1 WHAT MOVIE TO WATCH TONIGHT? AZURE ML BY MSFT CC AUGUST 5, 2015.

Post on 17-Jan-2016

212 views 0 download

Tags:

Transcript of 1 WHAT MOVIE TO WATCH TONIGHT? AZURE ML BY MSFT CC AUGUST 5, 2015.

1

WHAT MOVIE TO WATCH TONIGHT?

AZURE ML BY MSFT CC

AUGUST 5, 2015

2

ABOUT ME

PHOTO Pavel Golub Solution Architect at EPAM Microsoft Competency Center IT professional with 11+ years of experience. I build applications of enterprise level primarily in Financial Services business domain based on .NET platform.

At spare time Windows Phone enthusiast with several released applications.

3

AGENDA

Introduction to ML theory1

ML platforms overview2

Problem definition3

Problem solving4

Q&A5

4

SPOT THE SPOTTED CAT

5

Out of the box ML solutions

• Azure ML

• Amazon ML

• MetaMind

• IBM Watson

• etc.

6

MetaMind

7

“A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E” - Tom M. Mitchell

WHAT IS ML?

DEFENITION

DEFENITION FOR HUMANS

“The goal of machine learning is to program computers to use example data or past experience to solve a given problem.” – Introduction to Machine Learning, 2nd Edition, MIT Press

8

Poisonous or Edible?

WHAT ML CAN BE USED FOR?

CLASSIFICATION PREDICTION

Pavel Golub
Add Twitter Sentiment example, e.g. from here
Pavel Golub
or diabetic https://healthmind.io/

9

PRIMARY TECHNIQUES: SUPERVISED LEARNING

Used when you want to predict unknown answers from answers you already have

6. Select the best model

5. Evaluate results using

test set

Go to 3?

4. Apply algorithm(s) on data set

3. Build model

2. Split into training and

test sets

1. Clean up and

supplement data

0. Data

10

SUPERVISED LEARNING ON PRACTICE

• Two legs, two hands

• Commonly two eyes, but some of them have one only

• Wear safety glasses

• Yellow skin

• No neck

HOW TO RECOGNIZE A MINION?

OK

OK

11

PRIMARY TECHNIQUES: UNSUPERVISED LEARNING

1. Used when you want to find unknown answers – mostly groupings - directly from data

2. No simple way to evaluate accuracy of what you learn

3. Evaluates more vectors, groups into sets or classifications

Evaluate groups

Apply algorithm(s)Data

12

UNSUPERVISED LEARNING ON PRACTICE

13

AZURE ML STUDIO OVERVIEW

DEMO

14

PROBLEM DEFINITION

MOVIE RECOMMENDATION SYSTEM

?

15

PROBLEM SOLVING

DEMO

17

PROBLEM SOLVING RESULTS

MOVIE RECOMMENDATION SYSTEM

19

Q&A

Q&A