRI.

eess.IV↗

arXiv2017

High-energy mid-infrared sub-cycle pulse synthesis from a parametric amplifier

High-energy, carrier-envelope phase (CEP)-stable, sub-cycle, mid-infrared (mid-IR) pulses can provide unique opportunities of exploring phase-sensitive strong-field light-matter interactions in atoms, molecules, and solids. In the mid-IR wavelength, the ponderomotive energy of laser pulses is dramatically increased (versus the visible/near-infrared) and, therefore, the Keldysh parameter is much smaller than unity even at relatively modest laser intensities. This enables to study the sub-cycle electron dynamics in solids via high-harmonic generation (HHG) without damage. One can also control the electron emissions from nano-devices in the sub-cycle time scale. These efforts are opening a great opportunity towards petahertz electronics. Here, we present a high-energy, sub-cycle pulse synthesizer based on a mid-IR optical parametric amplifier (OPA), pumped by CEP-stable, 2.1 um femtosecond pulses, and its application to HHG in solids. The signal and idler combined spectrum spans from 2.5 to 9.0 um, which covers the whole midwave-infrared (MWIR) region. We coherently synthesize the passively CEP-stable few-cycle signal and idler pulses to generate 33 uJ, 0.88-cycle (12.4 fs), multi-GW pulses centered at ~4.2 um, which is further energy scalable. The in-line synthesis of the CEP-stable sub-cycle pulse is realized through the type-I collinear OPA with minimal temporal walk-off. The MWIR sub-cycle pulse is used for driving HHG in thin silicon samples, producing harmonics up to ~19th order with a continuous spectral coverage due to the isolated emission by the sub-cycle driver. Our demonstration offers an energy scalable and technically simple platform of laser sources generating CEP-stable sub-cycle pulses in the whole MWIR region for investigating isolated phase-sensitive strong-field interactions in solids and gases.

physics.optics↗

arXiv2012

Dipole-Dipole coupled double Rydberg molecules

We show that the dipole-dipole interaction between two Rydberg atoms can give rise to long range molecules. The binding potential arises from two states that converge to different separated atom asymptotes. These states interact weakly at large distances, but start to repel each other strongly as the van der Waals interaction turns into a resonant dipole-dipole interaction with decreasing separation between the atoms. This mechanism leads to the formation of an attractive well for one of the potentials. If the two separated atom asymptotes come from the small Stark splitting of an atomic Rydberg level, which lifts the Zeeman degeneracy, the depth of the well and the location of its minimum are controlled by the external electric field. We discuss two different geometries that result in a localized and a donut shaped potential, respectively.

physics.atom-ph↗