Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Rage Against The Learning of the Machine
Search
Errazudin Ishak
August 27, 2017
Technology
0
200
Rage Against The Learning of the Machine
Talk presented at Pycon APAC 2017, Kuala Lumpur, Malaysia.
Errazudin Ishak
August 27, 2017
Tweet
Share
More Decks by Errazudin Ishak
See All by Errazudin Ishak
The Spock Guide To Think Out of The Vagrant Box
errazudin
0
110
Develop and Deploy your Mobile API with Ruby on Rails, Nginx, Unicorn and Capistrano
errazudin
1
600
Rediscover Speed with Redis(and PHP)
errazudin
1
290
Other Decks in Technology
See All in Technology
【TiDB GAME DAY 2025】Shadowverse: Worlds Beyond にみる TiDB 活用術
cygames
0
1.1k
エンジニア向け技術スタック情報
kauche
1
290
マーケットプレイス版Oracle WebCenter Content For OCI
oracle4engineer
PRO
3
910
生成AIで小説を書くためにプロンプトの制約や原則について学ぶ / prompt-engineering-for-ai-fiction
nwiizo
4
2.6k
使いたいMCPサーバーはWeb APIをラップして自分で作る #QiitaBash
bengo4com
0
370
Кто отправит outbox? Валентин Удальцов, автор канала Пых
lamodatech
0
360
本が全く読めなかった過去の自分へ
genshun9
0
590
Snowflake Summit 2025 データエンジニアリング関連新機能紹介 / Snowflake Summit 2025 What's New about Data Engineering
tiltmax3
0
320
Model Mondays S2E03: SLMs & Reasoning
nitya
0
100
Clineを含めたAIエージェントを 大規模組織に導入し、投資対効果を考える / Introducing AI agents into your organization
i35_267
4
1.7k
25分で解説する「最小権限の原則」を実現するための AWS「ポリシー」大全 / 20250625-aws-summit-aws-policy
opelab
9
1.2k
GitHub Copilot の概要
tomokusaba
1
140
Featured
See All Featured
"I'm Feeling Lucky" - Building Great Search Experiences for Today's Users (#IAC19)
danielanewman
229
22k
CSS Pre-Processors: Stylus, Less & Sass
bermonpainter
357
30k
BBQ
matthewcrist
89
9.7k
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
507
140k
What’s in a name? Adding method to the madness
productmarketing
PRO
23
3.5k
GitHub's CSS Performance
jonrohan
1031
460k
Designing for humans not robots
tammielis
253
25k
The Web Performance Landscape in 2024 [PerfNow 2024]
tammyeverts
8
670
Why You Should Never Use an ORM
jnunemaker
PRO
58
9.4k
GraphQLの誤解/rethinking-graphql
sonatard
71
11k
jQuery: Nuts, Bolts and Bling
dougneiner
63
7.8k
Building Applications with DynamoDB
mza
95
6.5k
Transcript
RAGE AGAINST THE LEARNING OF THE MACHINE ERRAZUDIN ISHAK
PYCON APAC 2017
AGENDA ABOUT ME WHAT ON EARTH FOR WHAT REASON SO
HOW TO DO THAT SUMMARY PYCON APAC 2017
ABOUT ME Data Masseuse Solutions Architect DevOps Freak Bitcoin Farmer
:) PYCON APAC 2017
I WAS HERE 2009: foss.my, MyGOSSCON 2010: PHP North West
(UK), Entp. PHP Techtalk, BarcampKL, MOSC.my, MyGOSSCON 2011: Wordpress Conf. Asia, Joomla! Day KL, MOSC.my, OWASP Day KL PYCON APAC 2017
I WAS HERE 2012: OWASP AppSec APAC (Sydney), MOSC.my 2013:
OSDC (Auckland), MOSC.my 2016: SCM Workshop UMP PYCON APAC 2017
WHAT ON EARTH? PYCON APAC 2017
WHAT ON EARTH? PYCON APAC 2017 ML
–Tom M.Mitchell, CMU “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.” TYPICAL EXPLANATION…..
WHAT ON EARTH? PYCON APAC 2017 Source : NVIDIA
FOR WHAT REASON PYCON APAC 2017 “…it is now the
golden age of Machine Learning” –Random guy “… because big guys (Google and Facebook) work on it” –Another random guy
FOR WHAT REASON PYCON APAC 2017
FOR WHAT REASON Web Search & Recommendation Engines Finance :
Stock, Fraud, Credit Check Healthcare : Drug Discovery, Computational Biology Text, Speech, Object Recognition Space, Astronomy PYCON APAC 2017
FOR WHAT REASON PYCON APAC 2017 “Drawing lines through data”
FOR WHAT REASON PYCON APAC 2017
FOR WHAT REASON PYCON APAC 2017 Classification : “Draw lines
to separate data” Source : ML Berkeley Labelled Data Decision Boundary (D.B.) More complicated algo, More complicated D.B. FOR WHAT REASON
FOR WHAT REASON PYCON APAC 2017 Regression : “Draw lines
to describe data” Source : ML Berkeley Labelled Data Probability Predictor FOR WHAT REASON
FOR WHAT REASON PYCON APAC 2017 Source : Brown EDU
SO HOW TO DO THAT Formulate the problem Design the
solution Bring up the data Technology to master Build ML model Evaluate, fine tune the quality Package it nicely PYCON APAC 2017
FORMULATE YOUR PROBLEM What : Describe it Why : Benefits
How : The flow (step-by-step) PYCON APAC 2017
BRING UP THE DATA Prepare (the right) Data Identify Outliers
Data Pre-Processing PYCON APAC 2017
TECHNOLOGIES “Right tools for the right job” PYCON APAC 2017
BUILD THE MODEL PYCON APAC 2017 The most challenging part
Build, Train, Test, Repeat
FINE TUNING Test harness Measuring the performance Datasets (Test, Training)
PYCON APAC 2017
FINE TUNING “If You Knew Which Algorithm or Algorithm Configuration
To Use, You Would Not Need To Use Machine Learning” - Jason Brownlee, PhD PYCON APAC 2017
PRESENTATION PYCON APAC 2017
SAMPLE #1 PYCON APAC 2017
FORMULATE YOUR PROBLEM PYCON APAC 2017 Toyota’s stock price on
January 6th 2017
FORMULATE YOUR PROBLEM PYCON APAC 2017 Ford’s stock price on
January 4th 2017
DESIGN THE SOLUTION PYCON APAC 2017 trump2cash Python Google Cloud
Natural Language API Wikidata Query Service Tradeking API
BUILD (PLAY WITH) THE MODEL PYCON APAC 2017
PRESENTATION PYCON APAC 2017
PRESENTATION PYCON APAC 2017
SAMPLE #2 : SPAM DETECTION PYCON APAC 2017 ML problem:
text classification Algorithms: naive bayes, linear classifiers, tree classifiers, all-you-want classifiers Technologies: sklearn, nltk, scrapy Data: sms spam dataset, e-mail spam dataset , youtube comments spam dataset
SUMMARY PYCON APAC 2017 Source : Google Cloud Next 2017
THANK YOU We’re Hiring
[email protected]
PYCON APAC 2017