- PyCon Machine Learning
- Kyle Kastner - Machine Learning 101 - PyCon 2015
Click on text below to jump to the corresponding location in the video (except iOS)
My name is Kyle Kastner. I am here to talk about the basics of machine learning.
What is machine learning?I break it into two topics - automation and data analysis. The two approaches you see are people who want to automate - and people who want to inspect their processes and figure out what is making them tick and how to be more efficient and productive.
ApplicationsThe applications for machine learning have exploded in the last 5 years. Hardware has caught up with ideas from the 80s and the future is very bright.
Natural language processing
Automation spectrum-hand crafted rules
-machine learning proper
A testClassifying a point
The manifold hypothesis
ClassificationDrawing a boundary in some space.
RegressionYou are trying to predict what something will do - not merely classifying into a group of labels
Learning functionsThe core of machine learning is about learning functions.
Train - valid - testOn the left we have the red data - on the right blue data. You split a chunk of the red data and say it is the fake information I am going to test against. We really want our systems to work on new data.
What should I use?Anaconda and Canopy. They have all the packages installed for you already.
ExamplesRecommender systemsThe goal is - people rate some stuff and not other stuff. The goal is to automatically rate the stuff which has not been rated.
The example goes through a list of jokes.
Tip: Use pandas to read Excel files.
I implemented an algorithm called probabilistic matrix factorization.
Validation mean absolute error definition.
What the Recommender system reconstructs. The matrices tell us something - users are not completely unique. You can group them together.
Spam classificationWe use tf-idf vectorizer and bernoulli naive bayes.
DigitsShows the manifold structure we talked about.
Object detectionThis uses deep neural networks. Video outline created using VideoJots. Click and drag lower right corner to resize video. On iOS devices you cannot jump to video location by clicking on the outline.