Dse 2023-24 — различия между версиями

Материал из Wiki - Факультет компьютерных наук
Перейти к: навигация, поиск
(added github link)
(Log Book or Tentative Plan)
 
(не показаны 4 промежуточные версии 2 участников)
Строка 22: Строка 22:
 
==Log Book or Tentative Plan ==
 
==Log Book or Tentative Plan ==
  
[https://www.youtube.com/playlist?list=PLyjahhN4Wdd9rpesSCbxg5YdtF_cJ1SZC Lecture playlist]
+
[https://www.youtube.com/playlist?list=PLyjahhN4Wdd9rpesSCbxg5YdtF_cJ1SZC 🔗Lecture playlist]
 +
 
 +
[https://www.youtube.com/playlist?list=PLyjahhN4Wdd99xaZOCZ99hNcLiM66QR9E 🔗Consultations playlist]
  
 
'''Week 1. 2023-09-04''': Entropy, [https://github.com/Shuaynat/DSE-23-24/raw/main/03-lectures/Dse2023-L01.pdf pdf]
 
'''Week 1. 2023-09-04''': Entropy, [https://github.com/Shuaynat/DSE-23-24/raw/main/03-lectures/Dse2023-L01.pdf pdf]
Строка 51: Строка 53:
 
Class: group by, reshape and join
 
Class: group by, reshape and join
  
'''Week 3.''': Trees
+
'''Week 3.''': Trees, [https://github.com/Shuaynat/DSE-23-24/raw/main/03-lectures/Dse2023-L03.pdf pdf]
  
 
Class: Trees (regression + classification) + tree visualization
 
Class: Trees (regression + classification) + tree visualization
  
'''Week 4.''' Random forest + Data splitting strategies
+
More:
 +
 
 +
[http://www.r2d3.us/visual-intro-to-machine-learning-part-1/ tree visualization by r2d3]
 +
 
 +
[https://en.wikipedia.org/wiki/Receiver_operating_characteristic accuracy, recall, roc and all of that]
 +
 
 +
'''Week 4.''' Random forest [https://github.com/Shuaynat/DSE-23-24/raw/main/03-lectures/Dse2023-L04.pdf pdf]
 +
 
 +
More:
 +
 
 +
[http://www.r2d3.us/visual-intro-to-machine-learning-part-2/ bias-variance trade-off visualization for trees]
 +
 
 +
[https://arxiv.org/pdf/1411.5279.pdf Tim Hesterberg, What teachers should know about bootstrap?] Very well written text, for permutation test see sections 2.1 and 7.
  
 
Class: Random forest, cross-validation in sklearn, feature importance,  
 
Class: Random forest, cross-validation in sklearn, feature importance,  
  
'''Week 5.''': Gradient boosting
+
'''Week 5.''': Gradient boosting, Data splitting strategies
  
 
Class: XGBoost vs LightGBM, Dummy variables, categorical variables and Catboost  
 
Class: XGBoost vs LightGBM, Dummy variables, categorical variables and Catboost  

Текущая версия на 16:12, 31 октября 2023

General course info

16 lectures plus 16 classes

Fall grade = 0.2 Small HAs + 0.2 Group project + 0.3 Midterm + 0.3 Final

Each small HA consists of approximately 4 or 5 problems. Group project may be written by a group of 1-3 students. Midterm and Final are offline and hand written.

Lecturer: Boris Demeshev

Class teachers: Yana Khassan, Shuana Pirbudagova

tg group,

Lectures: Monday, 18:10 - 19:30 Moscow time, zoom

Classes:

  • Thursday, 13:00 - 14:20 Moscow time, D208, Shuana

Github repository of the class: [1]

Log Book or Tentative Plan

🔗Lecture playlist

🔗Consultations playlist

Week 1. 2023-09-04: Entropy, pdf

Guessing game, conditional entropy, joint entropy.

Class: data manipulation, data vizualization

More:

Cristopher Olah, Visual Information Theory https://colah.github.io/posts/2015-09-Visual-Information/

Grand Sanderson, Solving Wordle using information theory https://www.youtube.com/watch?v=v68zYyaEmEA

конспект аналогичной лекции на фкн: https://exuberant-arthropod-be8.notion.site/1-02-09-5e107ea1c4054594b8f37d955db8a2b0

Week 2.: Kelly criterion, pdf

How to calculate expected values using cross-entropy ideas, H(X) - H(X|Q) as long term interest rate.

More:

Kelly criterion

Class: group by, reshape and join

Week 3.: Trees, pdf

Class: Trees (regression + classification) + tree visualization

More:

tree visualization by r2d3

accuracy, recall, roc and all of that

Week 4. Random forest pdf

More:

bias-variance trade-off visualization for trees

Tim Hesterberg, What teachers should know about bootstrap? Very well written text, for permutation test see sections 2.1 and 7.

Class: Random forest, cross-validation in sklearn, feature importance,

Week 5.: Gradient boosting, Data splitting strategies

Class: XGBoost vs LightGBM, Dummy variables, categorical variables and Catboost

Week 6.: Naive bootstrap, t-stat bootstrap, permutation tests

Class: Hypothesis testing

More:

https://arch.readthedocs.io/en/latest/bootstrap/bootstrap.html

Week 7.: Matrices in regression

Class: (by hand) Differential in matrix form, derivation of formulas for beta.

Here will be dragons midterm!


Week 8.: SVD = PCA

Class: (by hand) Covariance matrices,


Week 9.: James Stein paradox

Class: Matrices in numpy, PCA in sklearn, SVD

Week 10.: L1, L2 regularization

Class: Regression in sklearn, different type of regularisation

Week 11.: Log regression + L1/L2

Class: Log regression (sklearn/statsmodels) + L1/L2

Week 12.: Hierarchical clustering + k-means

Class: Hierarchical clustering + k-means

Week 13.: ETS (Exponential Smoothing)

Class: Plotting time series, ETS (sktime)

More:

https://www.sktime.net/en/stable/examples/01_forecasting.html

Week 14.: Bayesian approach

Class: TS forecasting with grad boosting

Week 15.: Mention of MCMC + DLT

Class: DLT in python

More:

Mcmc visualization: https://chi-feng.github.io/mcmc-demo/app.html?algorithm=SVGD&target=banana&delay=0

https://www.uber.com/blog/orbit/

Week 16.: QA

Class: QA

Here will be dragons final!