87191 research outputs found
Sort by
Mixed-Precision Quantization: Make the Best Use of Bits Where They Matter Most
Get PDFMixed-precision quantization offers superior performance to fixed-precision quantization. It has been widely used in signal processing, communication systems, and machine learning. In mixed-precision quantization, bit allocation is essential. Hence, in this paper, we propose a new bit allocation framework for mixed-precision quantization from a search perspective. First, we formulate a general bit allocation problem for mixed-precision quantization. Then we introduce the penalized particle swarm optimization (PPSO) algorithm to address the integer consumption constraint. To improve efficiency and avoid iterations on infeasible solutions within the PPSO algorithm, a greedy criterion particle swarm optimization (GC-PSO) algorithm is proposed. The corresponding convergence analysis is derived based on dynamical system theory. Furthermore, we apply the above framework to some specific classic fields, i.e., finite impulse response (FIR) filters, receivers, and gradient descent. Numerical examples in each application underscore the superiority of the proposed framework to the existing algorithms
Individual Assembly of Two-Species Rydberg Molecules Using Optical Tweezers.
Get PDFWe present a new approach to investigating Rydberg molecules by demonstrating the formation and characterization of individual Rb^{*}Cs Rydberg molecules using optical tweezers. By employing single-atom detection of Rb and Cs, we observe molecule formation via correlated loss of both species and study the formation dynamics with single-particle resolution. We control the interatomic distances by manipulating the relative wave function of atom pairs using the tweezer intensity, optimizing the coupling to molecular states and exploring the effect of the tweezer on these states. Additionally, we demonstrate molecule association with atoms trapped in separate tweezers, paving the way for state-selective assembly of polyatomic molecules. The observed binding energies, molecular alignment, and bond lengths are in good agreement with theory. Our approach is broadly applicable to Rydberg tweezer platforms, expanding the range of available molecular systems and enabling the integration of Rydberg molecules into existing quantum science platforms
Words Without Intentions
Get PDFA commonly held position in the literature on the metaphysics of words holds that intentions are either jointly or independently necessary or sufficient for the tokening of a word. In this paper, using a modified version of an example case created by Pavese and Radulescu (2023), I propose a counterexample in which there is no intention to token a specific word, but a word is still successfully uttered
An Optimized Genetic Algorithm-Based Wavelet Image Fusion Technique for PCB Detection
Get PDFThis study proposes an optimized genetic algorithm-based wavelet image fusion technique for printed circuit board (PCB) detection, incorporating an improved Genetic Algorithm (GA) with the Elite Strategy and integrating it with discrete wavelet transform (DWT). The proposed method aims to enhance both the accuracy and efficiency of image fusion, which is crucial for defect detection in PCB inspection. A DWT is utilized to decompose images into multiple frequency components, where the low-frequency band preserves the structural integrity of the image, and the high-frequency band retains essential fine details such as edges and textures, which are critical for identifying defects. An improved genetic algorithm is applied to optimize the fusion process, incorporating the Elite Strategy to retain the best solutions in each evolutionary iteration. This strategy prevents the loss of optimal wavelet decomposition weights, and ensures steady convergence towards the global optimum. By maintaining superior solutions throughout the evolutionary process, the algorithm effectively enhances the fusion quality and computational efficiency. Experimental evaluations validate the effectiveness of the proposed approach, demonstrating superior performance over conventional fusion methods. The enhanced algorithm achieves significant improvements in key performance metrics, including relative standard deviation (RSD), peak signal-to-noise ratio (PSNR), image clarity, and processing efficiency. The team developed a prototype system and conducted simulations in a relatively realistic environment to validate the proposed method’s potential for high-precision PCB detection. The results demonstrate that the approach offers a robust solution for automated defect detection and quality assessment
The Influence of Water Molecules on the π* Shape Resonances of the Thymine Anion.
Get PDFLow-energy electrons have been shown to resonantly attach to DNA, inducing strand breakages and other damaging lesions. While computational studies have suggested that the nucleobase moieties can serve as the initial attachment site, there remains ambiguity over the exact character of the temporary anion resonances that form due to the unestablished role of the surrounding environment. Here, we investigate the influence of an aqueous environment on the low-lying anion shape resonances of the π* character of the thymine anion by applying frequency-resolved photoelectron spectroscopy to thymine-water cluster anions, T (H O) , with an increasing degree of hydration, . Our results indicate that spontaneous solvent rearrangement will stabilize the π * and π * states into bound electronic states, and we observe evidence for internal conversion to the anion ground state, further aiding long-term electron capture via these resonances
A random walk model for the evolution of the halo spin vector
Get PDFWe follow the spin vector evolutions of well resolved dark matter haloes (containing more than 300 particles) in merger tree main branches from the Millennium and Millennium-II N-body simulations, from to . We find that there seems to be a characteristic plane for the spin vector evolution along each main branch. In the direction perpendicular to it, spin vectors oscillate around the plane, while within the plane, spin vectors show a coherent direction change as well as a diffusion in direction (possibly corresponds to a Gaussian white noise). This plane may reflect the geometry of surrounding large-scale structures. We also construct a simple stochastic model in which halo spin vector evolution is assumed to be driven by accretion of halo mass and angular momentum. This model can reproduce major features of the results from N-body simulations
A Multiwavelength Technique for Estimating Galaxy Cluster Mass Accretion Rates
Get PDFThe mass accretion rate of galaxy clusters is a key factor in determining their structure, but a reliable observational tracer has yet to be established. We present a state-of-the-art machine learning model for constraining the mass accretion rate of galaxy clusters from only X-ray and thermal Sunyaev–Zeldovich observations. Using idealized mock observations of galaxy clusters from the MillenniumTNG simulation, we train a machine learning model to estimate the mass accretion rate. The model constrains 68% of the mass accretion rates of the clusters in our data set to within 33% of the true value without significant bias, a ∼58% reduction in the scatter over existing constraints. We demonstrate that the model uses information from both radial surface brightness density profiles and asymmetries
Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models
Get PDFSparse Autoencoders (SAEs) are a popular method for decomposing Large Language Model (LLM) activations into interpretable latents, however they have a substantial training cost and SAEs learned on different models are not directly comparable. Motivated by relative representation similarity measures, we introduce Inference-Time Decomposition of Activation models (ITDAs). ITDAs are constructed by greedily sampling activations into a dictionary based on an error threshold on their matching pursuit reconstruction. ITDAs can be trained in 1% of the time of SAEs, allowing us to cheaply train them on Llama-3.1 70B and 405B. ITDA dictionaries also enable cross-model comparisons, and outperform existing methods like CKA, SVCCA, and a relative representation method on a benchmark of representation similarity. Code available at https://github.com/pleask/itd
Data sources:
Access Repository Dashboard
Do you manage Open Research Online? Become a CORE Member to access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard! 👇