Browsing by Author "Ho, Yung-Han"

Now showing 1 - 4 of 4

A Comprehensive Study of 3-stage High Resistance State Retention Behavior for TMO ReRAMs from Single Cells to a Large Array
(2017-01-01)
Lin, Yu-Hsuan
;
Ho, Yung-Han
;
Lee, Ming-Hsiu
;
Wang, Chao-Hung
;
Lin, Yu-Yu
For the first time the retention of high resistance state (HRS) in resistive random access memory (ReRAM) is found to compose of three stages - extending tail-bits, distribution shift, and then distribution broadening. This work provides a comprehensive study on the HRS's retention behavior in WOx-based ReRAMs from single device characteristics to group distribution. Different from conventional activation energy (Ea) analysis, the mean and variance of the array distribution are presented to overcome the non-uniform Ea issue. Since the extracted Ea fits well with Vo(2+) (oxygen vacancy) migration characteristics, the three retention stages are suggested to be the competing results from the migration, recombination, and generation of the Vo(2+) and O2- in the gap region. A three-dimensional retention model with kinetic Monte Carlo simulator and trap-assisted tunneling conduction is proposed to discuss the dominating mechanism in different time and temperature scales. The mechanism for random telegraph noise is also included to illustrate the fluctuating nature of the HRS cells.
DEEP REINFORCEMENT LEARNING FOR VIDEO PREDICTION
(2019-01-01)
Ho, Yung-Han
;
Cho, Chuan-Yuan
;
Peng, Wen-Hsiao
;
資訊工程學系
;
Department of Computer Science
This paper introduces a hybrid video prediction scheme that combines the classic parametric overlapped block motion compensation (POBMC) technique with neural networks. Most learning-based video prediction methods rely on a black-box-like model for either direct generation of future video frames or estimation of a dense motion field. The model complexity often increases drastically with frame resolution. Departing from pure black-box approaches, this paper leverages the theoretically-grounded POBMC in a reinforcement learning framework to estimate a sparse motion field for future frame warping. Two neural networks are trained to identify critical points in the motion field for motion estimation. We train our model on 10k unlabeled frames in KITTI dataset and achieve the state-of-the-art SSIM score of 0.923 on CaltechPed and an average SSIM scroe of 0.856 on Common Intermediate Format (CIF) standard sequences.
Excellent High Temperature Retention of TiOXNV ReRAM by Interfacial Layer Engineering
(2018-01-01)
Lin, Yu-Hsuan
;
Lee, Dai-Ying
;
Wang, Chao-Hung
;
Lee, Ming-Hsiu
;
Ho, Yung-Han
Since the resistance switching of the transition metal oxide (TMO) resistive random access memory (ReRAM) is based on thc interaction between the oxygen ions and vacancies, the unintentional oxygen/vacancy reaction should be avoided during data retention. This work demonstrates significant improvements on the retention performance by inserting a Si layer in the TiO Ny ReRAM to block the diffusion of oxygen ions through the Ti/TiO,N interface. The mechanism and factors that influenced thc IIRS and LRS retention arc also studied. The retention performance of HRS is correlated with its RESET level while the LRS retention depends on the programming current. The proposed Ti/Si/TiO,N, ReRAMs can switch for more than 103 cycles from array testing results.
SME-Net: Sparse Motion Estimation for Parametric Video Prediction through Reinforcement Learning
(2019-01-01)
Ho, Yung-Han
;
Cho, Chuan-Yuan
;
Peng, Wen-Hsiao
;
Jin, Guo-Lun
;
資訊工程學系
This paper leverages a classic prediction technique, known as parametric overlapped block motion compensation (POBMC), in a reinforcement learning framework for video prediction. Learning-based prediction methods with explicit motion models often suffer from having to estimate large numbers of motion parameters with artificial regularization. Inspired by the success of sparse motion-based prediction for video compression, we propose a parametric video prediction on a sparse motion field composed of few critical pixels and their motion vectors. The prediction is achieved by gradually refining the estimate of a future frame in iterative, discrete steps. Along the way, the identification of critical pixels and their motion estimation are addressed by two neural networks trained under a reinforcement learning setting. Our model achieves the state-of-the-art performance on CaltchPed, UCF101 and CIF datasets in one-step and multi-step prediction tests. It shows good generalization results and is able to learn well on small training data.