hiyenwong/ai_collection - Page 64

snn-mlir-neuromorphic-compilation

SNN-MLIR编译方法论 — MLIR方言用于将神经形态SNN从NIR编译到裸机C代码。支持浮点和量化数据，自动重缩放操作，生成依赖-free C11代码。

Source text: Chinese

socratic-swe-self-evolving-coding-agents

Self-Evolving Coding Agents via Trace-Derived Agent Skills (Socratic-SWE). Closed-loop framework that reuses solving traces to distill agent skills, generate targeted repair tasks, and iteratively improve Solver performance. Achieves 50.40% on SWE-bench…

flow-matching-brain-dynamics

Flow Matching with In-Context Priors for Out-of-Distribution Brain Dynamics — compositional conditional generation of neural time series using continuous normalizing flows, enabling zero-shot generalization to novel experimental conditions

sparse-mamba-decoder-qec

Sparse Mamba Decoder (SMD) for quantum error correction — a defect-centric neural decoder using Mamba state-space model that processes only active detection events (k ≪ d²R) achieving O(k) complexity on surface codes. 95-467x faster than Tesseract near-MLD…

sparse-mamba-qec-decoder

Sparse Mamba Decoder (SMD) for quantum error correction — defect-centric neural decoder using state-space (Mamba) backbone. Processes only k active detection events (O(k) complexity) instead of full O(d²R) syndrome array. Reduces MWPM logical error rate by up…

sparse-temporal-context-reconfiguration

Joint sparse coding and temporal dynamics methodology for context reconfiguration in lifelong learning. Identifies how sparsity and temporal structure in neural populations enable stable adaptation without catastrophic forgetting. Applies to SNN design,…

sparsity-neuromorphic-impulse-radio

Sparsity-aware event-driven impulse radio transceivers for reliable wireless neuromorphic inference. Activation: neuromorphic wireless, sparsity-aware radio, event-driven impulse, spike transmission.

spatiotemporal-tdann-mt-direction-maps

Spatiotemporal TDANN framework for modeling the emergence of direction-selective maps in primate MT cortex via self-supervised contrastive optimization with spatial regularization. Unifies ventral and dorsal stream topographic self-organization. arXiv:…

spatiotemporal-tdann

Spatiotemporal Topographic Deep Artificial Neural Network (TDANN) methodology for modeling dorsal stream cortical self-organization. Extends TDANN to motion-sensitive MT area using 3D ResNet trained with MoCo self-supervised contrastive learning on…

spider-directed-connectivity

SPIDER: non-parametric frequency-domain framework for inferring directed effective connectivity from incomplete, asynchronous recordings via stitched power-spectra and spectral factorization.

spike-agreement-dependent-plasticity

Spike Agreement Dependent Plasticity (SADP) - biologically inspired learning rule for SNNs using population-level correlation metrics instead of precise spike timing. Activation triggers: spike agreement, synaptic plasticity, SNN learning, bio-inspired…

spike-forecast-behavioral-decoding

Implicit behavioral decoding from next-step spike forecasts at population scale. Methodology from paper 'Implicit Behavioral Decoding from Next-Step Spike Forecasts at Population Scale' (arXiv: 2605.12999). Demonstrates that a single Mamba forecaster, trained…

spike-forecast-behavioral

Implicit behavioral decoding from next-step spike forecasts at population scale. Joint learning of neural population forecasting and behavioral readout from spiking activity. Use when: closed-loop BCI systems, neural population modeling, behavioral decoding…

spike-image-decoder

从神经脉冲重建视觉场景的深度学习框架（SID）。端到端解码器， 从视网膜神经节细胞脉冲重建静态图像和动态视频。 触发词：神经解码、视觉重建、脉冲解码、脑机接口、神经假体、 neural decoding, visual reconstruction, spike decoding, brain-machine interface。

Source text: Mixed languages

spike-ptsd-adversarial

Adversarial robustness analysis for Spiking Neural Networks. Spike-Triggered Decoupled (SpikT) attack methodology exploiting temporal vulnerability and STDP-induced sensitivity. Covers vulnerability analysis, attack implementation, and defense strategies.

spikedecoder-snn-gpt-architecture

SpikeDecoder - Fully SNN-based implementation of Transformer decoder block for NLP applications, achieving 87-93% energy reduction while maintaining performance

spiketimer-snn-copyright-protection

Active copyright protection for Spiking Neural Networks via temporal backdoor regularization

spikevla-spiking-vla-embodied-navigation

Paper analysis: SpikeVLA — a spiking Vision-Language-Action architecture for embodied navigation with energy-efficient inference (ICML 2026). Replaces dense continuous layers with event-driven spiking layers across vision (Spike-V), language (Spike-L), and…

spiking-arm-locomotor-coordination

Spiking Neural Network architecture coordinating bipedal locomotion and arm control via NEF/SPA with biologically grounded basal ganglia for humanoid robots

spiking-neuron-biological-plausibility-assessment

脉冲神经元生物学合理性自动化评估框架，系统化评估人工脉冲神经元与生物神经元的相似度，提供量化指标和可解释分析

Source text: Chinese

spiking-temporal-memory-stm-sequence-timing

Spiking Temporal Memory (sTM) model for learning sequence timing and control of replay speed in networks of spiking neurons

Source text: Chinese

spin-wave-superconducting-coupling

Strong coupling between propagating spin waves and microwave photons in superconducting resonator-magnetic thin film hybrid circuits. Design methodology for hybrid magnonic quantum systems using YIG-on-substrate integration.

spintune-quantum-sensor-reliability

SpinTune: RL-based optimization of dynamical decoupling pulse sequences for quantum sensor network reliability. Enables adaptive noise-aware DD sequence optimization to mitigate environmental decoherence.

split-primes-elekes-ronyai

Split primes and Elekes-Rónyai problem methodology for number theory counterexamples and arithmetic combinatorics

sqdr-cnn-spiking-quantum

SQDR-CNN methodology — joint training of convolutional SNNs and quantum circuits with surrogate gradient and quantum data-reupload for parameter-efficient hybrid models.

srf-similarity-representation-factorization

Similarity-Based Representation Factorization (SRF) methodology for recovering low-dimensional, non-negative, interpretable embeddings from similarity matrices

stimulus-symmetries-rsm-confound

Stimulus symmetries can confound representational similarity analyses — demonstrates how stimulus symmetries in neural network inputs cause functionally-equivalent representations to produce different, drifting RSM geometries. Based on arXiv:2605.21324.

stochastic-momentum-tracking-push-pull

Stochastic Momentum Tracking Push-Pull (SMTPP) algorithm for decentralized optimization over directed graphs. Decouples variance reduction from graph connectivity for robust convergence. Use for: directed graph optimization, decentralized learning, push-pull…

stochastic-quantum-neural-network

Stochastic Quantum Neural Network (SQNN) methodology for modeling neural dynamics using quantum superposition, entanglement, and unitary evolution beyond Von Neumann architecture constraints. Explores neuro-quantum correspondence between artificial neural…

stochastic-resonance-tinnitus-review

随机共振耳鸣模型十年综述。从幻听感知到适应性感官优化的神经计算理论，整合信息论、适应性信号检测、多通道听觉处理和跨模态可塑性。

Source text: Chinese

strand-survival-topological-analysis

STRAND (Survival Topological Representation ANalysis of Diagrams) treats persistence diagrams as survival data for hypothesis testing, effect sizes, and vectorisation in neuroscience applications.

structured-light-turbulent-channel-information

Analytical framework for structured light propagation through turbulent atmospheric channels using split-step mode-based approach. Power transfer between spatial modes scales linearly with distance, yielding matrix exponential solution for arbitrary…

sumo-whole-body-locomanipulation

Sim-to-real approach for dynamic whole-body loco-manipulation with test-time steering. Use when: (1) Designing legged robot control systems, (2) Implementing dynamic manipulation with heavy objects, (3) Building generalizable locomotion policies, (4)…

supervised-memory-training

Supervised Memory Training (SMT) methodology for parallel RNN pretraining without backpropagation through time.

surface-code-lattice-surgery

Superconducting surface-code processor with lattice-surgery logical operations — experimental demonstration of fault-tolerant logical Bell state preparation, Deutsch-Jozsa algorithm, and magic-state injection for non-Clifford rotations. Logical gate fidelity…

svp-lattice-tail-bound

Rogers tails methodology for the lattice gamma parameter — uniform random-lattice tail bounds for the SVP kissing-profile parameter. Provides probabilistic guarantees for lattice-dependent parameters in quantum and classical SVP algorithms. Activation: SVP…

swpc-directed-functional-connectivity

Sliding-window prediction correlation (SWpC) for time-varying directed functional connectivity. Embeds directional LTI models within sliding windows to estimate time-resolved information flow in brain networks, going beyond undirected correlation.

symboliclight-spike-gated-language

SymbolicLight V1: Spike-Gated Dual-Path Language Modeling with High Activation Sparsity and Sub-Billion-Scale Pre-Training Evidence. Research methodology from arXiv:2605.21333 (May 2026). First natively trained spiking language model combining binary LIF…

synaptic-matrix-eigenvalues-analysis

Spectral analysis of synaptic matrix eigenvalues for stability, transient dynamics, and memory capacity analysis in sparsely connected neural networks

synaptic-motif-mean-field

Mean-field theory bridging microscale synaptic motifs to macroscale heterogeneous neural dynamics. Derives low-rank equations for P-population networks where chain motifs induce correlations in synaptic variability, enabling microscopic fluctuations to…