CMA-ES
PulseAugur coverage of CMA-ES — every cluster mentioning CMA-ES across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
Diffusion models enable efficient inverse design of metasurfaces for smart electromagnetic environments
Researchers have developed a diffusion-based framework for the inverse design of dielectric resonator metasurfaces, aiming to create smart electromagnetic environments for future wireless systems. This method, trained o…
-
LLM-guided framework enhances neural architecture search proxies
Researchers have developed Bi-EZP, a novel bilevel framework designed to improve the discovery of ensemble zero-cost proxies for neural architecture search (NAS). This framework separates the discrete structural optimiz…
-
New Integer Natural Evolution Strategy Developed for Optimization Problems
Researchers have developed a new optimization strategy called the Integer Natural Evolution Strategy (INES) that is natively designed for integer optimization problems. Unlike existing methods that adapt continuous Gaus…
-
JANUS module enhances black-box optimization with Jacobian alignment
Researchers have developed JANUS, a novel plug-and-play module designed to enhance population-based optimizers like CMA-ES. JANUS extracts local geometric signals from recent evaluation traces without replacing the host…
-
Social media simulation model calibrated for content moderation research
Researchers have developed a calibrated extension of the SimSoM agent-based model to simulate social media dynamics, specifically focusing on content moderation. This enhanced model, grounded in real-world data from the…
-
New hybrid framework enhances LLM optimization by decoupling structure and parameters
Researchers have developed a novel hybrid nested search framework designed to improve the efficiency of large language models (LLMs) in optimization tasks. This approach decouples the structural and parameter updates, a…
-
UAV photogrammetry method improves 3D reconstruction accuracy
Researchers have developed a new iterative hybrid discrete-continuous viewpoint planning method specifically for unmanned aerial vehicle (UAV) photogrammetry. This method aims to improve reconstruction accuracy and comp…
-
New PSO variants enhance optimization with adaptive gradients and hypergraph topologies · 4 sources tracked
Researchers have developed new approaches to enhance Particle Swarm Optimization (PSO), a metaheuristic algorithm. One method, Adaptive Hybrid PSO (AHPSO), intelligently modulates gradient influence based on swarm diver…
-
New QML method boosts trainable-frequency circuit performance
Researchers have introduced a new initialization technique called ternary grid initialization for trainable-frequency (TF) circuits in quantum machine learning (QML). This method addresses a gradient suppression issue t…
-
New benchmark BBOPlace-Bench advances AI for chip placement
Researchers have introduced BBOPlace-Bench, a novel benchmark designed to evaluate and advance black-box optimization (BBO) algorithms specifically for chip placement tasks. This benchmark addresses a gap in existing to…
-
New Hybrid Method Enhances Constrained Optimization via Evolutionary Algorithms
Researchers have developed a new Hybrid Augmented Lagrangian (HyAL) method that combines the strengths of Augmented Lagrangian frameworks with evolutionary algorithms for constrained optimization problems. This novel ap…
-
NeuronSoup architecture evolves temporal graphs without backpropagation
Researchers have developed NeuronSoup, a novel neural computation architecture that deviates from traditional layer-by-layer processing. Instead, it utilizes asynchronous, delay-mediated signal propagation through a sha…
-
New method enhances evolutionary algorithms for noisy optimization problems
Researchers have developed a new confidence-based ranking method to improve the efficiency of evolutionary algorithms in solving noisy black-box optimization problems. This method employs an adaptive sampling strategy t…
-
New NOTES method enhances inverse design for physical systems
Researchers have developed a new method called Neural Operator-enabled Topology-informed Evolutionary Strategy (NOTES) to improve the inverse design of physical systems governed by partial differential equations. This a…
-
New method uses evolutionary optimization for real-time surgical instrument tracking
Researchers have developed a new method for tracking surgical instruments in real-time during robot-assisted minimally invasive surgery. This approach utilizes CMA-ES, an evolutionary optimization strategy, integrated i…
-
CMA-ES primer explains gradient-free model optimization
This article explains the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), a method for optimizing models when gradients are unavailable. CMA-ES works by sampling potential solutions from a Gaussian distributio…
-
New Bézier Walk Evolution framework enhances optimization with adaptive geometry
Researchers have introduced Bézier Walk Evolution (BWE), a novel optimization framework that uses geometry-driven adaptive trajectory construction. This method integrates Bézier curves with a random walk mechanism to ba…
-
\chisao{} optimizer leverages GPUs for multimodal black-box function optimization
Researchers have developed \chisao{}, a novel GPU-native parallel optimizer designed to efficiently find all modes of multimodal black-box functions. Unlike sequential CPU-based methods such as basin-hopping or CMA-ES, …
-
Evolutionary algorithms discover novel reward schedules for reinforcement learning
Researchers have developed an evolutionary framework to discover developmental reward schedules in deep reinforcement learning, aiming to explore how motivational priorities can shift during training. This approach comb…
-
New theory analyzes discrete IGO convergence in continuous spaces
Researchers have developed a new theoretical framework for analyzing the convergence of Information-Geometric Optimization (IGO) in discrete, continuous spaces. This work focuses on IGO updates within the multivariate G…