Stanford cs224w notes pdf Encoder maps from nodes to embeddings 2. ¡In this lecture, we overview the traditional Start and end math equations with $$ for both inline and display equations!To make a display equation, put one newline before the starting $$ a newline after the ending $$. Reading for the class will from: • Networks, Crowds, and Markets: Reasoning About a Highly Connected World by D. ¡A: an !×! adjacency matrix. CS224W: Machine Learning with Graphs Stanford CS224W: Machine For external inquiries, personal matters, or in emergencies, you can email us at cs224w-aut2425-staff@lists. CS224W: Machine Learning with Graphs Stanford CS224W: Machine •Open-book, open-notes, but do not discussthe exam with any other students until after Saturday. For external inquiries, personal matters, or in emergencies, you can email us at cs224w-aut2425-staff@lists. Networks are a fundamental tool for modeling complex social, technological, and biological systems. Rev. Q: Are there any special submission instructions for reports? ¡1) New problem: Outbreak detection ¡ (2) Develop an approximation algorithm § It is a submodularopt. §Edge type for edge ¡Networks with equal number of nodes and edges: § ER random graph § Scale-free network ¡ Study the properties of the network as an increasing fraction of nodes are removed ¡So far: Decision Based Models § Utility based § Deterministic § “Node” centric: A node observes decisions of its neighbors and makes its own decision ¡Homework 1 recitation session was yesterday (Wed Oct 9th) §Check Ed for recording ¡Colab 1 due today ¡Homework 1 due in 1 week ¡Colab 2 will be released today by 9PM on our ¡Task: Map nodes into an embedding space §Similarity of embeddings between nodes indicates their similarity in the network. edu Granovetter makes a connection between social and structural role of an edge First point: Structurally embedded edges are also socially strong. Contribute to Yasoz/cs224w-zh development by creating an account on GitHub. edu Regulatory Networks Image credit: ese. The OAE will evaluate the request, recommend accommodations ¡Knowledge in graph form §Capture entities, types, and relationships ¡ Nodes are entities ¡ Nodes are labeled with their types ¡ Edges between two nodes capture relationships For personal matters, or in emergencies, e-mail us at cs224w-aut1920-staff@lists. nodes and edges are taken from the original graph G. Two ways to ask questions during lecture: In-person (encouraged) On Ed: At the beginning of class, we will open a new discussion thread dedicated to this lecture When to ask on Ed? ¡Many online settings where one person expresses an opinion about another (or about another’s content) § I trust you [Kamvar-Schlosser-Garcia-Molina ‘03] § I agree with you [Adamic-Glance ’04] CS224W: Social and Information Network Analysis Fall 2015 CS224W: Course Information Instructor Lada Adamic O ce Hours: Thursdays 10:30-11:30am, Location TBD ¡Homework 1 due today §Gradescopesubmissions close at 11:59 PM ¡Homework 2 will be released todayby 9PM on our course website ¡Homework 2: §Due Thursday, 11/02 (2 weeks from now) CS224W: Social and Information Network Analysis Fall 2014 CS224W: Course Information Instructor Jure Leskovec O ce Hours: Wednesdays 9-10am, Gates 418 In this lecture, we investigate graph analysis and learning from a matrix perspective. Easley and J. problem! ¡ (3) Speed-up greedy hill-climbing § Valid for optimizing general submodularfunctions ¡Goal: Recommend a list of possible friends ¡ Supervised machine learning setting: § Labeled training examples: § For every user E have a list of others she will create links to {G Colabs 0 and 1 will be released today (Thu 1/12) by 9PM on our course website Colab 1: Due on Thursday 1/26 (2 weeks from today) Submit written answers and code on Gradescope Lecture notes and further reading. , Ren, H. edu 11/14/23 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 10!!! "!! resides in a cycle with length 3! " resides in a cycle with length 4 …!! The computational graphs for nodes " # and " $ are always the same J. ¡S ∈ $!×#: scalar features. edu 3 Model: 2-dim grid where each node has one random edge This is a small-world Fact: A decentralized search algorithm in Watts-Strogatz model needs n2/3 steps to reach t in expectation CS224W: Machine Learning with Graphs Jure Leskovec, Weihua Hu, Stanford University http://cs224w. Stanford University ¡Intuition: Map nodes to !-dimensional embeddings such that similar nodes in the graph are embedded close together 3 f ( ) = Input graph 2D node embeddings CS224W: Analysis of Networks Fall 2017 CS224W: Course Information Instructor Jure Leskovec O ce Hours: Tuesdays 9:00-10:00AM, Gates 418 Lectures 1:30PM-2:50PM Tuesday and Thursday in NVIDIA Auditorium, Huang Engineering Center There is no official text for this course. Coupled with the emergence of online social networks and large-scale data availability in biological sciences, this course focuses on the analysis of massive networks which provide many computational, algorithmic, and modeling challenges. The OAE will evaluate the request, recommend accommodations ¡Using effective features over graphs is the key to achieving good test performance. ¡Decision based models (today!): § Models of product adoption, decision making § A node observes decisions of its neighbors and makes its own decision For example, last time we talked about Observations and Models for the Web graph: 1) We took a real system: the Web 2) We represented it as a directed graph 3) We used the language of graph theory ¡Some example datasets: § Author Citation/Collaboration Networks § ANetMinerand Microsoft Academic Graph § Pinterest (to be released): § Users: age, gender, boards they own Language practices (norms, etiquette, …) build collective identity foster individual expression Linguistic change captures the relation between Information Explosion in the era of Internet 10K+ movies in Netflix 12M products in Amazon 70M+ music tracks in Spotify 10B+ videos on YouTube Nodes arrive in order 1,2,3, … , 𝑛𝑛 When node 𝒋𝒋 is created it makes a single out-link to an earlier node 𝒊𝒊 chosen: 1) With prob. Decoder maps from embeddings to the These notes form a concise introductory course on machine learning with large-scale graphs. The OAE will evaluate the request, recommend accommodations ¡Social networks: §Facebook §Twitter §Instagram §Etc. CS224W: Machine Learning with Graphs Qian Huang, Stanford University Huang, Q. Runshort fixed-length random walks starting from each node on the graph using some strategy R 2. problem! ¡ (3) Speed-up greedy hill-climbing §Valid for optimizing general submodularfunctions ¡Observation 1 could also have issues: §Even though two nodes may have the same neighborhood structure, we may want to assign different embeddings to them Problem: (k is user-specified parameter)¡ Most influential set of size k: set Sof k nodes producing largest expected cascade size f(S) if activated [Domingos-Richardson ‘01] Q: How do I scan and create a PDF from a set of handwritten notes? Many printers and photocopiers have a create PDF feature. Make sure the PDF size is smaller than 10MB. Small-World Model [Watts-Strogatz ‘98] Two components to the model: ¡ (1) Start with a low-dimensional regular lattice § (In our case we are using a ring as a lattice) Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 26 Knowledge Graphs Image credit: Maximilian Nickel et al 3D Shapes Image credit: Wikipedia Code Graphs Image credit: ResearchGate Molecules Image credit: MDPI Scene Graphs Image credit: math. The class final project will offer you an opportunity to do exactly this. , Chen, P. , a measure of similarity in the original network) 3. ¡Traditional ML pipeline uses hand-designed features. In the following series of blog posts, I share my notes which I took watching lectures. hws. For external inquiries, personal matters, or in emergencies, you can email us at cs224w-aut2324-staff@lists. CS 246 Mining Massive Datasets also deals with interconnected data. D. , Liang, P. ⊕ The notes are still under construction! They will be written up as lectures continue to progress. The OAE will evaluate the request, recommend accommodations Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 7 Knowledge Graphs Image credit: Maximilian Nickel et al 3D Shapes Image credit: Wikipedia Code Graphs Image credit: ResearchGate Molecules Image credit: MDPI Scene Graphs Image credit: math. 𝒑𝒑, 𝒋𝒋 links to 𝒊𝒊 chosen uniformly at ¡Networks with equal number of nodes and edges: § ER random graph § Scale-free network ¡ Study the properties of the network as an increasing fraction of nodes are removed Definition: Networks with a power law tail in their degree distribution are called “scale-free networks” Where does the name come from? Scale invariance: There is no characteristic scale ¡We want to generate realistic graphs, using graph generative models ¡Applications: §Drug discovery, material design §Social network modeling 11/11/21 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs, cs224w. Submission Template for HW0 [pdf | tex | docx]. stanford. , a word 9/23/2013 Jure Leskovec, Stanford CS224W: Social and Information Network Analysis 3 Given real G, construct rewired network G’ Same degree distribution but random connections Consider G’ as multigraph The expected number of edge between nodes Nov 8, 2021 · There is no official text for this course. The OAE will evaluate the request, recommend accommodations •Open-book, open-notes, but do not discussthe exam with any other students until after Saturday. The Stanford Network Analysis Project (SNAP) provides a detailed explanation of the PageRank algorithm and its applications in network analysis. ¡ML tasks: §Friend recommendation (link-level) §User property prediction (node-level) 3/13/21 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 3 A: Reports (project proposal and final project report) should be submitted in pdf files via Gradescope. Each node has scalar attributes, e. Course Materials CS224W Stanford School of Engineering Stanford School of Engineering Winter 2024-25: Online, CS224W: Machine Learning with Graphs Joshua Robinson and Jure Leskovec, Stanford University http://cs224w. 10/10/24 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs, http://cs224w. ¡ML tasks: §Recommend items (link prediction) §Classify users/items (node classification) 11/16/23 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 3 GNNs & LLMs in PyG By: Rishi Puri, Junhao Shen, & Zack Aristei NVIDIA, Southern Methodist University, & Georgia Tech For external inquiries, personal matters, or in emergencies, you can email us at cs224w-win2223-staff@lists. Academic accommodations: If you need an academic accommodation based on a disability, you should initiate the request with the Office of Accessible Education (OAE). atom type for molecules. Submission Template for HW0 [pdf | tex Lecture notes from CS224W: Machine Learning with Graphs Jure Leskovec, Stanford University http://cs224w. Gomes-Selman, R. We define a flexible notion of node’s network neighborhood and design a biased ra. Networks: An introduction by Mark Newman. 9/22/2021 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 38 We are going to cover various topics in Machine Learning and Representation Learning for graph ¡Observation 1 could also have issues: §Even though two nodes may have the same neighborhood structure, we may want to assign different embeddings to them 1. The rest of the blog posts you can find here: 2, 3, CS224W-Chinese-Notes CS224W中文笔记. ¡Statusin a network [Davis-Leinhardt’68] §A B :: Bhas higherstatus than A §A B :: B has lowerstatus than A §Note: Here the notion of status is now implicit and governed by the ¡A heterogeneous graph is defined as !=#,%,&,’ §Nodes with node types (∈* §Node type for node !: §Edges with edge types (,,()∈. A: Reports (project proposal and final project report) should be submitted in pdf files via Gradescope. We created notes borrowing contents from the lecture slides and expanding on topics discussed in lecture. edu SCPD students can attend o ce hours remotely via a Google Hangout; the link will be posted on Piazza just before the o ce hours start. Generalize one-hop queries to path queries by adding more relations on the path. ), ’links to (chosen uniformly at CS224W Course | Stanford University Bulletin CS224W Download as PDF. 5 %ÐÔÅØ 2 0 obj /Type /ObjStm /N 100 /First 830 /Length 1732 /Filter /FlateDecode >> stream xÚÍZMsÔ8 ½Ï¯Ð ’¬/S U,즶v pÌÅ™ ! 2/28/2023 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 25 Position encoding for graphs: Represent a node’s position by its distance to randomly selected anchor-sets Dec 20, 2020 · Recently, I finished the Stanford course CS224W Machine Learning with Graphs. If you are interested in research, CS224W will also leave you well-qualified to do network science research. Define a node similarity function (i. 09/26: Introduction and Structure of Graphs Reading: Chapter 1: Overview. CS224W expects you to have decent knowledge in deep learning and all graph neural network techniques build on top of “typical” deep learning approaches. Q: Are there any special project requirements for SCPD students? A: Historically, SCPD students had no trouble finding project partners and perform the required work. The following books are recommended as optional reading: Networks, Crowds, and Markets: Reasoning About a Highly Connected World by David Easley and Jon Kleinberg. hoods of nodes. Identity-aware Graph Neural Networks, AAAI 2021 ¡ Geometric GraphLearning From Representation to Generation Guest lecture, Stanford CS224W Machine Learning with Graphs Minkai Xu Jure Leskovec 9/29/19 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs, cs224w. , a word Networks with positive and negative relationships Our basic unit of investigation will be signed triangles First we talk about undirected networks then directed Geometric GraphLearning From Representation to Generation Guest lecture, Stanford CS224W Machine Learning with Graphs Minkai Xu Jure Leskovec We will analyze the following model: ¡ Nodes arrive in order 1,2,3,…,& ¡ When node ’ is created it makes a single out-link to an earlier node ( chosen: § 1) With prob. General Advice for the Exam We suggest that you read through all lecture slides carefully Lectures that are relatively unimportant for the exam: lectures 1, 3, 5, 9, 12, 14 ¡Goal: identify a specific use case and demonstrate how GNNs and PyG can be used to solve this problem ¡Output: blog post, Google colab ¡Example use cases §Fraud detection A graph G = (A, S) is a set V of n nodes connected by edges. What if 𝑽′and 𝑬′come from a totally different ¡Homework 2 due today at midnight ¡Homework 3 will be released today by 9PM on our course website §Due Thursday, 11/16 (2 weeks from now) §TAs will hold a recitation session for HW 3: %PDF-1. PRODIGY: Enabling In-context Learning Over Graphs. The OAE will evaluate the request, recommend accommodations Lecture notes and further reading Pointers to the slides will be posted here just before the start of the class. In node2vec , we learn a mapping of nodes to a low-dimensional space of features that maximizes the likelihood of preserving distances between net-work neighbo. CS224W is a completion requirement for: CS224W Course | Stanford University Bulletin CS224W Download as PDF. CS224W is a completion requirement for: Sep 21, 2021 · For external inquiries, personal matters, or in emergencies, you can email us at cs224w-win2021-staff@lists. ), ’links to (chosen uniformly at A GNN will generate the same embedding for nodes 1 and 2 because: Computational graphs are the same. An J-hop path query Mcan be represented by M=(𝑣 , N1,…, N ) 𝑣 is an “anchor” entity, We will analyze the following simple model: ¡Nodes arrive in order 1,2,3,…,& ¡When node ’is created it makes a single out-linkto an earlier node (chosen: §1)With prob. You, J. Node features (colors) are identical. The OAE will evaluate the request, recommend accommodations 9/27/2011 Jure Leskovec, Stanford CS224W: Social and Information Network Analysis 12 . In the following series of blog posts, I share my notes which I took watching lectures. E2004 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 2 CS224W: Machine Learning with Graphs Jure Leskovec, Stanford University http://cs224w. 2, 2014). You can also use smartphone apps such as the Genius Scan app for iOS and Android to create a PDF. The OAE will evaluate the request, recommend accommodations ¡Main question today: Given a network with labels on some nodes, how do we assign labels to all other nodes in the network? ¡Example: In a network, some nodes are CS224W: Social and Information Network Analysis Homework 0 (Due at 9:30am Oct. edu 16 ¡ We specifically consider local neighborhood structures around each node in a graph. , & Leskovec, J. Homework 0 (Due at 11:59pm Oct. , KDD 2012 Last Lecture: Roles This Lecture: Communities Clauset, et al. Leskovec. g. edu. Kleinberg (PDF available online). Lecture notes from 9/23/2014 Jure Leskovec, Stanford CS224W: Social and Information Network Analysis 13 Facebook social graph 4-degrees of separation [Backstrom-Boldi-Rosa-Ugander-Vigna, 2011] 10/17/24 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 12 ¡ Transformers map 1D sequences of vectors to 1D sequences of vectors known as tokens §Tokens describe a ”piece” of data – e. 1 PDF/ZIP file (writeups, experimental results, code) ¡Homework 1 will be released todayby 9PM on our course website ¡Homework 1: §Due Thursday, 10/17 (2 weeks from now) §TAs will hold a recitation session for HW 1: For external inquiries, personal matters, or in emergencies, you can email us at cs224w-aut2122-staff@lists. Using one of the network analysis tools above load the Wikipedia voting network. See course website here and official notes here. Notes and reading assignments will be posted on the course web site. Notes and reading assignments will be posted periodically on the course web site. , Phys. Ying, J. edu Main question: How do we take advantage of Notes. , Zeng, D. es in networks. Oct 6, 2016 · CS224W: Social and Information Network Analysis Homeworks. , Graph Convolutional Neural Networks for Web-Scale Recommender Systems, KDD 2018 2/16/2023 CS224W (Fall 2019) was offered as a course on machine learning methods for networks. edu Note to other teachers and users of these slides: We would be CS224W: Social and Information Network Analysis - Problem Set 3 5 (i) [5 points] For each of the two files, plot the distribution of word frequencies on a log-log scale. CS224W: Machine Learning with Graphs Jure Leskovec, Stanford University http://cs224w. I hope it gives you a quick ¡Homework3 due today §Gradescope submissions close at 11:59 PM ¡Exam opens in one week §Ed post soon about Exam Recitation ¡Colab 5: will be released today §Due Thurs 12/5 ¡Homework3 due today §Gradescope submissions close at 11:59 PM ¡Exam opens in one week §Ed post soon about Exam Recitation ¡Colab 5: will be released today §Due Thurs 12/5 1. I hope it gives you a quick sneak peek overview of how ML applied for graphs. They mirror the topics topics covered by Stanford CS224W, and are written by the CS 224W TAs. NeurIPS 2023 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs Ying et al. wustl. Important Dates Assignment/Work Out Date Due Date Assignment 0 now October 4 Assignment 1 October 2 October 11 ¡Today we will talk about human behavior online ¡ We will try to understand how people express opinions about each other online § We will use data and network science theory to Stanford University (2) Aggregation (1) Message Putting things together: (1) Message: each node computes a message (2) Aggregation: aggregate messages from neighbors Nonlinearity (activation): Adds expressiveness ¡1)New problem:Outbreak detection ¡ (2)Develop an approximation algorithm §It is a submodularopt. BACKGROUND. edu 10 How to find connected components: • Start from random node and perform Breadth First Search (BFS) • Label the nodes that BFS visits • If all nodes are visited, the network is connected • Otherwise find an unvisited node and repeat BFS D C A B H F G 12/5/24 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 27 Task ! Task " Task %: graph-scalefree-const-path Task &: node-CoauthorPhysics Best design Encoder + Decoder Framework Shallow encoder: embedding lookup Parameters to optimize: 𝐙which contains node embeddings for all nodes ∈𝑉 ¡New office hours format: §1 hour of group office hour (general questions) §1 hour of individual office hour (questions/help with individuals’ code) ¡No class on November 7th(Election Day) §Lectures 13 (Advanced Topics in GNNs) to 17 (Link Prediction and Causality) will be pushed back by one For personal matters, or in emergencies, e-mail us at cs224w-aut1920-staff@lists. NOTE: Syllabus is still evolving and will be finalized by 09/22 For external inquiries, personal matters, or in emergencies, you can email us at cs224w-aut2425-staff@lists. 6, 2016). Treating a graph as a matrix allows us to: Determine node importance via random walk (PageRank) For external inquiries, personal matters, or in emergencies, you can email us at cs224w-aut2324-staff@lists. edu Main question: How do we take Dec 20, 2020 · Recently, I finished the Stanford course CS224W Machine Learning with Graphs. edu ¡Prior models and intuition say that the network diameter slowly grows (like log N) time r r size of the graph Internet Citations ¡ Diameter shrinks over time § As the network grows the CS224W: Analysis of Networks Fall 2017 Problem Set 2 Due 11:59pm PDT October 26, 2017 General Instructions These questions require thought, but do not require long answers. Stanford University ¡Goal: create long-lasting resources for your technical profiles + broader graph ML community ¡Three types of projects §1) Real-world applications of GNNs §2) Tutorial on PyGfunctionality ¡Examopens this Thursday 11/21 §11/21 5pm to 11/23 5am (36 hour window) §2 hours long (can't stop + start) §On gradescope – typeset your answers in Latex or upload images One of CS224W main goals is to prepare you to apply state-of-the-art network analysis tools and algorithms to an application. , Krvzmanc, G. Readings and the list of future lectures will be useful to you when you are thinking about the course project. For each node #collect 6 7(#), the multiset* of nodes visited on random walks starting CS224W: Social and Information Network Analysis Fall 2016 CS224W: Course Information Instructor Jure Leskovec O ce Hours: Thursdays 4:30PM-5:30PM, Gates 418 ¡Granovettermakes a connection between social and structural role of an edge ¡First point: Structure §Structurally embedded edges are also socially strong §Long-range edges spanning different parts of the ¡Motifs: §Help us understand how graphs work §Help us make predictions based on presence or lack of presence in a graph dataset ¡Examples: §Feed-forward loops:Found in networks of Modularity of partitioning 𝑺𝑺 of graph 𝑮𝑮: Q ∝ ∑ s∈ S [ (# edges within group s) – (expected # edges within group s) ] 𝑄𝑄𝐺𝐺,𝑆𝑆= 1 The preceding definitions define subgraphs when 𝑉′⊆𝑉and 𝐸′⊆𝐸, i. The OAE will evaluate the request, recommend accommodations CS224W: Social and Information Network Analysis - Problem Set 0 2 2. e. Pointers to the slides will be posted here just before the start of the class. Submission Template for HW0 [pdf | tex Lecture notes from 11/14/23 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 12 ¡ Transformers map 1D sequences of vectors to 1D sequences of vectors known as tokens §Tokens describe a ”piece” of data –e. CS224W: Social and Information Network Analysis Fall 2016 CS224W: Course Information Instructor Jure Leskovec O ce Hours: Thursdays 4:30PM-5:30PM, Gates 418 CS224W Course Notes. , a word 10/3/19 RolX Fast Modularity Henderson, et al. Contribute to snap-stanford/cs224w-notes development by creating an account on GitHub. edu Note to other teachers and users of these ¡Recommender systems: §Amazon §YouTube §Pinterest §Etc. For example: §Both nodes are close to each other (connected by an edge) Goal: create long-lasting resources for your technical profiles + broader graph ML community Three types of projects 1) Real-world applications of GNNs 2) Tutorial on PyG functionality 10/17/24 Jure Leskovec, Stanford CS224W: Machine Learning with Graphs 12 ¡ Transformers map 1D sequences of vectors to 1D sequences of vectors known as tokens §Tokens describe a ”piece” of data – e. We normally use Piazza for SCPD and Stanford students to find project partners. (2023). encbxzgl szv nzoy bwse zdc lrm hfcr orn aoucxxt uvn