Sixteen units, from what machine learning is to the parts that only matter in
finance — bars instead of the calendar, the triple barrier instead of "will it go up tomorrow",
purged cross-validation instead of a naive backtest. Every unit is a lecture and three
hands-on problems, with the data included.
16UNITS
66PDFs
1,729SLIDES
45DATA FILES
Every problem ships its data
There is no download step and no vendor account. Each unit's data/ directory holds
the small CSVs its three problems use, chosen so the hand-calculation in the problem set and the
code you write afterwards agree exactly. Clone the repository, or open any file straight in the
browser — GitHub renders CSV as a table.
Anomaly detection runs as its own thread across the algorithms — LOF in unit 5,
OCSVM in unit 6, Isolation Forest in unit 7, the autoencoder in unit 8 — because in production no
single detector is enough. The same three applications (product recommendation, fraud detection,
an AI chatbot) are revisited at every new level of machinery.
SPECSpecial sessionsoptional lectures beyond the sixteen units · each with a companion note4
SpecialRecommender SystemsThe third running app of the course · content-based, collaborative filtering, matrix factorization23 pp · 492 KB · with noteNoteViewDownload
SpecialRecommender Systems · Part 2Deep-learning recommenders — factorization machines up to graph and language models30 pp · 728 KB · with noteNoteViewDownload
SpecialThe AI-Native Asset ManagerMulti-agent investment firms — agents that research, trade and check risk26 pp · 565 KB · with noteNoteViewDownload
SpecialAI-Native Asset Manager · Part 2The implementation half — building the agent firm in LangGraph28 pp · 512 KB · with noteNoteViewDownload
LSDIn-class decksshort decks used in the face-to-face class · Week 7 · English and Korean6
SUPPSupplementary materialself-contained packages beside the units · deck, data and notebook · English and Korean18
SupplementLinear Regression on the KOSPISUP01 · English · class edition · least squares, metrics, Ridge / Lasso / ElasticNet, polynomial features, the KOSPI notebook cell by cell51 pp · 2.8 MBViewDownload
SupplementKOSPI_Prediction_EN.ipynbSUP01 · notebook · the lab the slides follow, §0–§872 KBColabViewDownload
SupplementKOSPI_Prediction_KO.ipynbSUP01 · 노트북 (한국어) · 슬라이드가 따르는 실습, §0–§872 KBColabViewDownload
SupplementKOSPI_Index_EN.csvSUP01 · data · 160 monthly rows, 29 indicators + KOSPI · with KOSPI_Data_Dictionary.csv26 KBViewDictionaryDownload
SupplementKOSPI_Index_KO.csvSUP01 · 데이터 (한국어 열 이름) · 160개월, 29개 지표 + KOSPI · KOSPI_Data_Dictionary_KO.csv 포함26 KBViewDictionaryDownload
SupplementLogistic Regression on the KOSPISUP02 · English · Week 6 make-up class · sigmoid and maximum likelihood, confusion matrix, ROC / AUC, precision–recall, the KOSPI direction lab58 pp · 4.0 MBViewDownload
Supplement로지스틱 회귀모델 — KOSPI 방향 실습SUP02 · 한국어 · 6주차 보강 수업 · 시그모이드와 최대우도, 혼동행렬, ROC / AUC, 정밀도–재현율, 코스피 방향 예측 실습58 pp · 4.1 MBViewDownload
SupplementKOSPI_Logit_EN.ipynbSUP02 · notebook · parts A–C: the original lab, a fuller evaluation, the lecture examples32 KBColabViewDownload
SupplementKOSPI_Logit_KO.ipynbSUP02 · 노트북 (한국어) · 파트 A–C: 원본 실습, 평가 보강, 강의 예제32 KBColabViewDownload
SupplementKOSPI_Index_EN.csvSUP02 · data · the SUP01 file, unchanged · with KOSPI_Data_Dictionary.csv26 KBViewDictionaryDownload
SupplementKOSPI_Index_KO.csvSUP02 · 데이터 (한국어 열 이름) · SUP01과 같은 파일 · KOSPI_Data_Dictionary_KO.csv 포함26 KBViewDictionaryDownload
SupplementAnomaly DetectionSUP03 · English · One-Class SVM and SVDD, Isolation Forest, LOF, autoencoders — formulas and scikit-learn examples30 pp · 1.1 MBViewDownload
Supplement이상치 탐지SUP03 · 한국어 · 원본 강의 덱 · One-Class SVM, SVDD, Isolation Forest, LOF, 오토인코더24 pp · 0.9 MBViewDownload
SupplementMeta-LabelingSUP04 · English · three stages: side (recall), trade decision (precision), bet size (Kelly)10 pp · 0.2 MBViewDownload
SupplementML Algorithm Family TreeSUP05 · English · four families — supervised, unsupervised, ensembles, reinforcement — and their objectives21 pp · 2.6 MBViewDownload
Supplement머신러닝 알고리즘 계보도SUP05 · 한국어 · 지도·비지도·앙상블·강화학습 네 가족과 각 목적함수21 pp · 2.7 MBViewDownload
MIDMidterma practice set on units 1–7, its answer key, and a cheat sheet for review3
MidtermMidterm Practice SetUnits 1–7 · 48 multiple-choice questions · about 50 minutes7 pp · 109 KBViewDownload
MidtermMidterm Practice Set — Answer Keyanswer grid, formula sheet, and a worked explanation for every question21 pp · 526 KBViewDownload
U05Classification and Anomaly DetectionPart II · Core Algorithms · taught wk 54
PrimerPrimer — From a Number to a VerdictWatch first if probability, odds and distance are hazy · twenty minutes of groundwork17 pp · 161 KBViewDownload
LectureClassification & Our First Anomaly DetectorLogistic, KNN, Naive Bayes — and LOF for the odd one out38 pp · 761 KBViewDownload
ProblemsThree Hands-On ProblemsLogistic by Hand · KNN Vote · Naive Bayes Screening22 pp · 468 KBViewDownload
U09Unsupervised LearningPart II · Core Algorithms · taught wk 94
PrimerPrimer — No Answer Key This TimeWatch first if centroids, variance and eigenvectors are hazy · twenty minutes of groundwork17 pp · 163 KBViewDownload
LectureUnsupervised LearningGrouping, simplifying — and cleaning the noise38 pp · 631 KBViewDownload
ProblemsThree Hands-On ProblemsK-Means by Hand · PCA Eigenvalues · The Noise Edge22 pp · 455 KBViewDownload
U10Honest BacktestingPart III · Financial ML in Depth · taught wk 104
PrimerPrimer — How to Fool YourselfWatch first if the Sharpe ratio and cross-validation are new · twenty minutes of groundwork17 pp · 161 KBViewDownload
LectureHow Not to Fool YourselfBacktesting pitfalls, purged CV — and the Deflated Sharpe38 pp · 674 KBViewDownload
ProblemsThree Hands-On ProblemsScore a Strategy · Purge a Timeline · Price the Luck22 pp · 450 KBViewDownload
U11Portfolio ConstructionPart III · Financial ML in Depth · taught wk 114
PrimerPrimer — Why Two Is Safer Than OneWatch first if portfolio variance and correlation are hazy · twenty minutes of groundwork17 pp · 160 KBViewDownload
LectureBuilding Portfolios That Don't BreakDiversification, HRP, NCO — using the clean covariance from Week 938 pp · 668 KBViewDownload
ProblemsThree Hands-On ProblemsPrice the Free Lunch · Break Markowitz · Run HRP by Hand22 pp · 416 KBViewDownload
U14Causality, XAI and FairnessPart IV · Frontier and Causality · taught wk 144
PrimerPrimer — Correlation Is Not EnoughWatch first if confounders and conditional probability are hazy · twenty minutes of groundwork17 pp · 158 KBViewDownload
LectureCause, Explanation & FairnessCorrelation vs causation, SHAP — and responsible AI38 pp · 656 KBViewDownload
ProblemsThree Hands-On ProblemsDissolve a Signal · Compute a Shapley Value · Measure a Gap22 pp · 422 KBViewDownload