- Advanced Memory and Neural Computing
- Neural Networks Stability and Synchronization
- Natural Language Processing Techniques
- Topic Modeling
- stochastic dynamics and bifurcation
- China's Ethnic Minorities and Relations
- Advanced Neural Network Applications
- Asian Geopolitics and Ethnography
- Advanced Computational Techniques and Applications
- Handwritten Text Recognition Techniques
- Neural dynamics and brain function
- Speech Recognition and Synthesis
- Multimodal Machine Learning Applications
- CCD and CMOS Imaging Sensors
- Remote-Sensing Image Classification
- Neuroscience and Neural Engineering
- Distributed Control Multi-Agent Systems
- Music and Audio Processing
- Neural Networks and Applications
- Image Processing and 3D Reconstruction
- Web Data Mining and Analysis
- Linguistics and Cultural Studies
- Text and Document Classification Technologies
- Domain Adaptation and Few-Shot Learning
- Chaos control and synchronization
Tibet University
2017-2025
Tibetan Traditional Medical College
2023-2024
Tibet Autonomous Region People's Hospital
2023
University of the Humanities
2008
University of Oslo
2008
Activation functions facilitate deep neural networks by introducing non-linearity to the learning process. The feature gives network ability learn complex patterns. Recently, most widely used activation function is Rectified Linear Unit (ReLU). Though, other various existing including hand-designed alternatives ReLU have been proposed. However, none has succeeded in replacing due their inconsistencies. In this work, called ReLUMemristor-like Function (RMAF) proposed leverage benefits of...
This article investigates the problem of robust exponential stability fuzzy switched memristive inertial neural networks (FSMINNs) with time-varying delays on mode-dependent destabilizing impulsive control protocol. The model presented here is treated as a system rather than employing theory differential inclusion and set-value map. To optimize exponentially stable process reduce cost time, hybrid adaptive feedback controllers are simultaneously applied to stabilize FSMINNs. In new model,...
This paper proposes a mem-computing model of memristive network-based genetic algorithm (MNGA) by building up the relationship between network (MN) and (GA), new edge detection where image pixels are defined as individuals population. First, computing MNGA is designed to perform mem-computing, which brings possibility hardware implementation GA. Secondly, MNGA-based integrating filter GA operator deployed MN proposed. Finally, simulation results demonstrate that figure merit (FoM) our better...
This article investigates the heterogeneous impulsive synchronization for T-S fuzzy probabilistic coupled delayed neural networks (CDNNs) with mode-dependent parameters and piecewise membership functions. To begin with, a novel CDNNs model adjustable coupling strength delays is designed to ensure accuracy of model. Meanwhile, generalized isolated node four types mismatched parameters, named network, first considered extend problem. Then, rules are introduced design model, which implies that...
This paper highlights the memristor bridge-based lowpass filter (LPF) and improved image processing algorithms along with a novel adaptive Gaussian for denoising new pyramid scale invariant feature transform (SIFT). First, kind of LPF based on bridge is designed, whose cut-off frequency other traits are demonstrated to change different time memristance. In light changeable parameter LPF, an SIFT algorithm presented. Finally, experiment results show that peak signalto-noise ratio (PSNR) our...
This article addresses the problem of dynamic pinning synchronization fuzzy-dependent-switched (Fds) coupled memristive neural networks (CMNNs) with mismatched dimensions on time scales. To begin with, probabilistic coupling delays, scales, dimensions, and function projective rules are considered to design novel CMNNs improve reliability generalization ability model. Then Fds control (DPC) method adopted CMNNs, which can effectively promote information exchange between switching signals...
Brain tumor recently is considered among the deadliest cancers according to research statistics and have several categories, based on different characteristics of tumor. Early detection types help devise treatment plans achieve high survival rate. Human inspection noted be cost effective, error prone time-consuming, which led interest in Convolutional Neural Networks (CNNs) automatize problem. However, CNNs fail consider precise location features as beneficial, harmful, because its...
This article is concerned with the exponential synchronization of coupled memristive neural networks (CMNNs) multiple mismatched parameters and topology-based probability impulsive mechanism (TPIM) on time scales. To begin with, a novel model designed by taking into account three types parameters, including: 1) dimensions; 2) connection weights; 3) time-varying delays. Then, method auxiliary-state variables adopted to deal model, which implies that presented can not only use any isolated...
This article investigates the problem of relaxed exponential stabilization for coupled memristive neural networks (CMNNs) with connection fault and multiple delays via an optimized elastic event-triggered mechanism (OEEM). The two or some nodes can result in other cause iterative faults CMNNs. Therefore, method backup resources is considered to improve fault-tolerant capability survivability In order robustness enhance ability process noise signals, time-varying bounded threshold matrices,...
Availability of massive amounts data is a key contributing factor that influences the performance deep learning models. Convolutional Neural Networks for instance, require large in different variations to enable them generalize well viewpoints. However, health and other application domains, generation processing tasks are time-consuming requires annotation by experts. Capsule Network (CapsNet) have been proposed curtail limitations (CNNs). Due problem crowding, capsule perform badly on...
Abstract The issue of how to enhance cooperation has been a hot topic research in evolutionary games for long time. A mechanism is proposed facilitate the behavior groups on networks three game models, including prisoner's dilemma, snowdrift game, and stag hunt game. core lies in: 1) Each player length memory uses information elite span update its strategy. 2) chance with certain neighbor more than once each round. 3) accumulative payoff consists two parts, one from playing elites another...
Addressed in this paper is the stability issue of discrete-time monotone nonlinear systems. With assumption considered vector field being a one-to-one mapping, we established equivalence between locally asymptotic and existence max-separable Lyapunov function on compact set. This result an extension that linear Furthermore, it counterpart continuous-time
Active filters based on memristor and memcapacitor (AFMM) including low-pass, high-pass, bandpass, band-stop filter, are studied in this paper. By replacing resistor capacitor with memcapacitor, the novel AFMMs designed. Then transfer function, cut-off frequency, Bode plot demonstrated. Experiment results show that amplitude frequency phase-frequency responses change over time.
Abstract The scarcity of resources, lack labeled texts, and insufficient corpora pose significant challenges to many specialized classification problems. It increases the difficulty small language domain-specific In this paper, we propose a new approach address problem: by assigning multi-dimensional additional weights words using external knowledge, thereby enhancing text features in low-resource domains. This is achieved introducing concepts 'prior domains' 'Adjusted Term Frequency Vectors...
Summary This paper addresses the strong consensus problem of convex second‐order discrete‐time multi‐agent systems (MASs) with time‐varying topologies. The MAS model is derived from Langevin equation and therefore has a certain physical significance. here means that all first‐ states converge to an identical value. Some fully distributed control protocols are designed weights randomly chosen arbitrary finite set. These applicable several cases where changing topologies may be directed or...