资源推荐
按学习路径整理的参考资源,自动聚合自各篇文章的引用。
Transformer 核心机制 53 个资源
📄 论文
Attention Is All You Need
arxiv.org · 来源:
Transformer 网络结构总览 , QKV 的数据结构与直觉 , Attention 计算详解 , Multi-Head Attention , Positional Encoding — 让 Transformer 理解顺序
📄 论文
Language Models are Unsupervised Multitask Learners (GPT-2)
cdn.openai.com · 来源:
Transformer 网络结构总览 , Sampling & Decoding — 从概率到文本
🌐 网站
The Illustrated Transformer
jalammar.github.io · 来源:
Transformer 网络结构总览 , QKV 的数据结构与直觉 , Attention 计算详解 , Multi-Head Attention , KV Cache 原理
🌐 网站
LLM Visualization — Brendan Bycroft
bbycroft.net · 来源:
Transformer 网络结构总览 , QKV 的数据结构与直觉 , Attention 计算详解 , Multi-Head Attention , MQA 与 GQA , KV Cache 原理 , Prefill vs Decode 阶段
🌐 网站
Transformer Explainer — Georgia Tech / Polo Club
poloclub.github.io · 来源:
Transformer 网络结构总览 , QKV 的数据结构与直觉 , Attention 计算详解 , Multi-Head Attention , MQA 与 GQA
📄 论文
GLU Variants Improve Transformer
arxiv.org · 来源:
Transformer 网络结构总览
📄 论文
GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints
arxiv.org · 来源:
MQA 与 GQA
📄 论文
Fast Transformer Decoding: One Write-Head is All You Need
arxiv.org · 来源:
MQA 与 GQA
📄 论文
Mistral 7B
arxiv.org · 来源:
Attention 变体:从 Sliding Window 到 MLA
📄 论文
Gemma 2 Technical Report
arxiv.org · 来源:
Attention 变体:从 Sliding Window 到 MLA
📄 论文
Jamba: A Hybrid Transformer-Mamba Language Model
arxiv.org · 来源:
Attention 变体:从 Sliding Window 到 MLA , Hybrid 架构:Mamba 与 Attention 的融合
📄 论文
Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer (T5)
arxiv.org · 来源:
Attention 变体:从 Sliding Window 到 MLA
📄 论文
Flamingo: a Visual Language Model for Few-Shot Learning
arxiv.org · 来源:
Attention 变体:从 Sliding Window 到 MLA
📄 论文
DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
arxiv.org · 来源:
Attention 变体:从 Sliding Window 到 MLA , Mixture of Experts:稀疏激活的大模型架构
📄 论文
Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention
arxiv.org · 来源:
Attention 变体:从 Sliding Window 到 MLA
📄 论文
Retentive Network: A Successor to Transformer for Large Language Models
arxiv.org · 来源:
Attention 变体:从 Sliding Window 到 MLA
📄 论文
Parallelizing Linear Transformers with the Delta Rule over Sequence Length
arxiv.org · 来源:
Attention 变体:从 Sliding Window 到 MLA , Qwen3-Coder-Next 架构解析:当 SSM、Attention 与 MoE 三合一
📄 论文
Gated Delta Networks: Improving Mamba2 with Delta Rule
arxiv.org · 来源:
Attention 变体:从 Sliding Window 到 MLA , Qwen3-Coder-Next 架构解析:当 SSM、Attention 与 MoE 三合一
📄 论文
Efficient Memory Management for Large Language Model Serving with PagedAttention
arxiv.org · 来源:
KV Cache 原理
📄 论文
Efficiently Scaling Transformer Inference
arxiv.org · 来源:
Prefill vs Decode 阶段
📄 论文
LLM Inference Unveiled: Survey and Roofline Model Insights
arxiv.org · 来源:
Prefill vs Decode 阶段
📄 论文
FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
arxiv.org · 来源:
Flash Attention 分块原理
📄 论文
FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning
arxiv.org · 来源:
Flash Attention 分块原理
📄 论文
Self-Attention with Relative Position Representations
arxiv.org · 来源:
Positional Encoding — 让 Transformer 理解顺序
📄 论文
RoFormer: Enhanced Transformer with Rotary Position Embedding
arxiv.org · 来源:
Positional Encoding — 让 Transformer 理解顺序
📄 论文
Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation
arxiv.org · 来源:
Positional Encoding — 让 Transformer 理解顺序
📄 论文
The Curious Case of Neural Text Degeneration
arxiv.org · 来源:
Sampling & Decoding — 从概率到文本
📄 论文
Hierarchical Neural Story Generation
arxiv.org · 来源:
Sampling & Decoding — 从概率到文本
📄 论文
Perplexity — a Measure of the Difficulty of Speech Recognition Tasks
ieeexplore.ieee.org · 来源:
Sampling & Decoding — 从概率到文本
📄 论文
Fast Inference from Transformers via Speculative Decoding
arxiv.org · 来源:
Speculative Decoding — 猜测式解码加速
📄 论文
Accelerating Large Language Model Decoding with Speculative Sampling
arxiv.org · 来源:
Speculative Decoding — 猜测式解码加速
📄 论文
Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads
arxiv.org · 来源:
Speculative Decoding — 猜测式解码加速
📄 论文
Better & Faster Large Language Models via Multi-Token Prediction
arxiv.org · 来源:
Speculative Decoding — 猜测式解码加速
📄 论文
EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty
arxiv.org · 来源:
Speculative Decoding — 猜测式解码加速
📄 论文
EAGLE-2: Faster Inference of Language Models with Dynamic Draft Trees
arxiv.org · 来源:
Speculative Decoding — 猜测式解码加速
📄 论文
Break the Sequential Dependency of LLM Inference Using Lookahead Decoding
arxiv.org · 来源:
Speculative Decoding — 猜测式解码加速
🌐 网站
DeepSeek-V3 Technical Report
arxiv.org · 来源:
Speculative Decoding — 猜测式解码加速 , Mixture of Experts:稀疏激活的大模型架构
📄 论文
EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Test
arxiv.org · 来源:
Speculative Decoding — 猜测式解码加速
📄 论文
Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity
arxiv.org · 来源:
Mixture of Experts:稀疏激活的大模型架构
📄 论文
Mixtral of Experts
arxiv.org · 来源:
Mixture of Experts:稀疏激活的大模型架构
📄 论文
GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding
arxiv.org · 来源:
Mixture of Experts:稀疏激活的大模型架构
📄 论文
Efficiently Modeling Long Sequences with Structured State Spaces (S4)
arxiv.org · 来源:
状态空间模型与 Mamba
📄 论文
Mamba: Linear-Time Sequence Modeling with Selective State Spaces
arxiv.org · 来源:
状态空间模型与 Mamba
📄 论文
Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality
arxiv.org · 来源:
状态空间模型与 Mamba
📄 论文
HiPPO: Recurrent Memory with Optimal Polynomial Projections
arxiv.org · 来源:
状态空间模型与 Mamba
📄 论文
Hungry Hungry Hippos: Towards Language Modeling with State Space Models (H3)
arxiv.org · 来源:
状态空间模型与 Mamba
📄 论文
On the Parameterization and Initialization of Diagonal State Space Models (S4D)
arxiv.org · 来源:
状态空间模型与 Mamba
🌐 网站
Zamba2-Small: A Hybrid SSM-Transformer Model
zyphra.com · 来源:
Hybrid 架构:Mamba 与 Attention 的融合
📄 论文
Hymba: A Hybrid-head Architecture for Small Language Models
arxiv.org · 来源:
Hybrid 架构:Mamba 与 Attention 的融合
📄 论文
An Empirical Study of Mamba-based Language Models
arxiv.org · 来源:
Hybrid 架构:Mamba 与 Attention 的融合
📄 论文
Repeat After Me: Transformers are Better than State Space Models at Copying
arxiv.org · 来源:
Hybrid 架构:Mamba 与 Attention 的融合
📄 论文
Qwen3 Technical Report
arxiv.org · 来源:
Qwen3-Coder-Next 架构解析:当 SSM、Attention 与 MoE 三合一
💻 代码
Ollama - Qwen3-Next 模型实现
github.com · 来源:
Qwen3-Coder-Next 架构解析:当 SSM、Attention 与 MoE 三合一
Transformer 跨模态应用 39 个资源
📄 论文
Efficient Estimation of Word Representations in Vector Space
arxiv.org · 来源:
从文本到向量:Tokenization 与词嵌入
📄 论文
Neural Machine Translation of Rare Words with Subword Units
arxiv.org · 来源:
从文本到向量:Tokenization 与词嵌入
📄 论文
GloVe: Global Vectors for Word Representation
nlp.stanford.edu · 来源:
从文本到向量:Tokenization 与词嵌入
📄 论文
SentencePiece: A simple and language independent subword tokenizer
arxiv.org · 来源:
从文本到向量:Tokenization 与词嵌入
🌐 网站
The Illustrated Word2Vec
jalammar.github.io · 来源:
从文本到向量:Tokenization 与词嵌入
🌐 网站
Hugging Face Tokenizer Summary
huggingface.co · 来源:
从文本到向量:Tokenization 与词嵌入
📄 论文
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
arxiv.org · 来源:
BERT 与 GPT:理解与生成的两条路线
📄 论文
Improving Language Understanding by Generative Pre-Training
cdn.openai.com · 来源:
BERT 与 GPT:理解与生成的两条路线
📄 论文
Language Models are Unsupervised Multitask Learners
cdn.openai.com · 来源:
BERT 与 GPT:理解与生成的两条路线
📄 论文
Language Models are Few-Shot Learners
arxiv.org · 来源:
BERT 与 GPT:理解与生成的两条路线
📄 论文
BERT for Joint Intent Classification and Slot Filling
arxiv.org · 来源:
BERT 与 GPT:理解与生成的两条路线
📄 论文
Scaling Laws for Neural Language Models
arxiv.org · 来源:
BERT 与 GPT:理解与生成的两条路线
📄 论文
Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
arxiv.org · 来源:
句子嵌入:从 Token 级到语义检索
📄 论文
Text Embeddings by Weakly-Supervised Contrastive Pre-training
arxiv.org · 来源:
句子嵌入:从 Token 级到语义检索
📄 论文
C-Pack: Packaged Resources To Advance General Chinese Embedding
arxiv.org · 来源:
句子嵌入:从 Token 级到语义检索
📄 论文
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
arxiv.org · 来源:
句子嵌入:从 Token 级到语义检索
📄 论文
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
arxiv.org · 来源:
Vision Transformer:当图像变成 Token 序列
📄 论文
Training data-efficient image transformers & distillation through attention
arxiv.org · 来源:
Vision Transformer:当图像变成 Token 序列
📄 论文
Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
arxiv.org · 来源:
Vision Transformer:当图像变成 Token 序列
📄 论文
Learning Transferable Visual Models From Natural Language Supervision
arxiv.org · 来源:
多模态对齐:CLIP 与跨模态嵌入空间
📄 论文
Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision
arxiv.org · 来源:
多模态对齐:CLIP 与跨模态嵌入空间
📄 论文
Sigmoid Loss for Language Image Pre-Training
arxiv.org · 来源:
多模态对齐:CLIP 与跨模态嵌入空间
📄 论文
Visual Instruction Tuning
arxiv.org · 来源:
多模态对齐:CLIP 与跨模态嵌入空间
📄 论文
Denoising Diffusion Probabilistic Models
arxiv.org · 来源:
扩散模型基础:从噪声中生成
📄 论文
Denoising Diffusion Implicit Models
arxiv.org · 来源:
扩散模型基础:从噪声中生成
📄 论文
High-Resolution Image Synthesis with Latent Diffusion Models
arxiv.org · 来源:
扩散模型基础:从噪声中生成
📄 论文
Classifier-Free Diffusion Guidance
arxiv.org · 来源:
扩散模型基础:从噪声中生成
📄 论文
Scalable Diffusion Models with Transformers
arxiv.org · 来源:
Diffusion Transformer:用 Transformer 做图像生成
📄 论文
Scaling Rectified Flow Transformers for High-Resolution Image Synthesis
arxiv.org · 来源:
Diffusion Transformer:用 Transformer 做图像生成
📄 论文
Video generation models as world simulators
openai.com · 来源:
视频生成:时空注意力与 Sora 架构
📄 论文
Make-A-Video: Text-to-Video Generation without Text-Video Data
arxiv.org · 来源:
视频生成:时空注意力与 Sora 架构
📄 论文
Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models
arxiv.org · 来源:
视频生成:时空注意力与 Sora 架构
📄 论文
Robust Speech Recognition via Large-Scale Weak Supervision
arxiv.org · 来源:
语音与 Transformer:从 Whisper 到 VALL-E
📄 论文
Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers
arxiv.org · 来源:
语音与 Transformer:从 Whisper 到 VALL-E
📄 论文
High Fidelity Neural Audio Compression
arxiv.org · 来源:
语音与 Transformer:从 Whisper 到 VALL-E
📄 论文
Simple and Controllable Music Generation
arxiv.org · 来源:
音乐生成:当 Transformer 学会作曲
📄 论文
Jukebox: A Generative Model for Music
arxiv.org · 来源:
音乐生成:当 Transformer 学会作曲
📄 论文
MusicLM: Generating Music From Text
arxiv.org · 来源:
音乐生成:当 Transformer 学会作曲
📄 论文
Fast Timing-Conditioned Latent Audio Diffusion
arxiv.org · 来源:
音乐生成:当 Transformer 学会作曲
LLM 量化技术 27 个资源
📄 论文
A Survey of Quantization Methods for Efficient Neural Network Inference
arxiv.org · 来源:
量化基础
📄 论文
Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation
arxiv.org · 来源:
量化基础
📄 论文
FP8 Formats for Deep Learning
arxiv.org · 来源:
量化基础 , 推理时量化:KV Cache 与 Activation 量化
📄 论文
GPTQ: Accurate Post-Training Quantization for Generative Pre-Trained Transformers
arxiv.org · 来源:
PTQ 权重量化:从 GPTQ 到 AWQ , llama.cpp 量化方案
📄 论文
AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
arxiv.org · 来源:
PTQ 权重量化:从 GPTQ 到 AWQ , llama.cpp 量化方案
📄 论文
SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models
arxiv.org · 来源:
PTQ 权重量化:从 GPTQ 到 AWQ
📄 论文
Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference
arxiv.org · 来源:
量化感知训练 (QAT)
📄 论文
BitNet: Scaling 1-bit Transformers for Large Language Models
arxiv.org · 来源:
量化感知训练 (QAT)
📄 论文
The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits
arxiv.org · 来源:
量化感知训练 (QAT)
📄 论文
QLoRA: Efficient Finetuning of Quantized LLMs
arxiv.org · 来源:
量化感知训练 (QAT)
📄 论文
LQ-LoRA: Low-rank Plus Quantized Matrix Decomposition for Efficient Language Model Finetuning
arxiv.org · 来源:
量化感知训练 (QAT)
📄 论文
KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization
arxiv.org · 来源:
推理时量化:KV Cache 与 Activation 量化
📄 论文
KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache
arxiv.org · 来源:
推理时量化:KV Cache 与 Activation 量化
💻 代码
llama.cpp Quantization Types
github.com · 来源:
llama.cpp 量化方案
💻 代码
K-quant PR
github.com · 来源:
llama.cpp 量化方案
🌐 网站
NVIDIA Model Optimizer GitHub
github.com · 来源:
量化与模型转换工具链全景
🌐 网站
vLLM Quantization - LLM Compressor
github.com · 来源:
量化与模型转换工具链全景
🌐 网站
Microsoft Olive Documentation - Why Olive
microsoft.github.io · 来源:
量化与模型转换工具链全景
🌐 网站
Apple coremltools Optimization Overview
apple.github.io · 来源:
量化与模型转换工具链全景
🌐 网站
AMD Quark Documentation
quark.docs.amd.com · 来源:
量化与模型转换工具链全景
🌐 网站
Google AI Edge Torch GitHub
github.com · 来源:
量化与模型转换工具链全景
🌐 网站
NNCF GitHub Repository
github.com · 来源:
量化与模型转换工具链全景
🌐 网站
Optimum Intel Documentation
huggingface.co · 来源:
量化与模型转换工具链全景 , 动手:HF → GGUF / ONNX / OpenVINO 三条路径端到端
🌐 网站
llama.cpp GitHub Repository
github.com · 来源:
动手:HF → GGUF / ONNX / OpenVINO 三条路径端到端
🌐 网站
ONNX Runtime Documentation
onnxruntime.ai · 来源:
动手:HF → GGUF / ONNX / OpenVINO 三条路径端到端
🌐 网站
OpenVINO Documentation
docs.openvino.ai · 来源:
动手:HF → GGUF / ONNX / OpenVINO 三条路径端到端
🌐 网站
lm-evaluation-harness
github.com · 来源:
动手:HF → GGUF / ONNX / OpenVINO 三条路径端到端
vLLM + SGLang 推理引擎深度解析 13 个资源
📄 论文
Efficient Memory Management for Large Language Model Serving with PagedAttention
arxiv.org · 来源:
LLM 推理引擎全景:vLLM、SGLang、Ollama 与 TensorRT-LLM , PagedAttention 与 Continuous Batching , 调度与抢占:推理引擎的 Scheduler
📄 论文
SGLang: Efficient Execution of Structured Language Model Programs
arxiv.org · 来源:
LLM 推理引擎全景:vLLM、SGLang、Ollama 与 TensorRT-LLM , 前缀缓存与 RadixAttention , SGLang 编程模型与结构化输出
🌐 网站
NVIDIA TensorRT-LLM Documentation
nvidia.github.io · 来源:
LLM 推理引擎全景:vLLM、SGLang、Ollama 与 TensorRT-LLM
🌐 网站
Ollama GitHub Repository
github.com · 来源:
LLM 推理引擎全景:vLLM、SGLang、Ollama 与 TensorRT-LLM
📄 论文
Orca: A Distributed Serving System for Transformer-Based Generative Models
arxiv.org · 来源:
PagedAttention 与 Continuous Batching
🌐 网站
vLLM: Easy, Fast, and Cheap LLM Serving with PagedAttention
blog.vllm.ai · 来源:
PagedAttention 与 Continuous Batching
📄 论文
Sarathi: Efficient LLM Inference by Piggybacking Decodes with Chunked Prefills
arxiv.org · 来源:
调度与抢占:推理引擎的 Scheduler
📄 论文
Taming Throughput-Latency Tradeoff in LLM Inference with Sarathi-Serve
arxiv.org · 来源:
调度与抢占:推理引擎的 Scheduler
🌐 网站
vLLM Automatic Prefix Caching
docs.vllm.ai · 来源:
前缀缓存与 RadixAttention
📄 论文
Trie Memory
dl.acm.org · 来源:
前缀缓存与 RadixAttention
📄 论文
Efficient Guided Generation for Large Language Models
arxiv.org · 来源:
SGLang 编程模型与结构化输出
🌐 网站
Fast JSON Decoding for Local LLMs with Compressed Finite State Machine
lmsys.org · 来源:
SGLang 编程模型与结构化输出
🌐 网站
SGLang Documentation — Structured Outputs
docs.sglang.io · 来源:
SGLang 编程模型与结构化输出
📄 论文
RouteLLM: Learning to Route LLMs with Preference Data
arxiv.org · 来源:
Model Routing 全景:为什么一个模型不够 , 路由分类器:让小模型决定谁来回答 , RouteLLM 实战:从偏好数据到生产路由 , 因子分解机与 LLM 路由:从 FM 理论到 MF 路由器 , 在线学习与成本优化:路由也需要持续进化
📄 论文
FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance
arxiv.org · 来源:
Model Routing 全景:为什么一个模型不够 , 级联与自验证:先试便宜的,不行再升级
📄 论文
AutoMix: Automatically Mixing Language Models
arxiv.org · 来源:
Model Routing 全景:为什么一个模型不够 , 级联与自验证:先试便宜的,不行再升级
🌐 网站
RouteLLM GitHub Repository
github.com · 来源:
Model Routing 全景:为什么一个模型不够 , RouteLLM 实战:从偏好数据到生产路由
📄 论文
Evaluating Small Language Models for Front-Door Routing
arxiv.org · 来源:
路由分类器:让小模型决定谁来回答
🌐 网站
semantic-router: Superfast Decision-Making Layer
github.com · 来源:
路由分类器:让小模型决定谁来回答
📄 论文
Factorization Machines
csie.ntu.edu.tw · 来源:
因子分解机与 LLM 路由:从 FM 理论到 MF 路由器
📄 论文
Factorization Machines with libFM
dl.acm.org · 来源:
因子分解机与 LLM 路由:从 FM 理论到 MF 路由器
📄 论文
Confidence-Driven LLM Router
arxiv.org · 来源:
级联与自验证:先试便宜的,不行再升级
📄 论文
ConsRoute: Consistency-Driven LLM Routing for Cloud-Edge-Device
arxiv.org · 来源:
Hybrid LLM:本地与云端的智能路由
📄 论文
HybridFlow: Subtask-level DAG Routing
arxiv.org · 来源:
Hybrid LLM:本地与云端的智能路由
📄 论文
PRISM: Privacy-Sensitive Entity-Level LLM Routing
arxiv.org · 来源:
Hybrid LLM:本地与云端的智能路由
📄 论文
Bridging On-Device and Cloud LLMs for Collaborative Reasoning
arxiv.org · 来源:
Hybrid LLM:本地与云端的智能路由
📄 论文
Robust Batch-Level LLM Routing
arxiv.org · 来源:
在线学习与成本优化:路由也需要持续进化
📄 论文
Council Mode: Multi-LLM Collaboration for Hallucination Reduction
arxiv.org · 来源:
多模型协作:从选一个到用多个
🌐 网站
Mixture of Agents - Together AI
together.ai · 来源:
多模型协作:从选一个到用多个
LLM 评估与 Benchmark 深度解析 27 个资源
📄 论文
Measuring Massive Multitask Language Understanding (MMLU)
arxiv.org · 来源:
Benchmark 全景与评估方法论
🌐 网站
lm-evaluation-harness
github.com · 来源:
Benchmark 全景与评估方法论 , 优化对精度的影响 , lm-eval-harness 实操指南
📄 论文
Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
arxiv.org · 来源:
Benchmark 全景与评估方法论
🌐 网站
LiveBench
livebench.ai · 来源:
Benchmark 全景与评估方法论 , 排行榜解读与模型选型
📄 论文
MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark
arxiv.org · 来源:
知识与推理 Benchmark
📄 论文
GPQA: A Graduate-Level Google-Proof Q&A Benchmark
arxiv.org · 来源:
知识与推理 Benchmark
📄 论文
Measuring Mathematical Problem Solving With the MATH Dataset
arxiv.org · 来源:
知识与推理 Benchmark
📄 论文
Evaluating Large Language Models Trained on Code
arxiv.org · 来源:
代码 Benchmark
📄 论文
SWE-bench: Can Language Models Resolve Real-World GitHub Issues?
arxiv.org · 来源:
代码 Benchmark , SWE-bench 实操指南
📄 论文
Is Your Code Generated by ChatGPT Really Correct? (EvalPlus)
arxiv.org · 来源:
代码 Benchmark
🌐 网站
Berkeley Function Calling Leaderboard (BFCL)
gorilla.cs.berkeley.edu · 来源:
Agent 与 Tool Use Benchmark , BFCL 实操指南
📄 论文
GAIA: A Benchmark for General AI Assistants
arxiv.org · 来源:
Agent 与 Tool Use Benchmark
📄 论文
WebArena: A Realistic Web Environment for Building Autonomous Agents
arxiv.org · 来源:
Agent 与 Tool Use Benchmark
🌐 网站
Google Gemma 2 Technical Report
ai.google.dev · 来源:
模型发布 Benchmark 标配解析
🌐 网站
Microsoft Phi-3 Technical Report
arxiv.org · 来源:
模型发布 Benchmark 标配解析
📄 论文
Qwen2.5 Technical Report
arxiv.org · 来源:
模型发布 Benchmark 标配解析
🌐 网站
Meta Llama 3.1 Model Card
huggingface.co · 来源:
模型发布 Benchmark 标配解析
🌐 网站
Open LLM Leaderboard
huggingface.co · 来源:
模型发布 Benchmark 标配解析 , 排行榜解读与模型选型
🌐 网站
OpenVINO Neural Network Compression Framework (NNCF)
github.com · 来源:
优化对精度的影响
🌐 网站
Optimum Intel
huggingface.co · 来源:
优化对精度的影响
🌐 网站
llama.cpp
github.com · 来源:
优化对精度的影响
🌐 网站
Chatbot Arena (LMSYS)
lmarena.ai · 来源:
排行榜解读与模型选型
🌐 网站
Artificial Analysis LLM Leaderboard
artificialanalysis.ai · 来源:
排行榜解读与模型选型
🌐 网站
lm-eval Documentation
lm-evaluation-harness.readthedocs.io · 来源:
lm-eval-harness 实操指南
🌐 网站
SWE-bench GitHub
github.com · 来源:
SWE-bench 实操指南
🌐 网站
SWE-agent GitHub
github.com · 来源:
SWE-bench 实操指南
🌐 网站
Gorilla / BFCL GitHub
github.com · 来源:
BFCL 实操指南
Ollama + llama.cpp 深度解析 20 个资源
💻 代码
Ollama GitHub
github.com · 来源:
Ollama + llama.cpp 架构总览 , 一次推理的完整旅程 , KV Cache 与 Batch 调度 , 服务层与调度
💻 代码
llama.cpp GitHub
github.com · 来源:
Ollama + llama.cpp 架构总览 , 一次推理的完整旅程 , 计算图与推理引擎
💻 代码
GGML GitHub
github.com · 来源:
Ollama + llama.cpp 架构总览 , 计算图与推理引擎 , 硬件后端
📄 论文
Qwen3 Technical Report
arxiv.org · 来源:
一次推理的完整旅程
💻 代码
GGUF Specification
github.com · 来源:
GGUF 模型格式
🌐 网站
Safetensors Documentation
huggingface.co · 来源:
GGUF 模型格式
🌐 网站
ONNX
onnx.ai · 来源:
GGUF 模型格式
💻 代码
llama.cpp Quantization Types
github.com · 来源:
llama.cpp 量化方案
💻 代码
K-quant PR
github.com · 来源:
llama.cpp 量化方案
📄 论文
GPTQ: Accurate Post-Training Quantization
arxiv.org · 来源:
llama.cpp 量化方案
📄 论文
AWQ: Activation-aware Weight Quantization
arxiv.org · 来源:
llama.cpp 量化方案
📄 论文
FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
arxiv.org · 来源:
计算图与推理引擎
📄 论文
Efficient Memory Management for LLM Serving with PagedAttention
arxiv.org · 来源:
KV Cache 与 Batch 调度
🌐 网站
CUDA Programming Guide
docs.nvidia.com · 来源:
硬件后端
🌐 网站
Metal Shading Language
developer.apple.com · 来源:
硬件后端
🌐 网站
Vulkan Compute
khronos.org · 来源:
硬件后端
💻 代码
Ollama FAQ
github.com · 来源:
服务层与调度
💻 代码
Ollama Modelfile
github.com · 来源:
模型生态
💻 代码
Ollama API
github.com · 来源:
模型生态
📄 论文
LLaVA: Visual Instruction Tuning
arxiv.org · 来源:
模型生态
llama.cpp 源码精读 1 个资源
AI Compute Stack 21 个资源
🌐 网站
NVIDIA CUDA C++ Programming Guide
docs.nvidia.com · 来源:
AI Compute Stack 全景 — 从推理框架到硬件指令集 , GPU Architecture — 从晶体管到线程 , CUDA 编程模型 — 从代码到硬件
🌐 网站
Khronos OpenCL Specification
khronos.org · 来源:
AI Compute Stack 全景 — 从推理框架到硬件指令集
🌐 网站
Khronos SYCL Specification
khronos.org · 来源:
AI Compute Stack 全景 — 从推理框架到硬件指令集
🌐 网站
Intel oneAPI Level Zero Specification
spec.oneapi.io · 来源:
AI Compute Stack 全景 — 从推理框架到硬件指令集
🌐 网站
AMD ROCm HIP Programming Guide
rocm.docs.amd.com · 来源:
AI Compute Stack 全景 — 从推理框架到硬件指令集
🌐 网站
Apple Metal Shading Language Specification
developer.apple.com · 来源:
AI Compute Stack 全景 — 从推理框架到硬件指令集
💻 代码
ggml / llama.cpp
github.com · 来源:
AI Compute Stack 全景 — 从推理框架到硬件指令集
🌐 网站
ONNX Runtime Documentation
onnxruntime.ai · 来源:
AI Compute Stack 全景 — 从推理框架到硬件指令集
🌐 网站
NVIDIA H100 Tensor Core GPU Architecture Whitepaper
resources.nvidia.com · 来源:
GPU Architecture — 从晶体管到线程 , 矩阵加速单元 — Tensor Core 与 XMX
📄 论文
Why Systolic Architectures? — H.T. Kung
cs.virginia.edu · 来源:
矩阵加速单元 — Tensor Core 与 XMX
🌐 网站
NVIDIA PTX ISA — Matrix Multiply-Accumulate
docs.nvidia.com · 来源:
矩阵加速单元 — Tensor Core 与 XMX
🌐 网站
Intel Xe2 Architecture — Xe-Core and XMX
intel.com · 来源:
矩阵加速单元 — Tensor Core 与 XMX , CUDA 编程模型 — 从代码到硬件
📄 论文
DeepSeek-V3 Technical Report
arxiv.org · 来源:
矩阵加速单元 — Tensor Core 与 XMX
🌐 网站
NVIDIA Kernel Profiling Guide — Memory Coalescing
docs.nvidia.com · 来源:
CUDA 编程模型 — 从代码到硬件
🌐 网站
CUDA Occupancy Calculator
docs.nvidia.com · 来源:
CUDA 编程模型 — 从代码到硬件
🌐 网站
SYCL 2020 Specification
registry.khronos.org · 来源:
CUDA 编程模型 — 从代码到硬件
🌐 网站
CUTLASS: Fast Linear Algebra in CUDA C++
github.com · 来源:
GEMM 优化 — 从 Naive 到极致
🌐 网站
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance
siboehm.com · 来源:
GEMM 优化 — 从 Naive 到极致
🌐 网站
CUDA C++ Programming Guide — Warp Matrix Functions
docs.nvidia.com · 来源:
GEMM 优化 — 从 Naive 到极致
🌐 网站
Intel oneAPI DPC++ — joint_matrix Extension
github.com · 来源:
GEMM 优化 — 从 Naive 到极致
📄 论文
Dissecting the NVIDIA Volta GPU Architecture via Microbenchmarking
arxiv.org · 来源:
GEMM 优化 — 从 Naive 到极致
图编译与优化 79 个资源
📝 博客
PyTorch 2.0: Our next generation release
pytorch.org · 来源:
全景图:ML 编译器的世界 , 计算图捕获:TorchDynamo、AOTAutograd 与 Functionalization
🌐 网站
MLIR: Multi-Level Intermediate Representation
mlir.llvm.org · 来源:
全景图:ML 编译器的世界
🌐 网站
Triton Language and Compiler
triton-lang.org · 来源:
全景图:ML 编译器的世界
📄 论文
TVM: An Automated End-to-End Optimizing Compiler for Deep Learning
arxiv.org · 来源:
全景图:ML 编译器的世界
📄 论文
MLIR: A Compiler Infrastructure for the End of Moore's Law
arxiv.org · 来源:
全景图:ML 编译器的世界 , IR 设计(上):SSA、FX IR 与 MLIR Dialect , IR 设计(下):Progressive Lowering 与多层 IR
📝 博客
TorchDynamo: An Experiment in Dynamic Python Bytecode Transformation
dev-discuss.pytorch.org · 来源:
全景图:ML 编译器的世界 , 计算图捕获:TorchDynamo、AOTAutograd 与 Functionalization
🌐 网站
PEP 523 – Adding a frame evaluation API to CPython
peps.python.org · 来源:
计算图捕获:TorchDynamo、AOTAutograd 与 Functionalization
🌐 网站
torch.compiler — PyTorch Documentation
pytorch.org · 来源:
计算图捕获:TorchDynamo、AOTAutograd 与 Functionalization
🌐 网站
AOT Autograd — How to use and optimize?
pytorch.org · 来源:
计算图捕获:TorchDynamo、AOTAutograd 与 Functionalization
📄 论文
Efficiently Computing Static Single Assignment Form and the Control Dependence Graph
dl.acm.org · 来源:
IR 设计(上):SSA、FX IR 与 MLIR Dialect , 图优化 Pass(上):数据流分析基础与通用 Pass 模式
🌐 网站
torch.fx — PyTorch Documentation
pytorch.org · 来源:
IR 设计(上):SSA、FX IR 与 MLIR Dialect , 图优化 Pass(上):数据流分析基础与通用 Pass 模式
🌐 网站
MLIR Language Reference
mlir.llvm.org · 来源:
IR 设计(上):SSA、FX IR 与 MLIR Dialect
🌐 网站
MLIR Dialects
mlir.llvm.org · 来源:
IR 设计(上):SSA、FX IR 与 MLIR Dialect
🌐 网站
MLIR Dialect Conversion
mlir.llvm.org · 来源:
IR 设计(下):Progressive Lowering 与多层 IR
🌐 网站
MLIR Bufferization
mlir.llvm.org · 来源:
IR 设计(下):Progressive Lowering 与多层 IR
🌐 网站
MLIR Pass Infrastructure
mlir.llvm.org · 来源:
IR 设计(下):Progressive Lowering 与多层 IR , 图优化 Pass(上):数据流分析基础与通用 Pass 模式
💻 代码
torch-mlir: PyTorch to MLIR compiler
github.com · 来源:
IR 设计(下):Progressive Lowering 与多层 IR
📄 论文
A Unified Approach to Global Program Optimization
dl.acm.org · 来源:
图优化 Pass(上):数据流分析基础与通用 Pass 模式
🌐 网站
MLIR Canonicalization
mlir.llvm.org · 来源:
图优化 Pass(上):数据流分析基础与通用 Pass 模式
📄 论文
Constant Propagation with Conditional Branches
dl.acm.org · 来源:
图优化 Pass(上):数据流分析基础与通用 Pass 模式
🌐 网站
PyTorch FX Subgraph Rewriter
pytorch.org · 来源:
图优化 Pass(上):数据流分析基础与通用 Pass 模式
🌐 网站
MLIR Declarative Rewrite Rules (DRR)
mlir.llvm.org · 来源:
图优化 Pass(中):高级优化与 Pattern Matching
🌐 网站
MLIR PDL — Pattern Description Language
mlir.llvm.org · 来源:
图优化 Pass(中):高级优化与 Pattern Matching
🌐 网站
torch.fx — Subgraph Rewriting
pytorch.org · 来源:
图优化 Pass(中):高级优化与 Pattern Matching
📄 论文
Tensor Comprehensions: Framework-Agnostic High-Performance Machine Learning Abstractions
arxiv.org · 来源:
图优化 Pass(中):高级优化与 Pattern Matching
🌐 网站
NVIDIA Tensor Core Programming
docs.nvidia.com · 来源:
图优化 Pass(中):高级优化与 Pattern Matching
📄 论文
A Practical Automatic Polyhedral Parallelizer and Locality Optimizer
dl.acm.org · 来源:
图优化 Pass(下):Polyhedral 优化与循环变换
🌐 网站
MLIR Affine Dialect
mlir.llvm.org · 来源:
图优化 Pass(下):Polyhedral 优化与循环变换
📄 论文
Polyhedral Compilation as a Design Pattern for Compiler Construction
link.springer.com · 来源:
图优化 Pass(下):Polyhedral 优化与循环变换
🌐 网站
MLIR Transform Dialect
mlir.llvm.org · 来源:
图优化 Pass(下):Polyhedral 优化与循环变换 , 自动调优与端到端实战
📄 论文
Optimizing Compilers for Modern Architectures
elsevier.com · 来源:
图优化 Pass(下):Polyhedral 优化与循环变换
🌐 网站
Polly - Polyhedral optimizations for LLVM
polly.llvm.org · 来源:
图优化 Pass(下):Polyhedral 优化与循环变换
📄 论文
Integer Set Library: A Library for Manipulating Integer Sets
libisl.sourceforge.io · 来源:
图优化 Pass(下):Polyhedral 优化与循环变换
🌐 网站
MLIR Linalg Dialect
mlir.llvm.org · 来源:
图优化 Pass(下):Polyhedral 优化与循环变换 , 算子融合(下):Cost Model 与融合实战
📄 论文
FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
arxiv.org · 来源:
算子融合(上):融合类型学与判定算法 , 算子融合(下):Cost Model 与融合实战 , Tiling 策略与内存层次优化
📄 论文
PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transformation and Graph Compilation
dl.acm.org · 来源:
算子融合(上):融合类型学与判定算法 , 算子融合(下):Cost Model 与融合实战 , Dynamic Shapes:从捕获到执行的全链路挑战
🌐 网站
TorchInductor: a PyTorch-native Compiler
dev-discuss.pytorch.org · 来源:
算子融合(上):融合类型学与判定算法
📄 论文
XLA: Optimizing Compiler for Machine Learning
tensorflow.org · 来源:
算子融合(上):融合类型学与判定算法
🌐 网站
Roofline Model
docs.nersc.gov · 来源:
算子融合(上):融合类型学与判定算法 , 算子融合(下):Cost Model 与融合实战
📄 论文
FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning
arxiv.org · 来源:
算子融合(下):Cost Model 与融合实战
📄 论文
FlashAttention-3: Fast and Accurate Attention with Asynchrony and Low-precision
arxiv.org · 来源:
算子融合(下):Cost Model 与融合实战
📄 论文
Roofline: An Insightful Visual Performance Model for Multicore Architectures
www2.eecs.berkeley.edu · 来源:
Tiling 策略与内存层次优化
🌐 网站
NVIDIA CUDA C++ Programming Guide — Shared Memory
docs.nvidia.com · 来源:
Tiling 策略与内存层次优化
🌐 网站
CUTLASS: CUDA Templates for Linear Algebra Subroutines
github.com · 来源:
Tiling 策略与内存层次优化 , 代码生成(上):指令选择、Vectorization 与 Register Allocation
🌐 网站
Triton Language Documentation
triton-lang.org · 来源:
Tiling 策略与内存层次优化 , 代码生成(下):Triton Pipeline、编译器后端与数值正确性
🌐 网站
NVIDIA A100 GPU Architecture Whitepaper
images.nvidia.com · 来源:
Tiling 策略与内存层次优化
🌐 网站
torch.compile Dynamic Shapes Documentation
pytorch.org · 来源:
Dynamic Shapes:从捕获到执行的全链路挑战
🌐 网站
TorchDynamo Deep Dive
pytorch.org · 来源:
Dynamic Shapes:从捕获到执行的全链路挑战
🌐 网站
MLIR Tensor Type — Dynamic Dimensions
mlir.llvm.org · 来源:
Dynamic Shapes:从捕获到执行的全链路挑战
🌐 网站
NVIDIA CUDA C++ Programming Guide — PTX ISA
docs.nvidia.com · 来源:
代码生成(上):指令选择、Vectorization 与 Register Allocation
🌐 网站
LLVM Code Generator Documentation
llvm.org · 来源:
代码生成(上):指令选择、Vectorization 与 Register Allocation
🌐 网站
NVIDIA GPU Architecture — Execution Units
docs.nvidia.com · 来源:
代码生成(上):指令选择、Vectorization 与 Register Allocation
📄 论文
Triton: An Intermediate Language and Compiler for Tiled Neural Network Computations
eecs.harvard.edu · 来源:
代码生成(上):指令选择、Vectorization 与 Register Allocation , 代码生成(下):Triton Pipeline、编译器后端与数值正确性 , 自动调优与端到端实战
🌐 网站
MLIR GPU Dialect
mlir.llvm.org · 来源:
代码生成(下):Triton Pipeline、编译器后端与数值正确性
🌐 网站
IREE Compiler and Runtime
iree.dev · 来源:
代码生成(下):Triton Pipeline、编译器后端与数值正确性
🌐 网站
TensorRT Developer Guide
docs.nvidia.com · 来源:
代码生成(下):Triton Pipeline、编译器后端与数值正确性
🌐 网站
What Every Computer Scientist Should Know About Floating-Point Arithmetic
docs.oracle.com · 来源:
代码生成(下):Triton Pipeline、编译器后端与数值正确性
📄 论文
A Survey of Quantization Methods for Efficient Neural Network Inference
arxiv.org · 来源:
量化编译与混合精度优化
📄 论文
GPTQ: Accurate Post-Training Quantization for Generative Pre-Trained Transformers
arxiv.org · 来源:
量化编译与混合精度优化
📄 论文
AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
arxiv.org · 来源:
量化编译与混合精度优化
📄 论文
FP8 Formats for Deep Learning
arxiv.org · 来源:
量化编译与混合精度优化
🌐 网站
PyTorch Quantization Documentation
pytorch.org · 来源:
量化编译与混合精度优化
🌐 网站
TensorRT Quantization Toolkit
docs.nvidia.com · 来源:
量化编译与混合精度优化
📄 论文
GSPMD: General and Scalable Parallelization for ML Computation Graphs
arxiv.org · 来源:
分布式编译与图分割
📄 论文
Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism
arxiv.org · 来源:
分布式编译与图分割
📄 论文
GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism
arxiv.org · 来源:
分布式编译与图分割
📄 论文
PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallel
arxiv.org · 来源:
分布式编译与图分割
🌐 网站
PyTorch Distributed Overview
pytorch.org · 来源:
分布式编译与图分割
🌐 网站
XLA SPMD Partitioner
openxla.org · 来源:
分布式编译与图分割
🌐 网站
CUDA C++ Programming Guide — Streams
docs.nvidia.com · 来源:
调度与执行优化
🌐 网站
CUDA Graphs
docs.nvidia.com · 来源:
调度与执行优化
📄 论文
Checkmate: Breaking the Memory Wall with Optimal Tensor Rematerialization
arxiv.org · 来源:
调度与执行优化
📄 论文
Dynamic Tensor Rematerialization
arxiv.org · 来源:
调度与执行优化
🌐 网站
TorchInductor: A PyTorch Native Compiler
dev-discuss.pytorch.org · 来源:
调度与执行优化
🌐 网站
PyTorch Activation Checkpointing
pytorch.org · 来源:
调度与执行优化
📄 论文
Ansor: Generating High-Performance Tensor Programs for Deep Learning
arxiv.org · 来源:
自动调优与端到端实战
📄 论文
Learning to Optimize Tensor Programs
arxiv.org · 来源:
自动调优与端到端实战
🌐 网站
Triton Autotune Documentation
triton-lang.org · 来源:
自动调优与端到端实战
🌐 网站
torch.compile Troubleshooting
pytorch.org · 来源:
自动调优与端到端实战
强化学习:从基础到 LLM 对齐与推理 35 个资源
🌐 网站
Reinforcement Learning: An Introduction (Sutton & Barto, 2nd Edition)
incompleteideas.net · 来源:
强化学习基础:从 Agent 到 Bellman 方程
🌐 网站
David Silver UCL Reinforcement Learning Course
davidsilver.uk · 来源:
强化学习基础:从 Agent 到 Bellman 方程
🌐 网站
OpenAI Spinning Up: Introduction to RL
spinningup.openai.com · 来源:
强化学习基础:从 Agent 到 Bellman 方程
🌐 网站
Hugging Face Deep RL Course
huggingface.co · 来源:
强化学习基础:从 Agent 到 Bellman 方程 , Test-Time Scaling 与思维强化
🌐 网站
A (Long) Peek into Reinforcement Learning — Lilian Weng
lilianweng.github.io · 来源:
强化学习基础:从 Agent 到 Bellman 方程
🌐 网站
UC Berkeley CS285: Deep Reinforcement Learning
rail.eecs.berkeley.edu · 来源:
强化学习基础:从 Agent 到 Bellman 方程 , Policy Gradient:直接优化策略 , Actor-Critic 与 PPO:稳定的策略优化
🌐 网站
Deep Reinforcement Learning: Pong from Pixels — Andrej Karpathy
karpathy.github.io · 来源:
强化学习基础:从 Agent 到 Bellman 方程 , Test-Time Scaling 与思维强化
📄 论文
Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning (Williams, 1992)
link.springer.com · 来源:
Policy Gradient:直接优化策略
📄 论文
Policy Gradient Methods for Reinforcement Learning with Function Approximation (Sutton et al., 1999)
proceedings.neurips.cc · 来源:
Policy Gradient:直接优化策略
🌐 网站
Policy Gradient Algorithms — Lilian Weng
lilianweng.github.io · 来源:
Policy Gradient:直接优化策略 , Actor-Critic 与 PPO:稳定的策略优化 , 当 RL 遇上 LLM:从语言生成到策略优化
🌐 网站
OpenAI Spinning Up: Vanilla Policy Gradient
spinningup.openai.com · 来源:
Policy Gradient:直接优化策略
📄 论文
Proximal Policy Optimization Algorithms (Schulman et al., 2017)
arxiv.org · 来源:
Actor-Critic 与 PPO:稳定的策略优化
📄 论文
High-Dimensional Continuous Control Using Generalized Advantage Estimation (Schulman et al., 2016)
arxiv.org · 来源:
Actor-Critic 与 PPO:稳定的策略优化
📄 论文
Trust Region Policy Optimization (Schulman et al., 2015)
arxiv.org · 来源:
Actor-Critic 与 PPO:稳定的策略优化
🌐 网站
Hugging Face Deep RL Course: PPO
huggingface.co · 来源:
Actor-Critic 与 PPO:稳定的策略优化
📄 论文
Training language models to follow instructions with human feedback (Ouyang et al., 2022)
arxiv.org · 来源:
当 RL 遇上 LLM:从语言生成到策略优化 , RLHF:从人类反馈中学习
📄 论文
Direct Preference Optimization: Your Language Model is Secretly a Reward Model (Rafailov et al., 2023)
arxiv.org · 来源:
当 RL 遇上 LLM:从语言生成到策略优化 , 从 DPO 到 GRPO:直接偏好优化
📄 论文
DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models (Shao et al., 2024)
arxiv.org · 来源:
当 RL 遇上 LLM:从语言生成到策略优化 , 从 DPO 到 GRPO:直接偏好优化
📄 论文
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning (2025)
arxiv.org · 来源:
当 RL 遇上 LLM:从语言生成到策略优化 , Test-Time Scaling 与思维强化
📄 论文
A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning (Ross et al., 2011)
arxiv.org · 来源:
当 RL 遇上 LLM:从语言生成到策略优化
📄 论文
Fine-Tuning Language Models from Human Preferences (Ziegler et al., 2019)
arxiv.org · 来源:
当 RL 遇上 LLM:从语言生成到策略优化 , RLHF:从人类反馈中学习
📄 论文
Learning to summarize from human feedback (Stiennon et al., 2020)
arxiv.org · 来源:
当 RL 遇上 LLM:从语言生成到策略优化
🌐 网站
RLHF: Reinforcement Learning from Human Feedback — Chip Huyen
huyenchip.com · 来源:
当 RL 遇上 LLM:从语言生成到策略优化 , RLHF:从人类反馈中学习
📄 论文
Let's Verify Step by Step (Lightman et al., 2023)
arxiv.org · 来源:
当 RL 遇上 LLM:从语言生成到策略优化 , Reward 设计与 Scaling , Test-Time Scaling 与思维强化
📄 论文
Deep Reinforcement Learning from Human Preferences (Christiano et al., 2017)
arxiv.org · 来源:
RLHF:从人类反馈中学习
🌐 网站
RLHF 系列 — Nathan Lambert (interconnects.ai)
interconnects.ai · 来源:
RLHF:从人类反馈中学习 , Reward 设计与 Scaling
🌐 网站
Reward Hacking in Reinforcement Learning — Lilian Weng
lilianweng.github.io · 来源:
RLHF:从人类反馈中学习 , 从 DPO 到 GRPO:直接偏好优化 , Reward 设计与 Scaling
📄 论文
A General Theoretical Paradigm to Understand Learning from Human Feedback (Azar et al., 2023)
arxiv.org · 来源:
从 DPO 到 GRPO:直接偏好优化
📄 论文
KTO: Model Alignment as Prospect Theoretic Optimization (Ethayarajh et al., 2024)
arxiv.org · 来源:
从 DPO 到 GRPO:直接偏好优化
🌐 网站
Hugging Face TRL Documentation: DPO Trainer
huggingface.co · 来源:
从 DPO 到 GRPO:直接偏好优化
📄 论文
Training Verifiers to Solve Math Word Problems (Cobbe et al., 2021)
arxiv.org · 来源:
Reward 设计与 Scaling
📄 论文
Constitutional AI: Harmlessness from AI Feedback (Bai et al., 2022)
arxiv.org · 来源:
Reward 设计与 Scaling
📄 论文
Scaling Laws for Reward Model Overoptimization (Gao et al., 2022)
arxiv.org · 来源:
Reward 设计与 Scaling
📄 论文
Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters (Snell et al., 2024)
arxiv.org · 来源:
Test-Time Scaling 与思维强化
📄 论文
AlphaZero-like Tree-Search can Guide Large Language Model Decoding and Training (Feng et al., 2024)
arxiv.org · 来源:
Test-Time Scaling 与思维强化
🌐 网站
Intel Xe2 Architecture — Intel
intel.com · 来源:
Xe2 GPU 架构
🌐 网站
Intel Data Center GPU Max Series Architecture — Intel
intel.com · 来源:
Xe2 GPU 架构
🌐 网站
oneAPI GPU Optimization Guide — Intel
intel.com · 来源:
Xe2 GPU 架构
🌐 网站
oneAPI GPU Optimization Guide — Thread Hierarchy — Intel
intel.com · 来源:
Xe2 执行模型与编程抽象
🌐 网站
SYCL 2020 Specification — Khronos Group
registry.khronos.org · 来源:
Xe2 执行模型与编程抽象
🌐 网站
Intel GPU Occupancy Calculator — Intel
intel.com · 来源:
Xe2 执行模型与编程抽象
🌐 网站
DPC++ Language Extensions for SYCL — Intel
github.com · 来源:
Xe2 执行模型与编程抽象
🌐 网站
SPIR-V Specification — Khronos Group
registry.khronos.org · 来源:
SPIR-V 编译与 Level Zero 运行时
🌐 网站
oneAPI Level Zero Specification — Intel
spec.oneapi.io · 来源:
SPIR-V 编译与 Level Zero 运行时 , NPU 上的 LLM 推理:KV Cache 与软件栈 , NPU 执行模型与编程模型的边界
💻 代码
Intel Graphics Compiler (IGC) — GitHub
github.com · 来源:
SPIR-V 编译与 Level Zero 运行时
🌐 网站
SPIR-V Guide — Khronos
github.com · 来源:
SPIR-V 编译与 Level Zero 运行时
🌐 网站
oneDNN Developer Guide — Intel
oneapi-src.github.io · 来源:
oneDNN Primitive 体系
💻 代码
oneAPI Deep Neural Network Library (oneDNN) — GitHub
github.com · 来源:
oneDNN Primitive 体系
🌐 网站
oneDNN Programming Model — Intel
oneapi-src.github.io · 来源:
oneDNN Primitive 体系
🌐 网站
Memory Format Propagation — oneDNN
oneapi-src.github.io · 来源:
oneDNN Primitive 体系
🌐 网站
oneDNN Performance Profiling and Inspection — Intel
oneapi-src.github.io · 来源:
oneDNN GPU Kernel 优化
🌐 网站
oneAPI GPU Optimization Guide — GEMM — Intel
intel.com · 来源:
oneDNN GPU Kernel 优化
🌐 网站
XMX and XVE Architecture — Intel
intel.com · 来源:
oneDNN GPU Kernel 优化
🌐 网站
OpenVINO Architecture — Intel
docs.openvino.ai · 来源:
OpenVINO 图优化 Pipeline
🌐 网站
OpenVINO GPU Plugin — Intel
docs.openvino.ai · 来源:
OpenVINO 图优化 Pipeline
💻 代码
OpenVINO Toolkit — GitHub
github.com · 来源:
OpenVINO 图优化 Pipeline
🌐 网站
Optimum Intel Documentation
huggingface.co · 来源:
Intel 模型优化栈:Optimum Intel / NNCF / OpenVINO 三件套选型 , 动手:HF → GGUF / ONNX / OpenVINO 三条路径端到端
🌐 网站
NNCF GitHub Repository
github.com · 来源:
Intel 模型优化栈:Optimum Intel / NNCF / OpenVINO 三件套选型
🌐 网站
NNCF API Documentation
openvinotoolkit.github.io · 来源:
Intel 模型优化栈:Optimum Intel / NNCF / OpenVINO 三件套选型
🌐 网站
OpenVINO Model Conversion
docs.openvino.ai · 来源:
Intel 模型优化栈:Optimum Intel / NNCF / OpenVINO 三件套选型 , 动手:HF → GGUF / ONNX / OpenVINO 三条路径端到端
🌐 网站
Optimum Intel Source - Quantization
github.com · 来源:
Intel 模型优化栈:Optimum Intel / NNCF / OpenVINO 三件套选型
🌐 网站
Intel VTune Profiler — GPU Analysis — Intel
intel.com · 来源:
性能分析与瓶颈诊断
🌐 网站
OpenVINO Benchmark Tool — Intel
docs.openvino.ai · 来源:
性能分析与瓶颈诊断
🌐 网站
Intel GPU Top — intel_gpu_top man page
manpages.ubuntu.com · 来源:
性能分析与瓶颈诊断
🌐 网站
OpenVINO Multi-Device Execution — Intel
docs.openvino.ai · 来源:
NPU 架构与 GPU+NPU 协同推理
🌐 网站
OpenVINO AUTO Device — Intel
docs.openvino.ai · 来源:
NPU 架构与 GPU+NPU 协同推理
🌐 网站
Intel NPU Device — OpenVINO Documentation
docs.openvino.ai · 来源:
NPU 架构与 GPU+NPU 协同推理 , NPU 上的 LLM 推理:KV Cache 与软件栈
🌐 网站
Heterogeneous Execution — OpenVINO Docs
docs.openvino.ai · 来源:
NPU 架构与 GPU+NPU 协同推理
🌐 网站
OpenVINO GenAI — Stateful LLM Pipeline
docs.openvino.ai · 来源:
NPU 上的 LLM 推理:KV Cache 与软件栈
🌐 网站
openvinotoolkit/npu_compiler — GitHub
github.com · 来源:
NPU 上的 LLM 推理:KV Cache 与软件栈 , NPU 执行模型与编程模型的边界
🌐 网站
Flash Attention — Tri Dao et al.
arxiv.org · 来源:
NPU 执行模型与编程模型的边界
🌐 网站
CUTLASS 3.0 & CuTe — NVIDIA
github.com · 来源:
NPU 执行模型与编程模型的边界
🌐 网站
llama.cpp GitHub Repository
github.com · 来源:
动手:HF → GGUF / ONNX / OpenVINO 三条路径端到端
🌐 网站
ONNX Runtime Documentation
onnxruntime.ai · 来源:
动手:HF → GGUF / ONNX / OpenVINO 三条路径端到端
🌐 网站
lm-evaluation-harness
github.com · 来源:
动手:HF → GGUF / ONNX / OpenVINO 三条路径端到端
矩阵数学:从基础理论到现代 AI 架构 127 个资源
🌐 网站
MIT 18.06 Linear Algebra (Gilbert Strang)
ocw.mit.edu · 来源:
矩阵数学全景图:ML 的通用语言 , 核心性质速查:概念关系图与公式速查表 , 数据矩阵分解概述:问题、工具与方法谱系 , 向量空间的几何:内积、投影、秩与子空间 , 矩阵结构的几何:二次型、正定性与协方差 , 特征分解与对角化:万物之基 , 奇异值分解:核心中的核心 , 矩阵范数、内积与条件数:度量的艺术 , 矩阵微积分:从 Jacobian 到损失曲面 , 优化算法:从梯度下降到牛顿法 , PCA 与 Eigenfaces:从方差最大化到人脸识别 , 随机化 SVD:当精确分解算不动的时候 , NMF:非负约束下的 Parts-Based 分解 , Word2Vec 与 GloVe:隐式 vs 显式矩阵分解 , 算子矩阵全景:当矩阵不再装数据 , 马尔可夫链与转移矩阵:当矩阵编码概率 , 连续时间线性系统与 Kalman 滤波:从离散步进到平滑流动 , PageRank 与幂迭代:图上的马尔可夫链 , 随机游走与图嵌入:DeepWalk/Node2Vec , Kernel 矩阵与再生核:数据定义的给定算子 , 图 Laplacian 与谱聚类:从图结构到最优分割 , 图扩散、热核与 GNN 消息传递:从热方程到图神经网络
📄 论文
The Matrix Cookbook
math.uwaterloo.ca · 来源:
矩阵数学全景图:ML 的通用语言 , 核心性质速查:概念关系图与公式速查表 , 数据矩阵分解概述:问题、工具与方法谱系 , 向量空间的几何:内积、投影、秩与子空间 , 矩阵结构的几何:二次型、正定性与协方差 , 特征分解与对角化:万物之基 , 奇异值分解:核心中的核心 , 矩阵范数、内积与条件数:度量的艺术 , 矩阵微积分:从 Jacobian 到损失曲面 , PCA 与 Eigenfaces:从方差最大化到人脸识别 , 算子矩阵全景:当矩阵不再装数据 , 马尔可夫链与转移矩阵:当矩阵编码概率 , 隐马尔可夫模型:当状态看不见 , 连续时间线性系统与 Kalman 滤波:从离散步进到平滑流动
📄 论文
Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions
arxiv.org · 来源:
矩阵数学全景图:ML 的通用语言 , 数据矩阵分解概述:问题、工具与方法谱系 , 奇异值分解:核心中的核心 , 随机化 SVD:当精确分解算不动的时候
📄 论文
Exact Matrix Completion via Convex Optimization (Candès & Recht, 2009)
arxiv.org · 来源:
数据矩阵分解概述:问题、工具与方法谱系 , 矩阵范数、内积与条件数:度量的艺术 , 矩阵补全:从极少观测恢复低秩矩阵 , MF 与 FM:协同过滤的矩阵分解视角 , Robust PCA:低秩 + 稀疏分解
📄 论文
Robust Principal Component Analysis? (Candès, Li, Ma, Wright, 2011)
arxiv.org · 来源:
数据矩阵分解概述:问题、工具与方法谱系 , Robust PCA:低秩 + 稀疏分解
📄 论文
Efficient Estimation of Word Representations in Vector Space (Mikolov et al., 2013)
arxiv.org · 来源:
数据矩阵分解概述:问题、工具与方法谱系 , Word2Vec 与 GloVe:隐式 vs 显式矩阵分解
📄 论文
A Tutorial on Principal Component Analysis (Shlens, 2014)
arxiv.org · 来源:
数据矩阵分解概述:问题、工具与方法谱系 , 特征分解与对角化:万物之基 , PCA 与 Eigenfaces:从方差最大化到人脸识别 , Kernel 矩阵与再生核:数据定义的给定算子
🌐 网站
Inner product space — Wikipedia
en.wikipedia.org · 来源:
向量空间的几何:内积、投影、秩与子空间
🌐 网站
Projection (linear algebra) — Wikipedia
en.wikipedia.org · 来源:
向量空间的几何:内积、投影、秩与子空间
🌐 网站
Rank (linear algebra) — Wikipedia
en.wikipedia.org · 来源:
向量空间的几何:内积、投影、秩与子空间
🌐 网站
Kernel (linear algebra) — Wikipedia
en.wikipedia.org · 来源:
向量空间的几何:内积、投影、秩与子空间
🌐 网站
Orthogonal matrix — Wikipedia
en.wikipedia.org · 来源:
向量空间的几何:内积、投影、秩与子空间
🌐 网站
Essence of Linear Algebra — 3Blue1Brown
3blue1brown.com · 来源:
向量空间的几何:内积、投影、秩与子空间
🌐 网站
Quadratic form — Wikipedia
en.wikipedia.org · 来源:
矩阵结构的几何:二次型、正定性与协方差
🌐 网站
Definite matrix — Wikipedia
en.wikipedia.org · 来源:
矩阵结构的几何:二次型、正定性与协方差
🌐 网站
Covariance matrix — Wikipedia
en.wikipedia.org · 来源:
矩阵结构的几何:二次型、正定性与协方差
🌐 网站
Gramian matrix — Wikipedia
en.wikipedia.org · 来源:
矩阵结构的几何:二次型、正定性与协方差
🌐 网站
Determinant — Wikipedia
en.wikipedia.org · 来源:
矩阵结构的几何:二次型、正定性与协方差
🌐 网站
Interactive Linear Algebra — Eigenvalues and Eigenvectors (Georgia Tech)
textbooks.math.gatech.edu · 来源:
特征分解与对角化:万物之基
🌐 网站
Singular Value Decomposition — Wikipedia
en.wikipedia.org · 来源:
奇异值分解:核心中的核心
🌐 网站
Moore–Penrose Inverse — Wikipedia
en.wikipedia.org · 来源:
奇异值分解:核心中的核心
📄 论文
Implicit Regularization in Matrix Factorization (Gunasekar et al., 2017)
arxiv.org · 来源:
奇异值分解:核心中的核心
🌐 网站
Matrix norm — Wikipedia
en.wikipedia.org · 来源:
矩阵范数、内积与条件数:度量的艺术
🌐 网站
Condition number — Wikipedia
en.wikipedia.org · 来源:
矩阵范数、内积与条件数:度量的艺术
📄 论文
Spectral Normalization for Generative Adversarial Networks (Miyato et al., 2018)
arxiv.org · 来源:
矩阵范数、内积与条件数:度量的艺术
🌐 网站
What Is a Condition Number? (Nick Higham, 2020)
nhigham.com · 来源:
矩阵范数、内积与条件数:度量的艺术
🌐 网站
Matrix calculus — Wikipedia
en.wikipedia.org · 来源:
矩阵微积分:从 Jacobian 到损失曲面
🌐 网站
Jacobian matrix and determinant — Wikipedia
en.wikipedia.org · 来源:
矩阵微积分:从 Jacobian 到损失曲面
🌐 网站
Hessian matrix — Wikipedia
en.wikipedia.org · 来源:
矩阵微积分:从 Jacobian 到损失曲面
📄 论文
Measuring the Intrinsic Dimension of Objective Landscapes (Li et al., 2018)
arxiv.org · 来源:
矩阵微积分:从 Jacobian 到损失曲面 , 学习算子中的低秩结构:为什么神经网络权重是低秩的? , LoRA:低秩分解在 LLM 微调中的应用
🌐 网站
Gradient descent — Wikipedia
en.wikipedia.org · 来源:
优化算法:从梯度下降到牛顿法
🌐 网站
Newton's method in optimization — Wikipedia
en.wikipedia.org · 来源:
优化算法:从梯度下降到牛顿法
📄 论文
Introductory Lectures on Stochastic Optimization (Bottou, Curtis, Nocedal, 2018)
arxiv.org · 来源:
优化算法:从梯度下降到牛顿法
📄 论文
Adam: A Method for Stochastic Optimization (Kingma & Ba, 2015)
arxiv.org · 来源:
优化算法:从梯度下降到牛顿法
📄 论文
Numerical Optimization (Nocedal & Wright, 2006)
link.springer.com · 来源:
优化算法:从梯度下降到牛顿法
📄 论文
Eigenfaces for Recognition (Turk & Pentland, 1991)
face-rec.org · 来源:
PCA 与 Eigenfaces:从方差最大化到人脸识别
🌐 网站
Principal component analysis — Wikipedia
en.wikipedia.org · 来源:
PCA 与 Eigenfaces:从方差最大化到人脸识别
🌐 网站
Eigenface — Wikipedia
en.wikipedia.org · 来源:
PCA 与 Eigenfaces:从方差最大化到人脸识别
📄 论文
Lecture Notes on Randomized Linear Algebra (Mahoney, 2016)
arxiv.org · 来源:
随机化 SVD:当精确分解算不动的时候
🌐 网站
Johnson-Lindenstrauss lemma — Wikipedia
en.wikipedia.org · 来源:
随机化 SVD:当精确分解算不动的时候
🌐 网站
Random projection — Wikipedia
en.wikipedia.org · 来源:
随机化 SVD:当精确分解算不动的时候
📄 论文
The Power of Convex Relaxation: Near-Optimal Matrix Completion (Candès & Tao, 2010)
arxiv.org · 来源:
矩阵补全:从极少观测恢复低秩矩阵
📄 论文
A Simpler Approach to Matrix Completion (Recht, 2011)
arxiv.org · 来源:
矩阵补全:从极少观测恢复低秩矩阵
📄 论文
A Singular Value Thresholding Algorithm for Matrix Completion (Cai, Candès & Shen, 2010)
arxiv.org · 来源:
矩阵补全:从极少观测恢复低秩矩阵
📄 论文
Low-rank Matrix Completion using Alternating Minimization (Jain, Netrapalli & Sanghavi, 2013)
arxiv.org · 来源:
矩阵补全:从极少观测恢复低秩矩阵
📄 论文
Restricted Strong Convexity and Weighted Matrix Completion: Optimal Bounds with Noise (Negahban & Wainwright, 2012)
arxiv.org · 来源:
矩阵补全:从极少观测恢复低秩矩阵
🌐 网站
Netflix Prize — Wikipedia
en.wikipedia.org · 来源:
矩阵补全:从极少观测恢复低秩矩阵
🌐 网站
Matrix completion — Wikipedia
en.wikipedia.org · 来源:
矩阵补全:从极少观测恢复低秩矩阵
📄 论文
Learning the parts of objects by non-negative matrix factorization (Lee & Seung, 1999)
doi.org · 来源:
NMF:非负约束下的 Parts-Based 分解
📄 论文
Algorithms for Non-negative Matrix Factorization (Lee & Seung, 2000)
papers.nips.cc · 来源:
NMF:非负约束下的 Parts-Based 分解
📄 论文
The Why and How of Nonnegative Matrix Factorization (Gillis, 2014)
arxiv.org · 来源:
NMF:非负约束下的 Parts-Based 分解
🌐 网站
Non-negative matrix factorization — Wikipedia
en.wikipedia.org · 来源:
NMF:非负约束下的 Parts-Based 分解
📄 论文
Matrix Factorization Techniques for Recommender Systems (Koren, Bell, Volinsky, 2009)
doi.org · 来源:
MF 与 FM:协同过滤的矩阵分解视角
📄 论文
Factorization Machines (Rendle, 2010)
ismll.uni-hildesheim.de · 来源:
MF 与 FM:协同过滤的矩阵分解视角
📄 论文
Factorization Machines with libFM (Rendle, 2012)
dl.acm.org · 来源:
MF 与 FM:协同过滤的矩阵分解视角
🌐 网站
Matrix factorization (recommender systems) — Wikipedia
en.wikipedia.org · 来源:
MF 与 FM:协同过滤的矩阵分解视角
📄 论文
Neural Word Embedding as Implicit Matrix Factorization (Levy & Goldberg, 2014)
papers.nips.cc · 来源:
Word2Vec 与 GloVe:隐式 vs 显式矩阵分解 , 随机游走与图嵌入:DeepWalk/Node2Vec
📄 论文
Distributed Representations of Words and Phrases and their Compositionality (Mikolov et al., 2013)
arxiv.org · 来源:
Word2Vec 与 GloVe:隐式 vs 显式矩阵分解
📄 论文
GloVe: Global Vectors for Word Representation (Pennington, Socher & Manning, 2014)
nlp.stanford.edu · 来源:
Word2Vec 与 GloVe:隐式 vs 显式矩阵分解
📄 论文
Improving Distributional Similarity with Lessons Learned from Word Embeddings (Levy, Goldberg & Dagan, 2015)
aclanthology.org · 来源:
Word2Vec 与 GloVe:隐式 vs 显式矩阵分解
📄 论文
word2vec Explained: Deriving Mikolov et al.'s Negative-Sampling Word-Embedding Method (Goldberg & Levy, 2014)
arxiv.org · 来源:
Word2Vec 与 GloVe:隐式 vs 显式矩阵分解
📄 论文
The Augmented Lagrange Multiplier Method for Exact Recovery of Corrupted Low-Rank Matrices (Lin, Chen, Ma, 2010)
arxiv.org · 来源:
Robust PCA:低秩 + 稀疏分解
📄 论文
Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers (Boyd, Parikh, Chu, Peleato, Eckstein, 2011)
stanford.edu · 来源:
Robust PCA:低秩 + 稀疏分解
📄 论文
Stable Principal Component Pursuit (Zhou, Li, Wright, Candès, Ma, 2010)
arxiv.org · 来源:
Robust PCA:低秩 + 稀疏分解
🌐 网站
Robust principal component analysis — Wikipedia
en.wikipedia.org · 来源:
Robust PCA:低秩 + 稀疏分解
📄 论文
Tensor Decompositions and Applications (Kolda & Bader, 2009)
doi.org · 来源:
张量分解与知识图谱嵌入:从二维到高阶
📄 论文
Embedding Entities and Relations for Learning and Inference in Knowledge Bases (Yang et al., 2015) — DistMult
arxiv.org · 来源:
张量分解与知识图谱嵌入:从二维到高阶
📄 论文
Complex Embeddings for Simple Link Prediction (Trouillon et al., 2016) — ComplEx
arxiv.org · 来源:
张量分解与知识图谱嵌入:从二维到高阶
📄 论文
A Three-Way Model for Collective Learning on Multi-Relational Data (Nickel et al., 2011) — RESCAL
icml.cc · 来源:
张量分解与知识图谱嵌入:从二维到高阶
📄 论文
Analysis of individual differences in multidimensional scaling via an n-way generalization of 'Eckart-Young' decomposition (Tucker, 1966)
doi.org · 来源:
张量分解与知识图谱嵌入:从二维到高阶
🌐 网站
Tensor decomposition — Wikipedia
en.wikipedia.org · 来源:
张量分解与知识图谱嵌入:从二维到高阶
📄 论文
A Tutorial on Spectral Clustering (von Luxburg, 2007)
arxiv.org · 来源:
算子矩阵全景:当矩阵不再装数据 , 图 Laplacian 与谱聚类:从图结构到最优分割
📄 论文
An Intuitive Tutorial to Gaussian Process Regression (Wang, 2020)
arxiv.org · 来源:
算子矩阵全景:当矩阵不再装数据
📄 论文
Mamba: Linear-Time Sequence Modeling with Selective State Spaces (Gu & Dao, 2023)
arxiv.org · 来源:
算子矩阵全景:当矩阵不再装数据 , 连续时间线性系统与 Kalman 滤波:从离散步进到平滑流动 , 学习算子中的低秩结构:为什么神经网络权重是低秩的? , SSM / Mamba:矩阵对角化的胜利
📄 论文
A Comprehensive Survey on Graph Neural Networks (Wu et al., 2019)
arxiv.org · 来源:
算子矩阵全景:当矩阵不再装数据 , 图扩散、热核与 GNN 消息传递:从热方程到图神经网络
📄 论文
Markov Chains and Mixing Times (Levin, Peres & Wilmer, 2nd ed., AMS, 2017)
pages.uoregon.edu · 来源:
马尔可夫链与转移矩阵:当矩阵编码概率 , PageRank 与幂迭代:图上的马尔可夫链
🌐 网站
Perron–Frobenius Theorem — Wikipedia
en.wikipedia.org · 来源:
马尔可夫链与转移矩阵:当矩阵编码概率
🌐 网站
Markov Chain — Wikipedia
en.wikipedia.org · 来源:
马尔可夫链与转移矩阵:当矩阵编码概率 , 随机游走与图嵌入:DeepWalk/Node2Vec
📄 论文
An Introduction to MCMC for Machine Learning (Andrieu et al., 2003)
link.springer.com · 来源:
马尔可夫链与转移矩阵:当矩阵编码概率
📄 论文
A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition (Rabiner, 1989)
cs.ubc.ca · 来源:
隐马尔可夫模型:当状态看不见
📄 论文
An Introduction to Hidden Markov Models (Rabiner & Juang, IEEE ASSP Magazine, 1986)
ieeexplore.ieee.org · 来源:
隐马尔可夫模型:当状态看不见
📄 论文
Biological Sequence Analysis (Durbin, Eddy, Krogh & Mitchison, Cambridge University Press, 1998)
cambridge.org · 来源:
隐马尔可夫模型:当状态看不见
🌐 网站
Hidden Markov Model — Wikipedia
en.wikipedia.org · 来源:
隐马尔可夫模型:当状态看不见
🌐 网站
MIT 6.864 Advanced Natural Language Processing — HMM Lecture Notes
ocw.mit.edu · 来源:
隐马尔可夫模型:当状态看不见
📖 书籍
Feedback Systems: An Introduction for Scientists and Engineers (Åström & Murray, Princeton University Press, 2021)
cds.caltech.edu · 来源:
连续时间线性系统与 Kalman 滤波:从离散步进到平滑流动
📄 论文
A New Approach to Linear Filtering and Prediction Problems (Kalman, 1960)
cs.unc.edu · 来源:
连续时间线性系统与 Kalman 滤波:从离散步进到平滑流动
📄 论文
Efficiently Modeling Long Sequences with Structured State Spaces (Gu, Goel & Ré, ICLR 2022)
arxiv.org · 来源:
连续时间线性系统与 Kalman 滤波:从离散步进到平滑流动 , SSM / Mamba:矩阵对角化的胜利
🌐 网站
Kalman Filter — Wikipedia
en.wikipedia.org · 来源:
连续时间线性系统与 Kalman 滤波:从离散步进到平滑流动
📄 论文
The Anatomy of a Large-Scale Hypertextual Web Search Engine (Brin & Page, 1998)
infolab.stanford.edu · 来源:
PageRank 与幂迭代:图上的马尔可夫链
📄 论文
The PageRank Citation Ranking: Bringing Order to the Web (Page et al., 1999)
ilpubs.stanford.edu · 来源:
PageRank 与幂迭代:图上的马尔可夫链
🌐 网站
PageRank — Wikipedia
en.wikipedia.org · 来源:
PageRank 与幂迭代:图上的马尔可夫链
📄 论文
Deeper Inside PageRank (Langville & Meyer, 2004)
doi.org · 来源:
PageRank 与幂迭代:图上的马尔可夫链
📄 论文
DeepWalk: Online Learning of Social Representations (Perozzi, Al-Rfou & Skiena, 2014)
arxiv.org · 来源:
随机游走与图嵌入:DeepWalk/Node2Vec
📄 论文
node2vec: Scalable Feature Learning for Networks (Grover & Leskovec, 2016)
arxiv.org · 来源:
随机游走与图嵌入:DeepWalk/Node2Vec
📄 论文
Network Embedding as Matrix Factorization: Approximating Graph Context with Polynomial Functions (Qiu et al., 2018)
arxiv.org · 来源:
随机游走与图嵌入:DeepWalk/Node2Vec
📄 论文
DeepWalk is equivalent to computing a specific matrix factorization (Yang & Qiu, 2015, see Qiu et al. 2018 for unified treatment)
dl.acm.org · 来源:
随机游走与图嵌入:DeepWalk/Node2Vec
🌐 网站
Learning with Kernels (Schölkopf & Smola, 2002)
en.wikipedia.org · 来源:
Kernel 矩阵与再生核:数据定义的给定算子
📖 书籍
Gaussian Processes for Machine Learning (Rasmussen & Williams, 2006)
gaussianprocess.org · 来源:
Kernel 矩阵与再生核:数据定义的给定算子
🌐 网站
Mercer's theorem — Wikipedia
en.wikipedia.org · 来源:
Kernel 矩阵与再生核:数据定义的给定算子
🌐 网站
Kernel Principal Component Analysis (Schölkopf, Smola & Müller, 1998)
en.wikipedia.org · 来源:
Kernel 矩阵与再生核:数据定义的给定算子
🌐 网站
Reproducing kernel Hilbert space — Wikipedia
en.wikipedia.org · 来源:
Kernel 矩阵与再生核:数据定义的给定算子
📄 论文
Algebraic connectivity of graphs (Fiedler, 1973)
doi.org · 来源:
图 Laplacian 与谱聚类:从图结构到最优分割
📄 论文
Normalized Cuts and Image Segmentation (Shi & Malik, 2000)
people.eecs.berkeley.edu · 来源:
图 Laplacian 与谱聚类:从图结构到最优分割
📄 论文
On Spectral Clustering: Analysis and an Algorithm (Ng, Jordan & Weiss, 2001)
papers.nips.cc · 来源:
图 Laplacian 与谱聚类:从图结构到最优分割
📄 论文
Using the Nyström Method to Speed Up Kernel Machines (Williams & Seeger, 2001)
proceedings.neurips.cc · 来源:
图 Laplacian 与谱聚类:从图结构到最优分割
📄 论文
Semi-Supervised Classification with Graph Convolutional Networks (Kipf & Welling, ICLR 2017)
arxiv.org · 来源:
图扩散、热核与 GNN 消息传递:从热方程到图神经网络
📄 论文
Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering (Defferrard, Bresson & Vandergheynst, NeurIPS 2016)
arxiv.org · 来源:
图扩散、热核与 GNN 消息传递:从热方程到图神经网络
📄 论文
Diffusion Kernels on Graphs and Other Discrete Structures (Kondor & Lafferty, ICML 2002)
cs.cmu.edu · 来源:
图扩散、热核与 GNN 消息传递:从热方程到图神经网络
📄 论文
The Emerging Field of Signal Processing on Graphs (Shuman, Narang, Frossard, Ortega & Vandergheynst, 2013)
arxiv.org · 来源:
图扩散、热核与 GNN 消息传递:从热方程到图神经网络
📄 论文
Denoising Diffusion Probabilistic Models (Ho, Jain & Abbeel, NeurIPS 2020)
arxiv.org · 来源:
图扩散、热核与 GNN 消息传递:从热方程到图神经网络
🌐 网站
Heat Equation — Wikipedia
en.wikipedia.org · 来源:
图扩散、热核与 GNN 消息传递:从热方程到图神经网络
📄 论文
Intrinsic Dimensionality Explains the Effectiveness of Language Model Fine-Tuning (Aghajanyan et al., ACL 2021)
arxiv.org · 来源:
学习算子中的低秩结构:为什么神经网络权重是低秩的? , LoRA:低秩分解在 LLM 微调中的应用
📄 论文
LoRA: Low-Rank Adaptation of Large Language Models (Hu et al., ICLR 2022)
arxiv.org · 来源:
学习算子中的低秩结构:为什么神经网络权重是低秩的? , LoRA:低秩分解在 LLM 微调中的应用
📄 论文
Traditional and Heavy-Tailed Self Regularization in Neural Network Models (Martin & Mahoney, 2019)
arxiv.org · 来源:
学习算子中的低秩结构:为什么神经网络权重是低秩的? , Attention 的低秩结构与 Efficient Attention
📄 论文
Gradient Descent Happens in a Tiny Subspace (Gur-Ari et al., 2018)
arxiv.org · 来源:
学习算子中的低秩结构:为什么神经网络权重是低秩的?
📄 论文
Rethinking Attention with Performers (Choromanski et al., ICLR 2021)
arxiv.org · 来源:
学习算子中的低秩结构:为什么神经网络权重是低秩的? , Attention 的低秩结构与 Efficient Attention
📄 论文
QLoRA: Efficient Finetuning of Quantized LLMs (Dettmers et al., NeurIPS 2023)
arxiv.org · 来源:
LoRA:低秩分解在 LLM 微调中的应用
📄 论文
LoRA+: Efficient Low Rank Adaptation of Large Models (Hayou et al., ICML 2024)
arxiv.org · 来源:
LoRA:低秩分解在 LLM 微调中的应用
📄 论文
A Rank Stabilization Scaling Factor for Fine-Tuning with LoRA (Kalajdzievski, 2023)
arxiv.org · 来源:
LoRA:低秩分解在 LLM 微调中的应用
📄 论文
LQ-LoRA: Low-Rank Plus Quantized Matrix Decomposition for Efficient Language Model Finetuning (Guo et al., ICLR 2024)
arxiv.org · 来源:
LoRA:低秩分解在 LLM 微调中的应用
📄 论文
Attention Is All You Need (Vaswani et al., NeurIPS 2017)
arxiv.org · 来源:
Attention 的低秩结构与 Efficient Attention
📄 论文
Linformer: Self-Attention with Linear Complexity (Wang et al., 2020)
arxiv.org · 来源:
Attention 的低秩结构与 Efficient Attention
📄 论文
A Unified Taxonomy and Evaluation of Efficient Transformers (Tay et al., 2022)
arxiv.org · 来源:
Attention 的低秩结构与 Efficient Attention
📄 论文
HiPPO: Recurrent Memory with Optimal Polynomial Projections (Gu et al., NeurIPS 2020)
arxiv.org · 来源:
SSM / Mamba:矩阵对角化的胜利
📄 论文
On the Parameterization and Initialization of Diagonal State Space Models (Gu et al., NeurIPS 2022)
arxiv.org · 来源:
SSM / Mamba:矩阵对角化的胜利
📄 论文
It's Raw! Audio Generation with State-Space Models (Goel et al., ICML 2022)
arxiv.org · 来源:
SSM / Mamba:矩阵对角化的胜利
📄 论文
How to Train Your HiPPO: State Space Models with Generalized Orthogonal Basis Projections (Gu et al., ICLR 2023)
arxiv.org · 来源:
SSM / Mamba:矩阵对角化的胜利
图算法:从结构探索到组合优化 162 个资源
🌐 网站
Introduction to Algorithms (CLRS), 4th Edition — Part VI: Graph Algorithms
mitpress.mit.edu · 来源:
图算法全景图:从结构探索到组合优化 , BFS 与 DFS:图的两种基本呼吸方式 , 连通性:图能拆成几块? , 拓扑排序与 DAG:有依赖时的合法顺序 , 欧拉与哈密顿:遍历的两种完备性 , 树上算法:图的特殊骨架 , 最短路径:图上的距离 , 中心性:谁最重要? , 社区发现:哪些节点抱团? , 团与密子图:最紧密的子群 , 最小生成树:最便宜地连通所有人 , 网络流:管道能通多少? , 匹配:最优配对 , 着色与划分:最少几种颜色? , NP-hard 与近似算法:当最优解算不出来 , 图建模案例集:这个问题其实是图问题
🌐 网站
Graph Theory (Reinhard Diestel), 5th Edition — Free online version
diestel-graph-theory.com · 来源:
图算法全景图:从结构探索到组合优化 , BFS 与 DFS:图的两种基本呼吸方式 , 连通性:图能拆成几块? , 拓扑排序与 DAG:有依赖时的合法顺序 , 欧拉与哈密顿:遍历的两种完备性 , 树上算法:图的特殊骨架 , 团与密子图:最紧密的子群 , 着色与划分:最少几种颜色?
🌐 网站
Network Science (Albert-László Barabási) — Free online textbook
networksciencebook.com · 来源:
图算法全景图:从结构探索到组合优化 , 随机图与网络模型:真实网络长什么样?
🌐 网站
NetworkX Documentation
networkx.org · 来源:
图算法全景图:从结构探索到组合优化
🌐 网站
Jeff Erickson — Algorithms (2019), Chapter 5: Whatever-First Search
jeffe.cs.illinois.edu · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
🌐 网站
Sedgewick, R. — Algorithms in C, Part 5: Graph Algorithms (1990/2003)
cs.princeton.edu · 来源:
图上的通用迭代机器(上):从数学问题到求解框架
🌐 网站
Sedgewick, R. & Wayne, K. — Algorithms, 4th Edition (2011)
algs4.cs.princeton.edu · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
📄 论文
Pingali, K. et al. — The Tao of Parallelism in Algorithms (PLDI 2011)
doi.org · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
📄 论文
Kildall, G. — A Unified Approach to Global Program Optimization (POPL 1973)
doi.org · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
📄 论文
Cousot, P. & Cousot, R. — Abstract Interpretation: A Unified Lattice Model (POPL 1977)
doi.org · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
📄 论文
Dijkstra, E.W. et al. — On-the-fly Garbage Collection: An Exercise in Cooperation (CACM 1978)
doi.org · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
📄 论文
Dijkstra, E.W. — A note on two problems in connexion with graphs (1959)
doi.org · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
🌐 网站
Pearl, J. — Probabilistic Reasoning in Intelligent Systems (1988)
dl.acm.org · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
📄 论文
Tarski, A. — A lattice-theoretical fixpoint theorem and its applications (1955)
doi.org · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
📄 论文
Mohri, M. — Semiring Frameworks and Algorithms for Shortest-Distance Problems (2002)
doi.org · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
🌐 网站
Russell, S. & Norvig, P. — Artificial Intelligence: A Modern Approach, 4th Edition (2020)
aima.cs.berkeley.edu · 来源:
图上的通用迭代机器(上):从数学问题到求解框架 , 图上的通用迭代机器(下):范式、领域与边界
📄 论文
Xu, K. et al. — How Powerful are Graph Neural Networks? (ICLR 2019)
arxiv.org · 来源:
图上的通用迭代机器(下):范式、领域与边界 , 图嵌入与图神经网络:把图变成向量
📄 论文
Gondran, M. & Minoux, M. — Graphs, Dioids and Semirings (2008)
link.springer.com · 来源:
图上的通用迭代机器(下):范式、领域与边界
📄 论文
Malewicz, G. et al. — Pregel: A System for Large-Scale Graph Processing (SIGMOD 2010)
doi.org · 来源:
图上的通用迭代机器(下):范式、领域与边界
📄 论文
Kipf, T.N. & Welling, M. — Semi-Supervised Classification with Graph Convolutional Networks (ICLR 2017)
arxiv.org · 来源:
图上的通用迭代机器(下):范式、领域与边界 , 图嵌入与图神经网络:把图变成向量
🌐 网站
Algorithms, 4th Edition (Sedgewick & Wayne) — Section 4.1-4.2: Undirected/Directed Graphs
algs4.cs.princeton.edu · 来源:
BFS 与 DFS:图的两种基本呼吸方式 , 树上算法:图的特殊骨架
🌐 网站
BFS and DFS — Wikipedia
en.wikipedia.org · 来源:
BFS 与 DFS:图的两种基本呼吸方式
🌐 网站
NetworkX Documentation — Traversal
networkx.org · 来源:
BFS 与 DFS:图的两种基本呼吸方式
🌐 网站
Algorithms, 4th Edition (Sedgewick & Wayne) — Section 4.2: Directed Graphs
algs4.cs.princeton.edu · 来源:
连通性:图能拆成几块? , 拓扑排序与 DAG:有依赖时的合法顺序
🌐 网站
Robert Tarjan — Depth-First Search and Linear Graph Algorithms (1972)
doi.org · 来源:
连通性:图能拆成几块?
🌐 网站
NetworkX Documentation — Components
networkx.org · 来源:
连通性:图能拆成几块?
🌐 网站
2-SAT — CP-Algorithms
cp-algorithms.com · 来源:
连通性:图能拆成几块?
🌐 网站
Kahn, Arthur B. (1962) — Topological sorting of large networks, Communications of the ACM
dl.acm.org · 来源:
拓扑排序与 DAG:有依赖时的合法顺序
🌐 网站
Topological sorting — Wikipedia
en.wikipedia.org · 来源:
拓扑排序与 DAG:有依赖时的合法顺序
🌐 网站
NetworkX Documentation — DAG algorithms
networkx.org · 来源:
拓扑排序与 DAG:有依赖时的合法顺序
🌐 网站
Critical Path Method — Wikipedia
en.wikipedia.org · 来源:
拓扑排序与 DAG:有依赖时的合法顺序
🌐 网站
Euler, Leonhard (1736) — Solutio problematis ad geometriam situs pertinentis
scholarlycommons.pacific.edu · 来源:
欧拉与哈密顿:遍历的两种完备性
🌐 网站
Hierholzer, Carl (1873) — Ueber die Möglichkeit, einen Linienzug ohne Wiederholung und ohne Unterbrechung zu umfahren
en.wikipedia.org · 来源:
欧拉与哈密顿:遍历的两种完备性
🌐 网站
Hamiltonian path problem — Wikipedia
en.wikipedia.org · 来源:
欧拉与哈密顿:遍历的两种完备性
🌐 网站
Karp, Richard M. (1972) — Reducibility Among Combinatorial Problems
link.springer.com · 来源:
欧拉与哈密顿:遍历的两种完备性
🌐 网站
Pevzner, P.A. et al. (2001) — An Eulerian path approach to DNA fragment assembly
doi.org · 来源:
欧拉与哈密顿:遍历的两种完备性
🌐 网站
Chinese Postman Problem — Wikipedia
en.wikipedia.org · 来源:
欧拉与哈密顿:遍历的两种完备性
🌐 网站
Travelling salesman problem — Wikipedia
en.wikipedia.org · 来源:
欧拉与哈密顿:遍历的两种完备性 , NP-hard 与近似算法:当最优解算不出来
🌐 网站
NetworkX Documentation — Eulerian circuits
networkx.org · 来源:
欧拉与哈密顿:遍历的两种完备性
🌐 网站
Bender & Farach-Colton (2000) — The LCA Problem Revisited
ics.uci.edu · 来源:
树上算法:图的特殊骨架
🌐 网站
Tarjan, Robert E. (1979) — Applications of Path Compression on Balanced Trees, Journal of the ACM
dl.acm.org · 来源:
树上算法:图的特殊骨架
🌐 网站
Lengauer, Thomas; Tarjan, Robert E. (1979) — A Fast Algorithm for Finding Dominators in a Flowgraph
dl.acm.org · 来源:
树上算法:图的特殊骨架
🌐 网站
Lowest common ancestor — Wikipedia
en.wikipedia.org · 来源:
树上算法:图的特殊骨架
🌐 网站
Heavy path decomposition — Wikipedia
en.wikipedia.org · 来源:
树上算法:图的特殊骨架
🌐 网站
Dominator (graph theory) — Wikipedia
en.wikipedia.org · 来源:
树上算法:图的特殊骨架
🌐 网站
NetworkX Documentation — Tree algorithms
networkx.org · 来源:
树上算法:图的特殊骨架 , 最小生成树:最便宜地连通所有人
🌐 网站
Algorithms, 4th Edition (Sedgewick & Wayne) — Section 4.4: Shortest Paths
algs4.cs.princeton.edu · 来源:
最短路径:图上的距离
🌐 网站
Dijkstra, E. W. (1959). A note on two problems in connexion with graphs. Numerische Mathematik, 1(1), 269-271
link.springer.com · 来源:
最短路径:图上的距离
🌐 网站
Bellman, Richard (1958). On a routing problem. Quarterly of Applied Mathematics, 16(1), 87-90
ams.org · 来源:
最短路径:图上的距离
🌐 网站
Floyd, Robert W. (1962). Algorithm 97: Shortest path. Communications of the ACM, 5(6), 345
dl.acm.org · 来源:
最短路径:图上的距离
🌐 网站
Hart, P. E., Nilsson, N. J., & Raphael, B. (1968). A formal basis for the heuristic determination of minimum cost paths. IEEE Transactions on SSC, 4(2), 100-107
ieeexplore.ieee.org · 来源:
最短路径:图上的距离
🌐 网站
Dijkstra's algorithm — Wikipedia
en.wikipedia.org · 来源:
最短路径:图上的距离
🌐 网站
Bellman-Ford algorithm — Wikipedia
en.wikipedia.org · 来源:
最短路径:图上的距离
🌐 网站
Floyd-Warshall algorithm — Wikipedia
en.wikipedia.org · 来源:
最短路径:图上的距离
🌐 网站
A* search algorithm — Wikipedia
en.wikipedia.org · 来源:
最短路径:图上的距离
🌐 网站
NetworkX Documentation — Shortest Paths
networkx.org · 来源:
最短路径:图上的距离
🌐 网站
Brandes, Ulrik (2001). A Faster Algorithm for Betweenness Centrality. Journal of Mathematical Sociology, 25(2), 163-177
doi.org · 来源:
中心性:谁最重要?
🌐 网站
Page, Lawrence; Brin, Sergey; Motwani, Rajeev; Winograd, Terry (1999). The PageRank Citation Ranking: Bringing Order to the Web. Stanford InfoLab Technical Report
ilpubs.stanford.edu · 来源:
中心性:谁最重要?
🌐 网站
Freeman, Linton C. (1977). A set of measures of centrality based on betweenness. Sociometry, 40(1), 35-41
doi.org · 来源:
中心性:谁最重要?
🌐 网站
Katz, Leo (1953). A new status index derived from sociometric analysis. Psychometrika, 18(1), 39-43
doi.org · 来源:
中心性:谁最重要?
🌐 网站
Bonacich, Phillip (1987). Power and centrality: A family of measures. American Journal of Sociology, 92(5), 1170-1182
doi.org · 来源:
中心性:谁最重要?
🌐 网站
Centrality — Wikipedia
en.wikipedia.org · 来源:
中心性:谁最重要?
🌐 网站
NetworkX Documentation — Centrality
networkx.org · 来源:
中心性:谁最重要?
🌐 网站
Neo4j Graph Data Science — Centrality Algorithms
neo4j.com · 来源:
中心性:谁最重要?
📄 论文
Fast unfolding of communities in large networks (Blondel et al., 2008) — Louvain algorithm
arxiv.org · 来源:
社区发现:哪些节点抱团?
📄 论文
Near linear time algorithm to detect community structures in large-scale networks (Raghavan et al., 2007) — Label Propagation
arxiv.org · 来源:
社区发现:哪些节点抱团?
📄 论文
Modularity and community structure in networks (Newman, 2006)
arxiv.org · 来源:
社区发现:哪些节点抱团?
📄 论文
Resolution limit in community detection (Fortunato & Barthélemy, 2007)
doi.org · 来源:
社区发现:哪些节点抱团?
📄 论文
From Louvain to Leiden: guaranteeing well-connected communities (Traag et al., 2019)
arxiv.org · 来源:
社区发现:哪些节点抱团?
🌐 网站
NetworkX Community Detection Documentation
networkx.org · 来源:
社区发现:哪些节点抱团?
🌐 网站
Network Science (Barabási) — Chapter 9: Communities
networksciencebook.com · 来源:
社区发现:哪些节点抱团?
📄 论文
Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks (Morris et al., 2019)
arxiv.org · 来源:
相似性与同构:两个图/节点有多像? , 图嵌入与图神经网络:把图变成向量
📄 论文
Graph Isomorphism in Quasipolynomial Time (Babai, 2016)
arxiv.org · 来源:
相似性与同构:两个图/节点有多像?
📄 论文
SimRank: A Measure of Structural-Context Similarity (Jeh & Widom, 2002)
dl.acm.org · 来源:
相似性与同构:两个图/节点有多像?
📄 论文
Weisfeiler-Lehman Graph Kernels (Shervashidze et al., 2011)
jmlr.org · 来源:
相似性与同构:两个图/节点有多像?
📄 论文
An Improved Algorithm for Matching Large Graphs — VF2 (Cordella et al., 2004)
ieeexplore.ieee.org · 来源:
相似性与同构:两个图/节点有多像?
🌐 网站
NetworkX — Graph Isomorphism
networkx.org · 来源:
相似性与同构:两个图/节点有多像?
🌐 网站
GraKeL — Graph Kernels Library
ysig.github.io · 来源:
相似性与同构:两个图/节点有多像?
🌐 网站
Algorithm Design (Kleinberg & Tardos) — Chapter 13: Randomized Algorithms
cs.cornell.edu · 来源:
团与密子图:最紧密的子群 , 匹配:最优配对 , NP-hard 与近似算法:当最优解算不出来 , 图建模案例集:这个问题其实是图问题
📄 论文
Bron, Kerbosch — Algorithm 457: Finding All Cliques of an Undirected Graph (1973)
doi.org · 来源:
团与密子图:最紧密的子群
📄 论文
Eppstein, Löffler, Strash — Listing All Maximal Cliques in Large Sparse Real-World Graphs (2013)
doi.org · 来源:
团与密子图:最紧密的子群
📄 论文
Batagelj, Zaversnik — An O(m) Algorithm for Cores Decomposition of Networks (2003)
arxiv.org · 来源:
团与密子图:最紧密的子群
📄 论文
Cohen — Trusses: Cohesive Subgraphs for Social Network Analysis (2008)
doi.org · 来源:
团与密子图:最紧密的子群
🌐 网站
NetworkX Documentation — Cliques
networkx.org · 来源:
团与密子图:最紧密的子群
🌐 网站
Algorithms, 4th Edition (Sedgewick & Wayne) — Section 4.3: Minimum Spanning Trees
algs4.cs.princeton.edu · 来源:
最小生成树:最便宜地连通所有人
🌐 网站
Kruskal, Joseph B. (1956). On the Shortest Spanning Subtree of a Graph. Proceedings of the AMS, 7(1), 48-50
doi.org · 来源:
最小生成树:最便宜地连通所有人
🌐 网站
Prim, R. C. (1957). Shortest Connection Networks And Some Generalizations. Bell System Technical Journal, 36(6), 1389-1401
doi.org · 来源:
最小生成树:最便宜地连通所有人
🌐 网站
Borůvka, Otakar (1926). O jistém problému minimálním. Práce Moravské Přírodovědecké Společnosti, 3, 37-58
en.wikipedia.org · 来源:
最小生成树:最便宜地连通所有人
🌐 网站
Tarjan, Robert E. (1975). Efficiency of a Good But Not Linear Set Union Algorithm. JACM, 22(2), 215-225
doi.org · 来源:
最小生成树:最便宜地连通所有人
🌐 网站
Minimum spanning tree — Wikipedia
en.wikipedia.org · 来源:
最小生成树:最便宜地连通所有人
🌐 网站
Matroid — Wikipedia
en.wikipedia.org · 来源:
最小生成树:最便宜地连通所有人
🌐 网站
Christofides, Nicos (1976). Worst-Case Analysis of a New Heuristic for the Travelling Salesman Problem. Technical Report
en.wikipedia.org · 来源:
最小生成树:最便宜地连通所有人
🌐 网站
Steiner tree problem — Wikipedia
en.wikipedia.org · 来源:
最小生成树:最便宜地连通所有人
🌐 网站
Ford, L. R. & Fulkerson, D. R. (1956). Maximal flow through a network. Canadian Journal of Mathematics, 8, 399-404
doi.org · 来源:
网络流:管道能通多少?
🌐 网站
Edmonds, J. & Karp, R. M. (1972). Theoretical improvements in algorithmic efficiency for network flow problems. Journal of the ACM, 19(2), 248-264
dl.acm.org · 来源:
网络流:管道能通多少?
🌐 网站
Dinic, E. A. (1970). Algorithm for solution of a problem of maximum flow in networks with power estimation. Soviet Mathematics Doklady, 11, 1277-1280
en.wikipedia.org · 来源:
网络流:管道能通多少?
🌐 网站
Maximum flow problem — Wikipedia
en.wikipedia.org · 来源:
网络流:管道能通多少?
🌐 网站
Max-flow min-cut theorem — Wikipedia
en.wikipedia.org · 来源:
网络流:管道能通多少?
🌐 网站
NetworkX Documentation — Flow Algorithms
networkx.org · 来源:
网络流:管道能通多少?
🌐 网站
Boykov, Y. & Kolmogorov, V. (2004). An experimental comparison of min-cut/max-flow algorithms for energy minimization in vision. IEEE TPAMI, 26(9), 1124-1137
ieeexplore.ieee.org · 来源:
网络流:管道能通多少?
📄 论文
Hopcroft, J. E. & Karp, R. M. (1973). An n^(5/2) Algorithm for Maximum Matchings in Bipartite Graphs. SIAM J. Comput., 2(4), 225-231
doi.org · 来源:
匹配:最优配对
📄 论文
Edmonds, J. (1965). Paths, Trees, and Flowers. Canadian Journal of Mathematics, 17, 449-467
doi.org · 来源:
匹配:最优配对
📄 论文
Kuhn, H. W. (1955). The Hungarian Method for the Assignment Problem. Naval Research Logistics Quarterly, 2(1-2), 83-97
doi.org · 来源:
匹配:最优配对
📄 论文
Hall, P. (1935). On Representatives of Subsets. Journal of the London Mathematical Society, s1-10(1), 26-30
doi.org · 来源:
匹配:最优配对
📄 论文
König, D. (1931). Gráfok és mátrixok. Matematikai és Fizikai Lapok, 38, 116-119
en.wikipedia.org · 来源:
匹配:最优配对
📄 论文
Gale, D. & Shapley, L. S. (1962). College Admissions and the Stability of Marriage. The American Mathematical Monthly, 69(1), 9-15
doi.org · 来源:
匹配:最优配对
🌐 网站
Matching (graph theory) — Wikipedia
en.wikipedia.org · 来源:
匹配:最优配对
🌐 网站
NetworkX Documentation — Matching Algorithms
networkx.org · 来源:
匹配:最优配对
📄 论文
Brélaz — New Methods to Color the Vertices of a Graph (1979)
doi.org · 来源:
着色与划分:最少几种颜色?
📄 论文
Brooks — On Colouring the Nodes of a Network (1941)
doi.org · 来源:
着色与划分:最少几种颜色?
📄 论文
Appel, Haken — Every Planar Map is Four Colorable (1976)
doi.org · 来源:
着色与划分:最少几种颜色?
📄 论文
Karypis, Kumar — A Fast and High Quality Multilevel Scheme for Partitioning Irregular Graphs (1998)
doi.org · 来源:
着色与划分:最少几种颜色? , 图建模案例集:这个问题其实是图问题
🌐 网站
Kuratowski's Theorem — Wikipedia
en.wikipedia.org · 来源:
着色与划分:最少几种颜色?
🌐 网站
NetworkX Documentation — Coloring
networkx.org · 来源:
着色与划分:最少几种颜色?
🌐 网站
METIS — Serial Graph Partitioning and Fill-reducing Matrix Ordering
github.com · 来源:
着色与划分:最少几种颜色?
📄 论文
Christofides, N. (1976). Worst-case analysis of a new heuristic for the travelling salesman problem. Technical Report 388, Graduate School of Industrial Administration, CMU
doi.org · 来源:
NP-hard 与近似算法:当最优解算不出来
📄 论文
Karlin, Klein, Oveis Gharan (2021). A (Slightly) Improved Approximation Algorithm for Metric TSP. STOC 2021
doi.org · 来源:
NP-hard 与近似算法:当最优解算不出来
📄 论文
Vazirani, V. V. (2001). Approximation Algorithms. Springer
link.springer.com · 来源:
NP-hard 与近似算法:当最优解算不出来
📄 论文
Downey, R. G. & Fellows, M. R. (1999). Parameterized Complexity. Springer
link.springer.com · 来源:
NP-hard 与近似算法:当最优解算不出来
📄 论文
Courcelle, B. (1990). The Monadic Second-Order Logic of Graphs. I. Recognizable Sets of Finite Graphs. Information and Computation, 85(1), 12-75
doi.org · 来源:
NP-hard 与近似算法:当最优解算不出来
🌐 网站
Approximation algorithm — Wikipedia
en.wikipedia.org · 来源:
NP-hard 与近似算法:当最优解算不出来
🌐 网站
Fixed-parameter tractability — Wikipedia
en.wikipedia.org · 来源:
NP-hard 与近似算法:当最优解算不出来
🌐 网站
Treewidth — Wikipedia
en.wikipedia.org · 来源:
NP-hard 与近似算法:当最优解算不出来
🌐 网站
NetworkX Documentation — Approximation Algorithms
networkx.org · 来源:
NP-hard 与近似算法:当最优解算不出来
🌐 网站
Concorde TSP Solver
math.uwaterloo.ca · 来源:
NP-hard 与近似算法:当最优解算不出来
🌐 网站
Google OR-Tools — Vehicle Routing
developers.google.com · 来源:
NP-hard 与近似算法:当最优解算不出来
📄 论文
Erdos, P. & Renyi, A. (1959). On Random Graphs I. Publicationes Mathematicae, 6, 290-297
renyi.hu · 来源:
随机图与网络模型:真实网络长什么样?
📄 论文
Erdos, P. & Renyi, A. (1960). On the Evolution of Random Graphs. Publication of the Mathematical Institute of the Hungarian Academy of Sciences, 5, 17-61
renyi.hu · 来源:
随机图与网络模型:真实网络长什么样?
📄 论文
Watts, D. J. & Strogatz, S. H. (1998). Collective dynamics of 'small-world' networks. Nature, 393, 440-442
doi.org · 来源:
随机图与网络模型:真实网络长什么样?
📄 论文
Barabasi, A.-L. & Albert, R. (1999). Emergence of Scaling in Random Networks. Science, 286, 509-512
doi.org · 来源:
随机图与网络模型:真实网络长什么样?
📄 论文
Albert, R., Jeong, H. & Barabasi, A.-L. (2000). Error and attack tolerance of complex networks. Nature, 406, 378-382
doi.org · 来源:
随机图与网络模型:真实网络长什么样?
📄 论文
Newman, M. E. J. (2003). The Structure and Function of Complex Networks. SIAM Review, 45(2), 167-256
doi.org · 来源:
随机图与网络模型:真实网络长什么样?
🌐 网站
Random graph — Wikipedia
en.wikipedia.org · 来源:
随机图与网络模型:真实网络长什么样?
🌐 网站
Small-world network — Wikipedia
en.wikipedia.org · 来源:
随机图与网络模型:真实网络长什么样?
🌐 网站
Barabasi-Albert model — Wikipedia
en.wikipedia.org · 来源:
随机图与网络模型:真实网络长什么样?
🌐 网站
NetworkX Documentation — Graph Generators
networkx.org · 来源:
随机图与网络模型:真实网络长什么样?
📖 书籍
Probabilistic Graphical Models: Principles and Techniques (Koller & Friedman, 2009)
mitpress.mit.edu · 来源:
概率图模型:图上的不确定性推理
📖 书籍
Pattern Recognition and Machine Learning, Chapter 8: Graphical Models (Bishop, 2006)
microsoft.com · 来源:
概率图模型:图上的不确定性推理
📄 论文
An Introduction to Variational Methods for Graphical Models (Jordan et al., 1999)
doi.org · 来源:
概率图模型:图上的不确定性推理
📄 论文
Constructing Free-Energy Approximations and Generalized Belief Propagation Algorithms (Yedidia, Freeman & Weiss, 2005)
doi.org · 来源:
概率图模型:图上的不确定性推理
📄 论文
Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data (Lafferty, McCallum & Pereira, 2001)
repository.upenn.edu · 来源:
概率图模型:图上的不确定性推理
🌐 网站
pgmpy — Probabilistic Graphical Models library for Python
pgmpy.org · 来源:
概率图模型:图上的不确定性推理
🌐 网站
Bayesian network — Wikipedia
en.wikipedia.org · 来源:
概率图模型:图上的不确定性推理
🌐 网站
Belief propagation — Wikipedia
en.wikipedia.org · 来源:
概率图模型:图上的不确定性推理
📄 论文
DeepWalk: Online Learning of Social Representations (Perozzi et al., 2014)
arxiv.org · 来源:
图嵌入与图神经网络:把图变成向量
📄 论文
node2vec: Scalable Feature Learning for Networks (Grover & Leskovec, 2016)
arxiv.org · 来源:
图嵌入与图神经网络:把图变成向量
📄 论文
LINE: Large-scale Information Network Embedding (Tang et al., 2015)
arxiv.org · 来源:
图嵌入与图神经网络:把图变成向量
📄 论文
Graph Attention Networks (Veličković et al., 2018)
arxiv.org · 来源:
图嵌入与图神经网络:把图变成向量
📄 论文
Inductive Representation Learning on Large Graphs — GraphSAGE (Hamilton et al., 2017)
arxiv.org · 来源:
图嵌入与图神经网络:把图变成向量
🌐 网站
PyTorch Geometric (PyG) Documentation
pytorch-geometric.readthedocs.io · 来源:
图嵌入与图神经网络:把图变成向量
🌐 网站
Deep Graph Library (DGL) Documentation
dgl.ai · 来源:
图嵌入与图神经网络:把图变成向量
🌐 网站
Stanford CS224W: Machine Learning with Graphs
web.stanford.edu · 来源:
图嵌入与图神经网络:把图变成向量
📄 论文
Aho, A. V., Lam, M. S., Sethi, R., & Ullman, J. D. (2006). Compilers: Principles, Techniques, and Tools (2nd Edition). Addison-Wesley
suif.stanford.edu · 来源:
图建模案例集:这个问题其实是图问题
📄 论文
Chaitin, G. J. (1982). Register Allocation & Spilling via Graph Coloring. SIGPLAN Notices, 17(6), 98-101
doi.org · 来源:
图建模案例集:这个问题其实是图问题
📄 论文
Compeau, P. E. C., Pevzner, P. A. & Tesler, G. (2011). How to apply de Bruijn graphs to genome assembly. Nature Biotechnology, 29, 987-991
doi.org · 来源:
图建模案例集:这个问题其实是图问题
📄 论文
Pearl, J. (2009). Causality: Models, Reasoning, and Inference (2nd Edition). Cambridge University Press
doi.org · 来源:
图建模案例集:这个问题其实是图问题
📄 论文
Kempe, D., Kleinberg, J. & Tardos, E. (2003). Maximizing the Spread of Influence through a Social Network. KDD 2003
doi.org · 来源:
图建模案例集:这个问题其实是图问题
🌐 网站
NetworkX Documentation — Algorithms
networkx.org · 来源:
图建模案例集:这个问题其实是图问题
🌐 网站
Google OR-Tools — Optimization
developers.google.com · 来源:
图建模案例集:这个问题其实是图问题
🌐 网站
Graph neural network — Wikipedia
en.wikipedia.org · 来源:
图建模案例集:这个问题其实是图问题
🌐 网站
Causal inference — Wikipedia
en.wikipedia.org · 来源:
图建模案例集:这个问题其实是图问题
🌐 网站
De Bruijn graph — Wikipedia
en.wikipedia.org · 来源:
图建模案例集:这个问题其实是图问题