introducing machine learning

Post on 22-Mar-2017

49 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

MACHINE LEARNING

Machine can do what we do.

What is Machine Learning?A method of data analysis that automates analytical model building.

Using algorithms that iteratively learn from data.

In short, Machine Learning is make the machine learn something itself.

Applications of Machine Learning Face Detection used in Facebook and Coursera.

Language Understanding used in Slackbot.

Recommendations used in Netflix and Spotify.

Google and Baidu self driving car.

How it works?Typically classified into three boards categories:a. Supervised Learning

b. Unsupervised Learning

Supervised Learning “Right answer” given as final result - an input where the desired output is known.

Regression Problem - Predict continuous valued output

Supervised Learning “Right answer” given as final result - an input where the desired output is known.

Classification Problem

Supervised Learning Algorithms of supervised learning :

Logistic Regression

Linear Regression

Neural Network

ANOVA Quadratic Classifier

Decision Tree

Support Vector

Machine

Unsupervised Learning Used against data that has no historical labels. The algorithm must figure out what is being shown.

Clustering Problem

The teamAnswer the question, “Why are we the ones to solve the problem we identified?”

Market Segmentation

Social Network Analysis

News Summarisation

Recommender Systems

Unsupervised Learning Applications of unsupervised learning :

Unsupervised Learning Algorithms of unsupervised learning :

Hierarchical Clustering

K-mean Clustering

Gaussian Mixture Models

Hidden Markov Model

Quadratic Classifier

Decision Tree

Self-organizing Mapping

Recommender SystemsExample: Netflix Movie Recommendation

Alice (1) Brandon (2) Carmen (3) Dave (4)

Star War 1 1 5 4

Captain America

1 1 5 ?

Kung Fu Panda

? 4 1 ?

Zootopia 5 5 1 1

Lilo & Stitch 5 ? ? 1

nm = no. of

moviesnu = no. of users

r ( i , j ) = 1, if user j has rated movie iy ( i , j ) = rating given by user j to movie i

Recommender SystemsExample: Netflix Movie Recommendation

Alice (1) Brandon (2)

Carmen (3) Dave (4) Action (x1) Cartoon (x2)

Star War 1 1 5 4 1 0

Captain America

1 1 5 ? 1 0

Kung Fu Panda

? 4 1 ? 0 1

Zootopia 5 5 1 1 0 1

Lilo & Stitch 4 ? ? 1 0 1

Movie Features

Recommender SystemsProblem Formulation

r (i,j) = 1 if user j has rated movie iy (i,j) = rating given by user j to movie i

𝛳(j) = parameter vector for user jx(i) = feature vector for movie iFor user j and movie i, predicted rating = (𝛳(j))T(x(i))

To learn 𝛳(j):

Recommender SystemsProblem Formulation

To learn 𝛳(j) parameter for user j:

To learn 𝛳(1), 𝛳(1), … , 𝛳(nu):

Recommender SystemsCollaborative Filtering

Alice (1) Brandon (2)

Carmen (3) Dave (4) Action (x1) Cartoon (x2)

Star War 1 1 5 4 ? ?

Captain America

1 1 5 ? ? ?

Kung Fu Panda

? 4 1 ? ? ?

Zootopia 5 5 1 1 ? ?

Lilo & Stitch 4 ? ? 1 ? ?(𝛳(1))Tx(1) = 1 (𝛳(2))Tx(1)

= 1(𝛳(3))Tx(1) = 5 (𝛳(4))Tx(1)

= 4

Recommender SystemsProblem Formulation

Given 𝛳(1), 𝛳(2), … , 𝛳(nu), to learn x(i):

Given 𝛳(1), 𝛳(2), … , 𝛳(nu), to learn x(1), x(2), … x(nm), :

Recommender SystemsCollaborative Filtering

Given x(1), … x(nm) and movies rating, can estimate 𝛳(1), … , 𝛳(nu)Given 𝛳(1), … , 𝛳(nu), can estimate x(1), … x(nm)

Guess 𝛳 x 𝛳 x 𝛳 ...

How it works

Step 1Initialize x or with 𝛳small random values.

Step 2Minimize J(x, ) for every 𝛳 i = 1, …, nm and j = 1, …, nu

Step 3Predict movie rating with 𝛳Tx

How to get similar movies?Action (x1) Cartoon (x2)

Star War (x(1)) 1 0

Captain America (x(2))

1 0

Kung Fu Panda (x(3))

0 1

Zootopia (x(4)) 0 1

Lilo & Stitch (x(5)) 0 1Minimum value of | x(5) - x(i) |, i

from 1 - 4

Q & A

top related