Menu
Skills in this repository
SkillsMP has collected 4,114 skills from hiyenwong/ai_collection. Open a skill to review its source and details.
hiyenwong/ai_collectionShowing 40 of 4,114 collected skills.
Geometry-aware longitudinal flow matching framework for unpaired biological snapshot data. FLUX learns data-dependent metrics, constructs manifold-aware conditional paths, and uses mixture-of-experts velocity fields for joint transport modeling and…
Scalable fluxonium quantum processor architecture using tunable-coupler unit cells. Achieves 99.9% CZ gate fidelity and demonstrates 22-qubit GHZ state generation. Alternative to transmon-based superconducting quantum computers. Keywords: fluxonium,…
Novel approach to dictionary learning on fMRI data that explicitly accounts for individual brain geometry variability using optimal transport (Fused Gromov-Wasserstein distance) with amortized optimization for computational efficiency.
fMRI2GES: Dual brain decoding alignment framework for co-speech gesture reconstruction from fMRI signals. Maps brain responses to external stimuli and decodes gestural behavior through dual-alignment brain decoding. Activation: fMRI gesture reconstruction,…
Fused On-Register Gradient Elimination for memory-efficient LLM training. Folds optimizer step into backward pass, applies tile-by-tile in registers, eliminating materialized gradients.
Control adaptation with formal guarantees for ODD-resilient autonomous systems. Combines barrier functions, adaptive control, and reachability analysis to provide safety certificates during online controller parameter adjustment. Use when designing adaptive…
Foundation models for discovering robust biomarkers of neurological disorders from dynamic functional connectivity. Use when: building neurological biomarker discovery pipelines, applying foundation models to fMRI/EEG data, analyzing dynamic functional…
Fourier-based Linear Combination of Unitaries (LCU) methodology for efficient quantum circuit decomposition in optimization algorithms. Covers ancilla-free LCU constructions, Fourier decomposition of diagonal/non-diagonal unitaries, formal connection to…
Functional-Network Prior-Guided Mixture-of-Experts (MoE) framework for interpretable brain decoding from fMRI. Uses brain network topology as expert priors with adaptive routing for visual semantic reconstruction.
Design FPGA-based neural network decoders for real-time quantum error correction (QEC) in surface code architectures. Covers hardware-software co-design, deterministic low-latency decoding under 1 microsecond, NN decoder implementation on FPGA, closed-loop…
FPGA-based real-time quantum error correction decoding architecture. Combines hardware-integrated NN decoders on FPGA with superconducting quantum processors for low-latency closed-loop QEC. Use when: (1) Designing real-time QEC control systems, (2)…
Federated Quantum Neural Network methodology for privacy-preserving medical image diagnosis. Combines federated learning (FL) with quantum neural networks (QNN) for early disease detection without sharing patient data. Addresses medical data privacy…
Frequency-matching methodology for Spiking Neural Networks in mmWave sensing. LIF dynamics provide inherent low-pass filtering that suppresses high-frequency noise in mmWave signals. Derives principled criterion for membrane decay factor by matching LIF…
Statistical benchmarking methodology for EEG motor-imagery BCI decoders using Friedman-Nemenyi tests. Proves no single decoding pipeline dominates across subjects — personalized model selection adds ~7% accuracy over best fixed choice. Use when evaluating BCI…
Senior Front-End Developer guidance for ReactJS, NextJS, JavaScript, TypeScript, HTML, CSS and modern UI/UX frameworks. Use when building front-end applications, writing React/Next.js code, or implementing UI components. Triggers on: frontend, react, nextjs,…
Frontier decoder for quantum LDPC codes — pruned dynamic-programming approach with narrow frontier syndrome decoding. Uses prefix merging with same residual syndrome and logical label, approximates logical-coset posteriors by retaining scored frontier.…
Encoding circuit synthesis methodology for fault-tolerant quantum computation. Constructs optimized circuits that map arbitrary logical states into error-correcting codes, minimizing two-qubit gate count and circuit depth. Use when: (1) designing…
Full extractor construction for logical processing in Hypergraph Product (HGP) QLDPC codes. Enables Pauli-based computation without compilation overhead. Extractors 50-80% of base code size, max qubit degree 10, fault-tolerant. arXiv:2606.03507.
Full-Stack FP4 pretraining framework — first complete NVFP4 LLM pretraining resolving stability bottlenecks in linear projections (LoRA-SVD), optimizers (AdamW second-moment transform, Muon Newton-Schulz), and attention (mixed-precision with forward-backward…
Functional Connectivity Graph Neural Networks methodology combining structural and functional connectivity with persistent graph homology for brain-inspired graph classification. Activation triggers: functional connectivity, graph neural network, persistent…
功能性连接组指纹分析方法论。扩展 differential identifiability 框架， 检测个体指纹梯度和双胞胎指纹梯度。 触发词：脑指纹、连接组指纹、个体差异、可识别性、fingerprint、 differential identifiability, connectome fingerprint。
Source text: Mixed languages
Functional Ensembles as Units of Computation in Deep Spiking Networks. 1FC (first-order functionally-connected) ensembles framework for analyzing information encoding in SNNs through rare coordinated firing events.
Source text: Chinese
Functional ensembles as units of computation in deep spiking networks. First-order functionally-connected (1FC) groups based on pairwise correlations, aggregate cofiring predicts downstream responses, ReLU-like input-output relationship with ensemble-size…
Functional Proximity Law in Multilayer Networks: Hub importance scores persist more strongly between functionally similar layers. Validated across 17 pre-registered experiments including neuroscience (r=0.777 in C. elegans connectome). Activation: multilayer…
Source text: Chinese
Functional Whole-Brain Models (fWBMs) — unified framework integrating structural/dynamical realism with functional competence
Source text: Chinese
Funessian过程：一种连续时间正可分非马尔可夫过程，具有初始状态记忆。平稳态下关联函数指数衰减（通常被视为马尔可夫特征），但记忆贯穿演化。互信息作为非马尔可夫性度量。应用于随机游走，展示记忆效应打破遍历性并改变扩散系数。
Source text: Chinese
Game-theoretic energetic framework for excitatory-inhibitory neural circuits with asymmetric connectivity and stability analysis.
Game-theoretic frameworks for modeling, learning, and control in socio-technical systems. Covers cooperative/noncooperative paradigms, feedback learning, incentive mechanism design, and multi-agent resilience. Tutorial by Tamer Başar, Tomohisa Hayakawa,…
Quantum-inspired sequence learning using Gated QKAN-FWP (Quantum Fast Weight Programmers with variational quantum Kolmogorov-Arnold Networks). Use this skill for designing quantum-inspired sequence models, temporal encoding for quantum ML, fast weight…
Gaussian Group Relative Policy Optimization (G²RPO) for multimodal RL training. Replaces linear scaling with distributional matching to ensure gradient equity across diverse tasks. Use when training multimodal models, balancing perception vs reasoning, or…
Ge²mS-T: Multi-Dimensional Grouping for Ultra-High Energy Efficiency in Spiking Transformers. Temporal, spatial, and channel grouping for efficient S-ViT training and inference. Triggers: spiking transformer, S-ViT, energy efficiency, multi-dimensional…
Strategy Gene methodology for experience-driven test-time control in LLM agents. Compact control-oriented experience representation (~230 tokens) outperforms documentation-heavy Skill (~2500 tokens) by +3.0pp. Core principle: encode experience as control…
脑动力学生成模型综述框架。整合计算神经科学、非线性动力学、数据驱动方法的生成模型方法论，涵盖不同组织尺度和抽象层次。适用于脑动力学建模、神经数据分析、科学机器学习。触发词：脑动力学、生成模型、神经动力学、动态系统模型、brain dynamics、generative model、neural dynamics、computational neuroscience、Dynamical systems。
Source text: Chinese
Genetic and environmental architecture of human functional connectome using extended twin modeling. Separates measurement error from non-shared environment to estimate true connectivity heritability. Keywords: functional connectome, twin modeling,…
Geometric Basis Functions (GBF) methodology for noninvasive whole human brain dynamics mapping using participant-specific cortical eigenmodes. Reconstructs whole-brain spatiotemporal dynamics from EEG/MEG with anatomically-constrained source imaging.…
Geometric Basis Functions (GBF) framework for noninvasive whole-brain spatiotemporal dynamics reconstruction. Uses participant-specific eigenmodes from cortical surface for EEG/MEG source imaging. Trigger words: geometric basis functions, GBF, brain dynamics,…
Geometric Basis Functions (GBF) framework for noninvasive whole human brain dynamics mapping. Uses participant-specific eigenmodes from cortical surface to resolve inverse problem in EEG/MEG source imaging. Activation: brain dynamics, geometric basis…
Geometric Basis Functions (GBF) framework for noninvasive whole human brain dynamics mapping using participant-specific eigenmodes derived from cortical geometry. Use when working with EEG/MEG source imaging, brain dynamics reconstruction, neuroimaging…
Geometry-aware framework for noninvasive whole-brain spatiotemporal dynamics mapping using participant-specific Geometric Basis Functions (GBFs). Resolves EEG/MEG inverse problem via cortical-surface eigenmodes. Validated across Meta-Source Benchmark,…
Geometric origin of exact mean-field reductions using Möbius symmetry and the Lorentzian Ansatz — proving the Cauchy-Lorentz family uniquely emerges as invariant under projective transport, unifying Ott-Antonsen and Montbrió-Pazó-Roxin reductions.