5. Syllabus
• 10 lectures
• Basic ML algorithms and their applications
• Assignments and in-class practice
• Competitions
• Individual projects
• Tutorials
More info here https://mlcourse.ai/roadmap
6. What makes it different
• Lots and lots of practice
• Theoretical understanding of
applied techniques
• Delving into competitions
• Your own projects
• Really vibrant community!
7. Roadmap/logistics
• All communication in ODS Slack, #mlcourse_ai
• https://mlcourse.ai/roadmap
• 10 assignments – ~10 credits each
• Projects, competitions, tutorials – up to 40 crd. each
• Current rating is here https://goo.gl/TGGr3b
• All materials are stored on GitHub https://github.com/Yorko/
mlcourse.ai and https://mlcourse.ai
• Top-100 participants will be mentioned on a special Wiki page
9. Lecture 1
• Data analysis with Pandas
• Practice on first steps after
getting data
10. Lecture 2
• Visual data analysis with
Pandas and Seaborn
• Crucial plots for feature
exploration
• Practice on «drawing»
11. Lecture 3
• Foundations of Machine
Learning
• Supervised learning
• Decision trees
• k Nearest Neighbours
• Practice: first steps with
Scikit-learn
12. Lecture 4
• Linear classification models
• Regularization
• Cross-validation
• Practice on logistic regression
for a "real-world" task
13. Lecture 5
• Ensembles, random forest
• Feature importance
• Practice on random forest and
assessing feature importance
14. Lecture 6
• Regression task
• Linear and non-linear
regression models
• Practice on grasping core
ideas behind linear regression
15. Lecture 7
• Unsupervised Learning
• Principal Component Analysis
• Clustering
• Practice: clustering Samsung
Galaxy S3 sensor data into
types of human activity
16. Lecture 8
• Stochastic Gradient Descent
& Online learning
• Learning with a couple GB of
data
• Vowpal Wabbit
• Extracting simple features
from texts
• Practice: text classification
17. Lecture 9
• Time series
• Classical and modern
approaches
• Practice: ARIMA model,
Facebook Prophet
18. Lecture 10
• Gradient boosting: a modern
view
• Theoretical basis for gradient
boosting
• Best implementations
• Practice: beating a baseline in
a Kaggle Inclass competition
Regularization?
19. Assignments
• Full versions are announced
during course sessions https://
mlcourse.ai/assignments
• Demo versions are found in
course repo https://
github.com/Yorko/mlcourse.ai
• And in a Kaggle Dataset
mlcourse.ai https://
www.kaggle.com/kashnitsky/
mlcourse
20. Kaggle Inclass
• Alice - tracking visited websites
to distinguish Alice from all others
• Medium - predicting #claps for a
story on Medium
More info here https://mlcourse.ai/roadmap
21. Individual projects
• Throughout the whole course
• Straightforward instructions
• Your own data or just Kaggle
Datasets
• Peer review
• Very cool experience
More info here https://mlcourse.ai/roadmap
22. Project "Alice"
• A substitute for an individual
project if you don't have cool
ideas for one
• Clear instructions
• 6 weeks, 6 notebooks to
complete
• In cooperation with Yandex
and MIPT, specialization
"Machine Learning and Data
Analysis"
• Solutions are not shared
Tutorials
• Your own tutorials on pretty
much any topic around ML & DS
• Peer-voted
• Nice way to grasp something
yourself is to write a tutorial
More info here https://mlcourse.ai/roadmap
23. More info in Slack
#mlcourse.ai, pinned items
Good luck!
https://mlcourse.ai/news