Independent study companion · proof-based graph theory · West, Chapters 1–7

Find the certificate. Write the proof.

A proof-based first course in graph theory where your own code searches small graphs and checks certificates, and every lab ends in a proof you write.

24 modules planned · 9 out now · about 20 hours so farFreeHands-on labs

Assumes a first proofs course and basic Python. Your code, assembled proofs and counterexamples are checked automatically; written proofs you check yourself against explicit criteria and a model proof. Independent project. Not affiliated with or endorsed by the University of Illinois.

Your Learning Path

Nine of the 24 modules are out, in three runs. A good route: modules 1, 2 and 3 (the basics), then 10, 11 and 13 (matchings; 13 does not need 12), then 14, 15 and 16 (connectivity; module 16 uses König–Egerváry from module 11). Where a module leans on one that is not out yet, it recaps what it needs.

Unit 1 · Fundamental conceptsWest, Chapter 1 · 3 of 6 out

1

Graphs and Their Models — From raw data to a graph, and the first counts proved by bijection

Six exams and four students' course lists become a conflict graph. Learn what a graph is, the families that recur (paths, cycles, complete and complete bipartite graphs), count their edges by bijections, prove the complement edge count, and tell subgraphs from induced subgraphs. Then build models in code, give evidence either way, and assemble, repair and write proofs yourself.

You build: A conflict graph built from course lists, a check that tells a subgraph from an induced one, and a tournament schedule with a proof

  • graph models
  • complete and bipartite graphs
  • induced subgraphs
  • proof writing
Free110 minReading materialInteractive lab
2

Degrees, Representations, and Isomorphism — When two drawings are the same graph, and short certificates either way

Two sketches of one network: are they the same graph? Store graphs as adjacency lists and matrices, define isomorphism with both directions, prove that degree sequences and cycle counts are invariants and that color refinement colors are too, and prove the basic facts about the n-cube and the Petersen graph. Then search for isomorphisms, certify differences, and assemble, repair and write proofs yourself.

You build: A search that finds a renaming or shows none exists, a certificate that tells two graphs apart, and a relabeling of the cube with a proof

  • isomorphism
  • invariants
  • color refinement
  • the n-cube and the Petersen graph
Free125 minReading materialInteractive lab
3

Connectivity and Bipartite Graphs — Two shifts or an odd cycle: one search finds whichever certificate exists

Five volunteers cannot be split into two shifts without a clash, and a triangle proves it. Turn walks into paths, use a longest path to force long cycles, prove that odd closed walks contain odd cycles, and prove König's theorem that a graph is bipartite exactly when it has no odd cycle, as a search that returns a two-coloring or an odd cycle. Then assemble, repair and write proofs yourself.

You build: A two-coloring search that returns an odd cycle when it must, a long-cycle finder, and a two-room split with a proof

  • walks and paths
  • bipartite graphs
  • odd cycles
  • proof writing
Free115 minReading materialInteractive lab
4

Eulerian Circuits — Euler's theorem, Hierholzer's algorithm, Königsberg resolved, and Mantel's bound

Coming soon
5

Degree Sequences and Havel–Hakimi — Which sequences are graphic, and how to build the graph

Coming soon
6

Digraphs, Tournaments, and de Bruijn Sequences — Directed walks, strong connectivity, Eulerian digraphs, kings

Coming soon

Unit 2 · Trees and distanceWest, Chapter 2 · coming soon

7

Trees and Their Characterizations — Leaves, the four equivalent definitions, distance, and Jordan's center theorem

Coming soon
8

Counting Trees: Prüfer Codes and Cayley's Formula — A bijection that counts n^(n−2) labeled trees, and Kirchhoff's determinant

Coming soon
9

Minimum Spanning Trees — Kruskal, Prim, and the cut property that proves them both

Coming soon

Unit 3 · Matchings and factorsWest, Chapter 3 · 3 of 4 out

10

Matchings and Augmenting Paths — Maximal versus maximum, path repairs, and a full proof of Berge's theorem

A pairing can be stuck, with no edge left to add, and still not be the largest. Repair it along an alternating path, prove what the symmetric difference of two matchings looks like, prove Berge's theorem including its endpoint argument, and justify a complete alternating search in bipartite graphs. Then assemble, repair and write proofs yourself.

You build: An alternating search that repairs a stuck pairing, a symmetric-difference certificate, and an instrument repair with a proof

  • matchings
  • augmenting paths
  • Berge's theorem
  • proof writing
Free140 minReading materialInteractive lab
11

Hall's Theorem and König–Egerváry — When every applicant can be placed, and the short evidence that proves it either way

Five applicants cannot all be placed, and a short certificate proves it: a set of applicants who between them qualify for too few jobs. Prove Hall's theorem by turning a stuck search into that certificate, prove König–Egerváry by reading a minimum cover off the same search, then assemble, repair and write proofs yourself.

You build: An alternating search, a König cover, and a proof that no roster is forced

  • Hall's theorem
  • vertex covers
  • certificates
  • proof writing
Free145 minReading materialInteractive lab
12

Stable Matchings and Gale–Shapley — Preferences, blocking pairs, and the proposal algorithm

Coming soon
13

Tutte's Theorem and Factors — Why some graphs cannot be perfectly matched, the short evidence that proves it, and how to build a 2-factor

Twenty-two devices, each compatible with exactly three others, cannot all be paired off, and one deleted vertex proves it: three odd pieces are left, and one vertex cannot rescue all three. You will prove that such evidence always rules out a perfect matching, follow a marked sketch of why its absence guarantees one (Tutte's theorem), and use the Berge–Tutte formula to say exactly how far from perfect a graph is. Then you will prove that every 2k-regular graph has a 2-factor and build one in code from closed trails. In the lab you search for the evidence, assemble a proof from its steps, find the false step in another, and write one for a problem that does not say what it is.

You build: A 2-factor built from closed trails, a search for the set whose removal rules out a perfect matching, and an audit with a proof

  • Tutte's theorem
  • odd components
  • Berge–Tutte formula
  • 2-factors
Free170 minReading materialInteractive lab

Unit 4 · Connectivity and network flowsWest, Chapter 4 · 3 of 4 out

14

Connectivity and Edge-Connectivity — How many routers or cables must fail, and why κ ≤ κ′ ≤ δ

Every router has three cables, yet one router or two cables split the network. Define connectivity and edge-connectivity with their conventions, prove that every disconnecting set contains an edge cut, prove Whitney's inequality by charging a vertex cut to an edge cut, and see each inequality fail to be tight. Then assemble, repair and write proofs yourself.

You build: Whitney's construction in code, a weakest-link search, and a proof that a growing network cannot be cut cheaply

  • connectivity
  • edge cuts
  • Whitney's inequality
  • proof writing
Free135 minReading materialInteractive lab
15

2-Connected Graphs and Ear Decompositions — Two routes between any two routers, and a recipe that builds every robust network

No single router failure splits the prism, but why? Prove Whitney's theorem that a graph is 2-connected exactly when every two vertices have two routes sharing no stop, prove the expansion lemma, and prove that 2-connected graphs are exactly the graphs built from a cycle by adding open ears. Then grow ear decompositions in code, find routes where the obvious first choice fails, and assemble, repair and write proofs.

You build: An ear decomposition in code, two disjoint routes past a trap, and a proof that an inspector's loop always exists

  • 2-connected graphs
  • Whitney's theorem
  • ear decompositions
  • proof writing
Free130 minReading materialInteractive lab
16

Menger's Theorem — The fewest vertices that separate two vertices equals the most paths that avoid each other

How many routes between two routers share no router in between? A cut of the same size proves no more exist. Prove Menger's theorem by induction, with König–Egerváry as the base case, prove that k-connected means k disjoint paths between every pair, then grow disjoint routes by rerouting and read off the matching cut. Then assemble, repair and write proofs yourself.

You build: A rerouting search that returns the most disjoint routes and a cut of the same size, and a capacity version with a proof

  • Menger's theorem
  • disjoint paths
  • cuts
  • proof writing
Free155 minReading materialInteractive lab
17

Network Flows and Max-Flow Min-Cut — Feasible flows, residual networks, Ford–Fulkerson, and the theorem that pays for everything

Coming soon

Unit 5 · Planar graphsWest, Chapter 6 · coming soon

18

Planar Graphs and Euler's Formula — Embeddings, faces, duals, and the edge bound that rules out K5 and K3,3

Coming soon
19

Kuratowski, Wagner, and Outerplanar Graphs — Subdivisions, minors, and the two obstructions that characterize planarity

Coming soon

Unit 6 · ColoringWest, Chapter 5 and Section 7.1 · coming soon

20

Vertex Coloring and Greedy Bounds — Chromatic number, greedy coloring, degeneracy, and six colors for every map

Coming soon
21

Critical Graphs, Mycielski, and Brooks' Theorem — What forces colors: minimal k-chromatic graphs, triangle-free graphs of any chromatic number, and χ ≤ Δ

Coming soon
22

Edge Coloring — Line graphs, König, Shannon, Vizing, and the Four Color Theorem as an edge-coloring statement

Coming soon

Unit 7 · Hamiltonian cycles and extremal graphsWest, Sections 7.2 and 5.2 · coming soon

23

Hamiltonian Cycles — Hamilton's puzzle, a necessary condition, and Dirac's theorem

Coming soon
24

Turán's Theorem and the Ten Theorems — Extremal graphs, the Ershov–Kozhukhin bound, and the course in ten results

Coming soon

Who this course helps

  • Students in a proof-based graph theory course who want practice that gives feedback
  • Anyone working through West's Introduction to Graph Theory on their own

A practical body of work from Introduction to Graph Theory

  • State theorems with every hypothesis
  • Prove min–max results with certificates
  • Write complete proofs of statements nobody labels for you

Don't just watch graph algorithms. Explore small graphs with code, find the certificate, turn it into a proof, then solve a problem that does not tell you which theorem to use.

About This Course

This is a course in proving things about graphs: matchings, connectivity, flows, planarity, colorings. It follows the topics of a standard first course built on West's Introduction to Graph Theory, Chapters 1–7, the kind where the exams ask you to prove a statement, not run an algorithm.

Every module has two halves. First a lesson you read: one concrete graph, the theorem stated with every hypothesis, and a full proof written the way a graded solution should be, with notes on how the proof is built. Then a lab in your browser. About half of it is code that serves a proof: your program grows a search and shows where it gets stuck, or checks the certificate that settles a question. The other half is proofs themselves: put a proof's steps in order while avoiding the tempting wrong ones, find the false step in a convincing argument, and write a proof of your own against explicit criteria.

The full course has 24 modules in seven units, following the chapters of West. They are published in the order a fall semester reaches them, so the list below shows what is ready now and what is coming. Out so far: modules 1–3 (graphs and their models, isomorphism, walks, cycles and bipartite graphs), modules 10, 11 and 13 (matchings, Hall's theorem, Tutte's theorem and factors) and modules 14–16 (connectivity, through Menger's theorem); each recaps what it needs from earlier modules that are not out yet.

Topic list modeled on the public syllabus of UIUC Math 412 (Graph Theory), which follows D. B. West, Introduction to Graph Theory, 2nd ed., Chapters 1–7. All lessons, proofs, examples, code and exercises are original to SciMigo; no course materials are reproduced. This course is independent and not affiliated with or endorsed by the University of Illinois.

Prerequisites

  • A first proofs course: induction, contradiction, sets and functions
  • Basic Python: functions, lists, dictionaries, loops

What You Will Learn

  • State the core theorems of graph theory with every hypothesis and quantifier
  • Prove classification, extremal and min–max results in writing
  • Produce and verify certificates: Hall violators, vertex covers, cuts, colorings
  • Test a conjecture on every small graph before you try to prove it
  • Move between a property's algorithmic, structural and certificate characterizations

Introduction to Graph Theory: frequently asked questions

Will it help with my graph theory homework?

It teaches the same kind of reasoning on original problems: exact statements, certificates, complete proofs. It does not solve your course's homework or textbook exercises, and your course's collaboration policy still applies.

How are written proofs checked?

You check your own proof against explicit criteria, then compare it with a model proof. Assembled proofs and counterexamples are checked automatically.

Module 1: Graphs and Their Models

From raw data to a graph, and the first counts proved by bijection. About 110 minutes. Runs in your browser. Nothing to install, no account needed.

Start module 1 — free