A5064867350- Decision-Making and Behavioral Economics
- Theoretical and Computational Physics
- Traffic and Road Safety
- Constraint Satisfaction and Optimization
- Complex Network Analysis Techniques
- Crime Patterns and Interventions
- Neural Networks and Applications
- Advanced Memory and Neural Computing
- Gaze Tracking and Assistive Technology
- Topological and Geometric Data Analysis
- Advanced Database Systems and Queries
- Logic, programming, and type systems
- Quantum many-body systems
- Digital Image Processing Techniques
- EEG and Brain-Computer Interfaces
- VLSI and FPGA Design Techniques
- Model Reduction and Neural Networks
- Tensor decomposition and applications
- Speech and Audio Processing
- Quantum Computing Algorithms and Architecture
- Cellular Automata and Applications
- Formal Methods in Verification
- Semantic Web and Ontologies
- Speech Recognition and Synthesis
- Advanced NMR Techniques and Applications
University of California, San Diego
2018-2021
North China Electric Power University
2018
China Electric Power Research Institute
2018
We introduce Centaurus, a class of networks composed generalized state-space model (SSM) blocks, where the SSM operations can be treated as tensor contractions during training. The optimal order then systematically determined for every block to maximize training efficiency. This allows more flexibility in designing blocks beyond depthwise-separable configuration commonly implemented. new design choices will take inspiration from classical convolutional including group convolutions, full and...
Boolean satisfiability is a propositional logic problem of interest in multiple fields, e.g., physics, mathematics, and computer science. Beyond field research, instances the SAT problem, as it known, require efficient solution methods variety applications. It decision determining whether formula has satisfying assignment, believed to exponentially growing time for an algorithm solve worst-case instances. Yet, many classes formulae eludes even most successful algorithms, not only scenarios,...
Restricted Boltzmann machines (RBMs) are a powerful class of generative models, but their training requires computing gradient that, unlike supervised backpropagation on typical loss functions, is notoriously difficult even to approximate. Here, we show that properly combining standard updates with an off-gradient direction, constructed from samples the RBM ground state (mode), improves dramatically over traditional methods. This approach, which call mode training, promotes faster and...
Universal memcomputing machines (UMMs) [IEEE Trans. Neural Netw. Learn. Syst. 26, 2702 (2015)] represent a novel computational model in which memory (time non-locality) accomplishes both tasks of storing and processing information. UMMs have been shown to be Turing-complete, namely they can simulate any Turing machine. In this paper, using set theory cardinality arguments, we compare them with liquid-state (or "reservoir computing") quantum ("quantum computing"). We show that types machines,...
Many optimization problems can be cast into the maximum satisfiability (MAX-SAT) form, and many solvers have been developed for tackling such problems. To evaluate a MAX-SAT solver, it is convenient to generate hard instances with known solutions. Here, we propose method of generating weighted MAX-2-SAT inspired by frustrated-loop algorithm used quantum annealing community. We extend general bipartite couplings, associated problem being minimization restricted Boltzmann machine (RBM) energy...
With the rapid growth of wind power, speed forecasting becomes more and significant to ensure stable efficient operations power system. This paper proposes an improved hybrid methodology for short-term forecasting. After data preprocessing, MI algorithm is used select proper features, then Ensemble Empirical Mode Decomposition (EEMD) utilized decompose original series in order make chaotic stable. A novel model named ST-LSSVM proposed forecast decomposed sub-series, which combines Least...
Event-based data are commonly encountered in edge computing environments where efficiency and low latency critical. To interface with such leverage their rich temporal features, we propose a causal spatiotemporal convolutional network. This solution targets efficient implementation on edge-appropriate hardware limited resources three ways: 1) deliberately simple architecture set of operations (convolutions, ReLU activations) 2) can be configured to perform online inference efficiently via...
This survey reviews the AIS 2024 Event-Based Eye Tracking (EET) Challenge. The task of challenge focuses on processing eye movement recorded with event cameras and predicting pupil center eye. emphasizes efficient tracking to achieve good accuracy efficiency trade-off. During period, 38 participants registered for Kaggle competition, 8 teams submitted a factsheet. novel diverse methods from factsheets are reviewed analyzed in this advance future event-based research.
We introduce a class of models named PLEIADES (PoLynomial Expansion In Adaptive Distributed Event-based Systems), which contains temporal convolution kernels generated from orthogonal polynomial basis functions. focus on interfacing these networks with event-based data to perform online spatiotemporal classification and detection low latency. By virtue using structured data, we have the freedom vary sample rate along discretization step-size network without additional finetuning....
We present aTENNuate, a simple deep state-space autoencoder configured for efficient online raw speech enhancement in an end-to-end fashion. The network's performance is primarily evaluated on denoising, with additional assessments tasks such as super-resolution and de-quantization. benchmark aTENNuate the VoiceBank + DEMAND Microsoft DNS1 synthetic test sets. network outperforms previous real-time denoising models terms of PESQ score, parameter count, MACs, latency. Even waveform processing...
We develop a general model for finding the optimal penal strategy based on behavioral traits of offenders. focus how discount rate (level time discounting) affects criminal propensity individual level, and aggregation these effects influences activities population level. The are aggregated distribution among population. study this empirically through survey with 207 participants, we show that it follows zero-inflated exponential distribution. quantify effectiveness as its net utility...
We develop a general model for finding the optimal penal strategy based on behavioral traits of offenders. focus how discount rate (level time discounting) affects criminal propensity individual level, and aggregation these effects influences activities population level. The are aggregated distribution among population. study this empirically through survey with 207 participants, we show that it follows zero-inflated exponential distribution. quantify effectiveness as its net utility...
It is believed that the $\pm J$ Ising spin-glass does not order at finite temperatures in dimension $d=2$. However, using a graphical representation and contour argument, we prove rigorously existence of finite-temperature phase transition $d\geq 2$ with $T_c \geq 0.4$. In representation, low-temperature allows for coexistence multiple infinite clusters each rigidly aligned spin-overlap state. These correlate negatively other, are entropically stable without breaking any global symmetry....