CS 470 - Spring 2020.

Welcome to CS 470!

Video of the Day

Beethoven, Spring Sonata for Violin and Piano

Poll of the Day

Psychology has studied learning. In behavioral psychology, you can learn from positive reinforcement (reward) or negative reinforcement (punishment). According to experimental data, which is more effective? [Note: this is different from Friday's question.]

https://pollev.com/cs470 You may also download the app to your phone. Use the "cs470" poll id.

Coronavirus COVID-19 Global Cases by Johns Hopkins CSSE

Lecture : Learning. 4/6/2020

Administrivia

Zoom Guide for Yale Students Sign in using Yale zoom account.

Security and Privacy Implications of Zoom Some of you took my CS 257 course on Informatation Security. Apparently zoom is rife with security issues. This article is by Bruce Schneier who is a leading security guru, not some quack.

Widely used, including British cabinet meeting.
Bad Privacy Practices: selling data, sending data to Facebook, treating home pages as marketing websites.
Bad security practices: third party can turn on Mac user's camera overriding local security setting, can steal users' Windows credentials, display LinkedIn data, sloppy encryption: claim AES-256, but use AES-128, in ECB mode (electronic code book - not a good idea), use servers in China, have 700 developers in China, who could slip in backdoors a Chinese authorities request.
Bad user configurations: zoombombing, meeting id's are too short - hackers can randomly try them - there are tools for this!

I have office hours Wednesdays from 4-6 pm, via zoom, standard meeting id: 316-021-726

ULA office hours - also on zoom. See the [Contact Info and Schedule] as well as piazza. The zoom office hours meetings are posted on canvas, under zoom. If you have trouble connecting, post a note to piazza or email cs470help@cs.yale.edu.

[Assignments]. hw7 has been posted: hw7.html reinforcement learning.

CS 570 project assignment is also available in [Assignments]. So far thirteen have been submitted and approved.

Google COVID-19 dataset

Final Exam: Friday May 1, 2pm

The final exam will be open book and take home. I will provide more information about the scope in the coming weeks.

Lecture: Learning

Readings: chapters 18-20

2019 Scassellati Slides:

Stuart Russell, Berkeley Slides:

AIMA chapter 18 6up Learning from Observations
AIMA chapter 20a 6up Statistical Learning
AIMA chapter 20b 6up Neural Networks

0304.ipynb 0304.html

0306.ipynb 0306.html

Visualization
distance functions
plurality learner classifier
k-nearest neighbors classifier
decision tree learner
random forest learner

0323.ipynb 0323.html

naive Bayes learner.
Perceptron Classifier
Linear Learner
Ensemble Learner
Learner Evaluation
k-Nearest Neighbor Example
AdaBoost

Machine Learning using scikit learn sklearn.html

sklcompare.py compare various sci kit learn algorithms. Included in above notebook.
Question about white areas in sklcompare plots: Uses RdBu Diverging colormap There is a spectrum from red to Blue, with white in the middle.

learning_apps.ipynb learning_apps.html Continue with MNIST PL.

Modified NIST. Orginal training data from US census and testing data from US high school students. "Re-mixed".

0401.ipynb 0401.html Statistics / Naive Bayes. We (naively) assume that the features are independent to make computations easier.

Toward ML-Centric Cloud Platforms CACM February, 2020. Microsoft Azure guys. Focus is on the engineering aspects of the cloud, not the content. Sigh.

There are many potential uses of ML in cloud computing platforms. The challenge is in defining exactly how and where ML should be infused in these platforms.
Leveraging ML-derived predictions has shown promise for many resource managers in Azure Compute. Having a general and independent ML framework/ system has been key to increasing adoption quickly.
Many research challenges remain open, including how to make action prescribing ML general enough for wide applicability in cloud platforms, how to manage (potentially partial) feedback at scale, and how to debug misbehaviors (especially when the ML is tightly integrated with resource managers).

0403.ipynb 0403.html Naive Bayes, Perceptron, Linear, Adaboost.

0406.ipynb 0406.html Ensemble Learning, Adaboost.

0406nn.ipynb 0406nn.html neural networks - annotated

neural_nets.ipynb neural_nets.html

learning.ipynb learning.html

Jupyter notebook keyboard shortcuts

Lecture: Reinforcement Learning

Readings: chapter 21

2019 Scassellati Slides:

rl.html

[Home]