Skip to content

Commit

Permalink
added last lecture
Browse files Browse the repository at this point in the history
  • Loading branch information
cmry committed Oct 19, 2016
1 parent 12c04d5 commit 766c10c
Showing 1 changed file with 271 additions and 0 deletions.
271 changes: 271 additions & 0 deletions Week 8 - Applications/rec_slides.ipynb
Original file line number Diff line number Diff line change
@@ -0,0 +1,271 @@
{
"cells": [
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"collapsed": true,
"slideshow": {
"slide_type": "skip"
}
},
"outputs": [],
"source": [
"from IPython.display import Image"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"# Data Mining\n",
"\n",
"# Lecture 7b: Applications\n",
"\n",
"![img](https://udemy-images.udemy.com/course/750x422/918390_02a6.jpg)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"## Why Now?"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"![img1](http://www.dvhardware.net/news/2016/mooreslaw_visualized.png)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"![img2](http://cdn.wccftech.com/wp-content/uploads/2016/02/NVIDIA-Pascal-GPU-Performance.jpg)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"![img3](http://www.kdnuggets.com/wp-content/uploads/test.jpg)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"![img4](https://cbi-blog.s3.amazonaws.com/blog/wp-content/uploads/2017/04/AI_acqui_timeline_2016April.png)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### This Part\n",
"\n",
"- Recommender Systems"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"## Recommender Systems"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### What Would You Like?\n",
"\n",
"- Facebook\n",
"- Amazon\n",
"- Spotify\n",
"- Netflix\n",
"\n",
"![img_rec2](http://image.slidesharecdn.com/shah-june27-425pm-room210a-v2-130710105227-phpapp01/95/recommender-system-at-scale-using-hbase-and-hadoop-8-638.jpg?cb=1373453680)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Collaborative Filtering\n",
"\n",
"![img_col](https://upload.wikimedia.org/wikipedia/commons/5/52/Collaborative_filtering.gif)\n",
"\n",
"Pretty standard DM task: given this history, predict recommendation for user X."
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Why Does This Work?\n",
"\n",
"- Users that like the same content are deemed to be similar.\n",
"- Content that is liked by the same kind of users is deemed to be similar.\n",
"- Content doesn't matter: books, music, pictures, etc."
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Where Does It Fail?\n",
"\n",
"- Relies on data:\n",
" - Popular items have a lot of that.\n",
" - If this $\\neq$ then that.\n",
" - Cold start problem: new user / new item.\n",
" \n",
" \n",
"![img_rec3](http://insidebigdata.com/wp-content/uploads/2014/06/Humor_recommender.jpg)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Content-based Filtering\n",
"\n",
"- Uses content or meta-data for items.\n",
" - Text from books.\n",
" - Audio from music.\n",
" - Stills from films.\n",
" - Titles, tags, actors, year, theme, etc."
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Interpeting Content: Audio Example\n",
"\n",
"By Van den Oord, Dieleman, and Schrauwe (2013). See [blog](https://benanne.github.io/2014/08/05/spotify-cnns.html).\n",
"\n",
"[![img_mus](https://benanne.github.io/images/prentje_nips.png)](https://benanne.github.io/2014/08/05/spotify-cnns.html)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Observations From the Netflix Prize\n",
"\n",
"\\$1M to beat Netflix' RMSE for predicting user's movie ratings. Intuitive solutions?\n",
"\n",
"- Standard features:\n",
" - Predict according to relative quality.\n",
" - User's tendency of voting (critical users).\n",
" - Compatibility with the movie.\n",
" \n",
" \n",
"![nf_prize](http://4.bp.blogspot.com/-22y2c1qT3CA/T3-aKF-i6pI/AAAAAAAAAO0/nCJZ2OotiRw/s1600/NetflixPrize.png)"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"- Time-based features:\n",
" - People who rate in batch tend to rate older movies.\n",
" - Different rating criteria (nostalgia).\n",
" - Movies sometimes don't age well.\n",
" - Present time also plays a role.\n",
" - Same movies rated differently friday vs. monday.\n",
" - Moods are up on sundays, raising ratings."
]
}
],
"metadata": {
"celltoolbar": "Slideshow",
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.5.2"
}
},
"nbformat": 4,
"nbformat_minor": 1
}

0 comments on commit 766c10c

Please sign in to comment.