Tomorrow we'll cover how to overcome noise using bagging.   https://drive.google.com/drive/folders/0BzUXUMab8u_ZLTB6NnpiRUJZNUU

My new ML study group site can be found here http://www.research.ibm.com/haifa/dept/vst/ML-studygroup.shtml
Enjoy!

 

The following set of resources are aimed at motivating high school kids.   Only one link is in Hebrew.  

Self driving car concept.   No driver.

https://m.youtube.com/watch?v=aaOB-ErYq6Y

Prof. Shai Shalev-Shwartz demo on the road to Jerusalem 

https://m.youtube.com/watch?v=cYTVXfIH0MU

10 ml products demo, funny robot,  2.26 self driving car, 5.29 robots trained by example, 6.40 go

https://m.youtube.com/watch?v=dcZvhP-IqY4

Balancing balls demo

https://m.youtube.com/watch?v=fq0JBaiCYNA

3d simulation of neural network 

https://m.youtube.com/watch?v=3JQ3hYko51Y#fauxfullscreen

Better intersection handling 

https://m.youtube.com/watch?v=4pbAI40dK0A

זיהוי דימנציה דוגמה למסוג בינארי  

https://m.youtube.com/watch?v=9SvcktFFUR4

How a simple neural network recognize digits with a lot of animation 

https://m.youtube.com/watch?v=aircAruvnKk

Learning to walk

https://m.youtube.com/watch?v=gn4nRCC9TwQ

Introduction slides

https://courses.edx.org/asset-v1:ColumbiaX+CSMM.101x+1T2017+type@asset+block@AI_edx_ml_5.1intro.pdf

Conjugate prior

A recording of the conjugate prior presentation.  

We'll continue our journey into Bayesian thinking, introducing the concept of conjugates 

How to choose priors?

The logic of science  by Jaynes chapter eleven  covers the maximum entropy principle for choosing priors and gives it an axiomatic treatment 

http://home.fnal.gov/~paterno/images/jaynesbook/cc11g.pdf

See my summary slide 

Mit lecture start from 8 to 35.   " It's the humble way of assigning prior probabilities" - 

https://m.youtube.com/watch?v=ufU7Wyc96M0#fauxfullscreen

Fun lecture!

If you like it the full course on information is here 

https://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-050j-information-and-entropy-spring-2008/syllabus/

A presentation on different ways to choose priors that covers among other things the principle of maximum entropy under  constraints 

http://www.phas.ubc.ca/~oser/p509/Lec_06.pdf

And another one 

https://www.pas.rochester.edu/~sybenzvi/courses/phy403/2015s/p403_07_priors.pdf

Under the Bayesian setting, probabilities represent our brief on the state of the world which we can update incrementally after each experiment.   This naturally lead to sequential processes.     See https://drive.google.com/file/d/1Y_1jwI91TVBVVwYhPZPg2tezp3E0t0_E/view?usp=sharing