42
Black Magic How to apply ML to real-world problems

Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Embed Size (px)

Citation preview

Page 1: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Black MagicHow to apply ML to real-world problems

Page 2: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

It is great tool for some purposes

ML is (Magic) Hammer

Page 3: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

“If all you have is a hammer, everything looks like a nail

Page 4: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems
Page 5: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

I am Evion KimLead Machine Learning Engineer @ Mattermark

Senior Software Engineer /Data Scientist @ Linkedin

M.S. , Computer Science @ Stanford University

B.S., Computer Science @ KAIST

Hello!

Page 6: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Today’s Talk◇ Machine Learning - the concept◇ Mattermark?◇ Funding Extraction Problem @ Mattermark

& Some Magic Spells

Page 7: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Not about

# Deep academic technical knowledge about ML algorithms

# Data Infrastructure

About

# How to transform real-world problem into ML-problem

# Tips and tricks on Machine Learning based problem solving

This talk is...

Page 8: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Machine LearningThe powerful hammer

1

Page 9: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems
Page 10: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

def traditional(x): return x*(x+1)

Traditional wayy = x * (x+1)

Page 11: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

2 = 63 = 124 = 205 = 306 = 42

ML Wayy = x * (x+1)

Model

Page 12: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems
Page 13: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Data

DEEP LEARNING?Trained Model

Page 14: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

It is not SKYNET… at least not “yet”.

Page 15: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

It is toolthat can be used for some problems

Page 16: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

What is(just quick advertisement)

2

Page 17: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems
Page 18: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems
Page 19: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems
Page 20: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems
Page 21: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems
Page 22: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Case Study: Funding Extraction-And Dark Magic spells we learned

3

Page 23: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Small(er) company

# Much smaller training data points

# Very high precision requirement.

Big(ger) company

# Millions of Millions of training data points

# Precision requirement: not that high

ML @ Big(ger) vs. Small(er)

Page 24: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

What’s the bottleneck?

# Scalability or Accuracy?

# Precision or Recall?

# Engineering or Machine Learning?

Spell 1: Know your Enemy

Page 25: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

~$156 BillionTotal VC funding in year 2015

~8,532VC funding events in year 2015

Page 26: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Problem to solve

Page 27: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Divide big chunky problem into smaller ML-solvable problems.

Spell 2: Slice and Dice

Page 28: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Smaller Problems

Classify Funding Articles

Classify Funding

Sentences

Extract Funding Entities

Confidence Scorer

Page 29: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Classify Funding Article

TF-IDF + SVM Classifier

NO

YES

Page 30: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Analyze and understand the problem space you are working on.

Spell 3: Understand your Domain

Page 31: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Amount/Series/Investors

...has closed a $3.5m Series A funding round led by Inter Capital, ...

Investors

Intel Capital led the round with participation from other investors that included Horizons Ventures

Amount/ Series

...has raised $3.5 million in Series A Funding

Funding Sentences Patterns

Page 32: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Classify Funding Sentences

Word2Vec

+ Semantic Role Labeling (SRL)

+ Gradient Boosting Classsifier

Page 33: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Regex Parsing

+ Named Entity Recognition

Extract Funding Entities

Page 34: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Spell 4: ProbabilisticTrain and use the probabilistic models helps a lot sometimes.

Page 35: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

“What’s the probability of these extracted information to be correct?”

Confidence Scoring

0~1 probability score

Page 36: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Spell 5: Human + MachineLet some part of the job get the help from mighty human-being

Page 37: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Human Administration

Page 38: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

~$156 BillionTotal VC funding in year 2015

~8,532VC funding events in year 2015

Page 39: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Spell 5: Human + MachineSpell 4: ProbabilisticSpell 3: Understand your domain

Spell 2: Slice and DiceSpell 1: Know your enemyML is powerful Hammer

Summary

Page 40: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

We are hiring!

Page 41: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems
Page 42: Data By The Bay 2016 - Black Magic: How to apply Machine Learning to real-world problems

Thanks!Any questions?You can find me at:

◇ in/evionkim◇ twitter@evion12◇ [email protected]