SlideShare a Scribd company logo
1 of 48
Download to read offline
10MoreLessons
Learned from building real-life Machine Learning Systems
Xavier Amatriain (@xamat) 10/13/2015
Machine Learning
@Quora
Our Mission
โ€œTo share and grow the worldโ€™s knowledgeโ€
โ— Millions of questions & answers
โ— Millions of users
โ— Thousands of topics
โ— ...
Demand
What we care about
Quality
Relevance
Lots of data relations
ML Applications @ Quora
โ— Answer ranking
โ— Feed ranking
โ— Topic recommendations
โ— User recommendations
โ— Email digest
โ— Ask2Answer
โ— Duplicate Questions
โ— Related Questions
โ— Spam/moderation
โ— Trending now
โ— ...
Models
โ— Logistic Regression
โ— Elastic Nets
โ— Gradient Boosted Decision
Trees
โ— Random Forests
โ— (Deep) Neural Networks
โ— LambdaMART
โ— Matrix Factorization
โ— LDA
โ— ...
10MoreLessons
Learned from implementing real-life ML systems
1.Implicitsignalsbeat
explicitones
(almostalways)
Implicit vs. Explicit
โ— Many have acknowledged
that implicit feedback is more useful
โ— Is implicit feedback really always
more useful?
โ— If so, why?
โ— Implicit data is (usually):
โ—‹ More dense, and available for all users
โ—‹ Better representative of user behavior vs.
user reflection
โ—‹ More related to final objective function
โ—‹ Better correlated with AB test results
โ— E.g. Rating vs watching
Implicit vs. Explicit
โ— However
โ—‹ It is not always the case that
direct implicit feedback correlates
well with long-term retention
โ—‹ E.g. clickbait
โ— Solution:
โ—‹ Combine different forms of
implicit + explicit to better represent
long-term goal
Implicit vs. Explicit
2.YourModelwilllearn
whatyouteachittolearn
Training a model
โ— Model will learn according to:
โ—‹ Training data (e.g. implicit and explicit)
โ—‹ Target function (e.g. probability of user reading an answer)
โ—‹ Metric (e.g. precision vs. recall)
โ— Example 1 (made up):
โ—‹ Optimize probability of a user going to the cinema to
watch a movie and rate it โ€œhighlyโ€ by using purchase history
and previous ratings. Use NDCG of the ranking as final
metric using only movies rated 4 or higher as positives.
Example 2 - Quoraโ€™s feed
โ— Training data = implicit + explicit
โ— Target function: Value of showing a story to a
user ~ weighted sum of actions: v = โˆ‘a
va
1{ya
= 1}
โ—‹ predict probabilities for each action, then compute expected
value: v_pred = E[ V | x ] = โˆ‘a
va
p(a | x)
โ— Metric: any ranking metric
3.Supervisedvs.plus
UnsupervisedLearning
Supervised/Unsupervised Learning
โ— Unsupervised learning as dimensionality reduction
โ— Unsupervised learning as feature engineering
โ— The โ€œmagicโ€ behind combining
unsupervised/supervised learning
โ—‹ E.g.1 clustering + knn
โ—‹ E.g.2 Matrix Factorization
โ–  MF can be interpreted as
โ— Unsupervised:
โ—‹ Dimensionality Reduction a la PCA
โ—‹ Clustering (e.g. NMF)
โ— Supervised
โ—‹ Labeled targets ~ regression
Supervised/Unsupervised Learning
โ— One of the โ€œtricksโ€ in Deep Learning is how it
combines unsupervised/supervised learning
โ—‹ E.g. Stacked Autoencoders
โ—‹ E.g. training of convolutional nets
4.Everythingisanensemble
Ensembles
โ— Netflix Prize was won by an ensemble
โ—‹ Initially Bellkor was using GDBTs
โ—‹ BigChaos introduced ANN-based ensemble
โ— Most practical applications of ML run an ensemble
โ—‹ Why wouldnโ€™t you?
โ—‹ At least as good as the best of your methods
โ—‹ Can add completely different approaches (e.
g. CF and content-based)
โ—‹ You can use many different models at the
ensemble layer: LR, GDBTs, RFs, ANNs...
Ensembles & Feature Engineering
โ— Ensembles are the way to turn any model into a feature!
โ— E.g. Donโ€™t know if the way to go is to use Factorization
Machines, Tensor Factorization, or RNNs?
โ—‹ Treat each model as a โ€œfeatureโ€
โ—‹ Feed them into an ensemble
The Master Algorithm?
It definitely is an ensemble!
5.Theoutputofyourmodel
willbetheinputofanotherone
(andotherdesignproblems)
Outputs will be inputs
โ— Ensembles turn any model into a feature
โ—‹ Thatโ€™s great!
โ—‹ That can be a mess!
โ— Make sure the output of your model is ready to
accept data dependencies
โ—‹ E.g. can you easily change the distribution of the
value without affecting all other models
depending on it?
โ— Avoid feedback loops
โ— Can you treat your ML infrastructure as you would
your software one?
ML vs Software
โ— Can you treat your ML infrastructure as you would
your software one?
โ—‹ Yes and No
โ— You should apply best Software Engineering
practices (e.g. encapsulation, abstraction, cohesion,
low couplingโ€ฆ)
โ— However, Design Patterns for Machine Learning
software are not well known/documented
6.Thepains&gains
ofFeatureEngineering
Feature Engineering
โ— Main properties of a well-behaved ML feature
โ—‹ Reusable
โ—‹ Transformable
โ—‹ Interpretable
โ—‹ Reliable
โ— Reusability: You should be able to reuse features in different
models, applications, and teams
โ— Transformability: Besides directly reusing a feature, it
should be easy to use a transformation of it (e.g. log(f), max(f),
โˆ‘ft
over a time windowโ€ฆ)
Feature Engineering
โ— Main properties of a well-behaved ML feature
โ—‹ Reusable
โ—‹ Transformable
โ—‹ Interpretable
โ—‹ Reliable
โ— Interpretability: In order to do any of the previous, you
need to be able to understand the meaning of features and
interpret their values.
โ— Reliability: It should be easy to monitor and detect bugs/issues
in features
Feature Engineering Example - Quora Answer Ranking
What is a good Quora answer?
โ€ข truthful
โ€ข reusable
โ€ข provides explanation
โ€ข well formatted
โ€ข ...
Feature Engineering Example - Quora Answer Ranking
How are those dimensions translated
into features?
โ€ข Features that relate to the answer
quality itself
โ€ข Interaction features
(upvotes/downvotes, clicks,
commentsโ€ฆ)
โ€ข User features (e.g. expertise in topic)
7.Thetwofacesofyour
MLinfrastructure
Machine Learning Infrastructure
โ— Whenever you develop any ML infrastructure, you need to
target two different modes:
โ—‹ Mode 1: ML experimentation
โ–  Flexibility
โ–  Easy-to-use
โ–  Reusability
โ—‹ Mode 2: ML production
โ–  All of the above + performance & scalability
โ— Ideally you want the two modes to be as similar as possible
โ— How to combine them?
Machine Learning Infrastructure: Experimentation & Production
โ— Option 1:
โ—‹ Favor experimentation and only invest in productionizing
once something shows results
โ—‹ E.g. Have ML researchers use R and then ask Engineers
to implement things in production when they work
โ— Option 2:
โ—‹ Favor production and have โ€œresearchersโ€ struggle to figure
out how to run experiments
โ—‹ E.g. Implement highly optimized C++ code and have ML
researchers experiment only through data available in logs/DB
Machine Learning Infrastructure: Experimentation & Production
โ— Option 1:
โ—‹ Favor experimentation and only invest in productionazing once
something shows results
โ—‹ E.g. Have ML researchers use R and then ask Engineers to
implement things in production when they work
โ— Option 2:
โ—‹ Favor production and have โ€œresearchersโ€ struggle to figure out
how to run experiments
โ—‹ E.g. Implement highly optimized C++ code and have ML
researchers experiment only through data available in logs/DB
โ— Good intermediate options:
โ—‹ Have ML โ€œresearchersโ€ experiment on iPython Notebooks using
Python tools (scikit-learn, Theanoโ€ฆ). Use same tools in
production whenever possible, implement optimized versions
only when needed.
โ—‹ Implement abstraction layers on top of optimized
implementations so they can be accessed from regular/friendly
experimentation tools
Machine Learning Infrastructure: Experimentation & Production
8.Whyyoushouldcareabout
answeringquestions(aboutyourmodel)
Model debuggability
โ— Value of a model = value it brings to the product
โ— Product owners/stakeholders have expectations on
the product
โ— It is important to answer questions to why did
something fail
โ— Bridge gap between product design and ML algos
โ— Model debuggability is so important it can
determine:
โ—‹ Particular model to use
โ—‹ Features to rely on
โ—‹ Implementation of tools
Model debuggability
โ— E.g. Why am I seeing or not seeing
this on my homepage feed?
9.Youdonโ€™tneedtodistribute
yourMLalgorithm
Distributing ML
โ— Most of what people do in practice can fit into a multi-
core machine
โ—‹ Smart data sampling
โ—‹ Offline schemes
โ—‹ Efficient parallel code
โ— Dangers of โ€œeasyโ€ distributed approaches such
as Hadoop/Spark
โ— Do you care about costs? How about latencies?
Distributing ML
โ— Example of optimizing computations to fit them into
one machine
โ—‹ Spark implementation: 6 hours, 15 machines
โ—‹ Developer time: 4 days
โ—‹ C++ implementation: 10 minutes, 1 machine
โ— Most practical applications of Big Data can fit into
a (multicore) implementation
10.Theuntoldstoryof
DataScienceandvs.MLengineering
Data Scientists and ML Engineers
โ— We all know the definition of a Data Scientist
โ— Where do Data Scientists fit in an organization?
โ—‹ Many companies struggling with this
โ— Valuable to have strong DS who can bring value
from the data
โ— Strong DS with solid engineering skills are
unicorns and finding them is not scalable
โ—‹ DS need engineers to bring things to production
โ—‹ Engineers have enough on their plate to be willing to
โ€œproductionizeโ€ cool DS projects
The data-driven ML innovation funnel
Data Research
ML Exploration -
Product Design
AB Testing
Data Scientists and ML Engineers
โ— Solution:
โ—‹ (1) Define different parts of the innovation funnel
โ–  Part 1. Data research & hypothesis
building -> Data Science
โ–  Part 2. ML solution building &
implementation -> ML Engineering
โ–  Part 3. Online experimentation, AB
Testing analysis-> Data Science
โ—‹ (2) Broaden the definition of ML Engineers
to include from coding experts with high-level
ML knowledge to ML experts with good
software skills
Data Research
ML Solution
AB Testing
Data
Science
Data
Science
ML
Engineering
Conclusions
โ— Make sure you teach your model what you
want it to learn
โ— Ensembles and the combination of
supervised/unsupervised techniques are key
in many ML applications
โ— Important to focus on feature engineering
โ— Be thoughtful about
โ—‹ your ML infrastructure/tools
โ—‹ about organizing your teams
10 more lessons learned from building Machine Learning systems

More Related Content

What's hot

Machine Learning Deep Learning AI and Data Science
Machine Learning Deep Learning AI and Data Science Machine Learning Deep Learning AI and Data Science
Machine Learning Deep Learning AI and Data Science Venkata Reddy Konasani
ย 
ChatGPT for Academic
ChatGPT for AcademicChatGPT for Academic
ChatGPT for AcademicAndry Alamsyah
ย 
OpenAIโ€™s GPT 3 Language Model - guest Steve Omohundro
OpenAIโ€™s GPT 3 Language Model - guest Steve OmohundroOpenAIโ€™s GPT 3 Language Model - guest Steve Omohundro
OpenAIโ€™s GPT 3 Language Model - guest Steve OmohundroNumenta
ย 
ChatGPT Deck.pptx
ChatGPT Deck.pptxChatGPT Deck.pptx
ChatGPT Deck.pptxomornahid1
ย 
Artificial Intelligence and Machine Learning for business
Artificial Intelligence and Machine Learning for businessArtificial Intelligence and Machine Learning for business
Artificial Intelligence and Machine Learning for businessSteven Finlay
ย 
How Does Generative AI Actually Work? (a quick semi-technical introduction to...
How Does Generative AI Actually Work? (a quick semi-technical introduction to...How Does Generative AI Actually Work? (a quick semi-technical introduction to...
How Does Generative AI Actually Work? (a quick semi-technical introduction to...ssuser4edc93
ย 
Visualising Data with Code
Visualising Data with CodeVisualising Data with Code
Visualising Data with CodeRi Liu
ย 
The Power of Auto ML and How Does it Work
The Power of Auto ML and How Does it WorkThe Power of Auto ML and How Does it Work
The Power of Auto ML and How Does it WorkIvo Andreev
ย 
Automated Machine Learning
Automated Machine LearningAutomated Machine Learning
Automated Machine LearningYuriy Guts
ย 
Blueprint ChatGPT Lunch & Learn
Blueprint ChatGPT Lunch & LearnBlueprint ChatGPT Lunch & Learn
Blueprint ChatGPT Lunch & Learngnakan
ย 
Introduction to Data Science
Introduction to Data ScienceIntroduction to Data Science
Introduction to Data ScienceNiko Vuokko
ย 
AI 2023.pdf
AI 2023.pdfAI 2023.pdf
AI 2023.pdfDavidCieslak4
ย 
Introduction to data science.pptx
Introduction to data science.pptxIntroduction to data science.pptx
Introduction to data science.pptxSadhanaParameswaran
ย 
Llama-index
Llama-indexLlama-index
Llama-indexDenis973830
ย 
Episode 2: The LLM / GPT / AI Prompt / Data Engineer Roadmap
Episode 2: The LLM / GPT / AI Prompt / Data Engineer RoadmapEpisode 2: The LLM / GPT / AI Prompt / Data Engineer Roadmap
Episode 2: The LLM / GPT / AI Prompt / Data Engineer RoadmapAnant Corporation
ย 
Uses of AI text bot.pdf
Uses of AI text bot.pdfUses of AI text bot.pdf
Uses of AI text bot.pdfSreeNivas983124
ย 
Lesson 1 intro to ai
Lesson 1   intro to aiLesson 1   intro to ai
Lesson 1 intro to aiankit_ppt
ย 
Generative AI for the rest of us
Generative AI for the rest of usGenerative AI for the rest of us
Generative AI for the rest of usMassimo Ferre'
ย 

What's hot (20)

Machine Learning Deep Learning AI and Data Science
Machine Learning Deep Learning AI and Data Science Machine Learning Deep Learning AI and Data Science
Machine Learning Deep Learning AI and Data Science
ย 
ChatGPT for Academic
ChatGPT for AcademicChatGPT for Academic
ChatGPT for Academic
ย 
OpenAIโ€™s GPT 3 Language Model - guest Steve Omohundro
OpenAIโ€™s GPT 3 Language Model - guest Steve OmohundroOpenAIโ€™s GPT 3 Language Model - guest Steve Omohundro
OpenAIโ€™s GPT 3 Language Model - guest Steve Omohundro
ย 
ChatGPT Deck.pptx
ChatGPT Deck.pptxChatGPT Deck.pptx
ChatGPT Deck.pptx
ย 
Artificial Intelligence and Machine Learning for business
Artificial Intelligence and Machine Learning for businessArtificial Intelligence and Machine Learning for business
Artificial Intelligence and Machine Learning for business
ย 
How Does Generative AI Actually Work? (a quick semi-technical introduction to...
How Does Generative AI Actually Work? (a quick semi-technical introduction to...How Does Generative AI Actually Work? (a quick semi-technical introduction to...
How Does Generative AI Actually Work? (a quick semi-technical introduction to...
ย 
Python for Data Science
Python for Data SciencePython for Data Science
Python for Data Science
ย 
Visualising Data with Code
Visualising Data with CodeVisualising Data with Code
Visualising Data with Code
ย 
The Power of Auto ML and How Does it Work
The Power of Auto ML and How Does it WorkThe Power of Auto ML and How Does it Work
The Power of Auto ML and How Does it Work
ย 
Automated Machine Learning
Automated Machine LearningAutomated Machine Learning
Automated Machine Learning
ย 
Blueprint ChatGPT Lunch & Learn
Blueprint ChatGPT Lunch & LearnBlueprint ChatGPT Lunch & Learn
Blueprint ChatGPT Lunch & Learn
ย 
Introduction to Data Science
Introduction to Data ScienceIntroduction to Data Science
Introduction to Data Science
ย 
AI 2023.pdf
AI 2023.pdfAI 2023.pdf
AI 2023.pdf
ย 
Introduction to data science.pptx
Introduction to data science.pptxIntroduction to data science.pptx
Introduction to data science.pptx
ย 
Artificial Intelligence (AI) in media applications and services
Artificial Intelligence (AI) in media applications and servicesArtificial Intelligence (AI) in media applications and services
Artificial Intelligence (AI) in media applications and services
ย 
Llama-index
Llama-indexLlama-index
Llama-index
ย 
Episode 2: The LLM / GPT / AI Prompt / Data Engineer Roadmap
Episode 2: The LLM / GPT / AI Prompt / Data Engineer RoadmapEpisode 2: The LLM / GPT / AI Prompt / Data Engineer Roadmap
Episode 2: The LLM / GPT / AI Prompt / Data Engineer Roadmap
ย 
Uses of AI text bot.pdf
Uses of AI text bot.pdfUses of AI text bot.pdf
Uses of AI text bot.pdf
ย 
Lesson 1 intro to ai
Lesson 1   intro to aiLesson 1   intro to ai
Lesson 1 intro to ai
ย 
Generative AI for the rest of us
Generative AI for the rest of usGenerative AI for the rest of us
Generative AI for the rest of us
ย 

Similar to 10 more lessons learned from building Machine Learning systems

Strata 2016 - Lessons Learned from building real-life Machine Learning Systems
Strata 2016 -  Lessons Learned from building real-life Machine Learning SystemsStrata 2016 -  Lessons Learned from building real-life Machine Learning Systems
Strata 2016 - Lessons Learned from building real-life Machine Learning SystemsXavier Amatriain
ย 
Scaling Recommendations at Quora (RecSys talk 9/16/2016)
Scaling Recommendations at Quora (RecSys talk 9/16/2016)Scaling Recommendations at Quora (RecSys talk 9/16/2016)
Scaling Recommendations at Quora (RecSys talk 9/16/2016)Nikhil Dandekar
ย 
BIG2016- Lessons Learned from building real-life user-focused Big Data systems
BIG2016- Lessons Learned from building real-life user-focused Big Data systemsBIG2016- Lessons Learned from building real-life user-focused Big Data systems
BIG2016- Lessons Learned from building real-life user-focused Big Data systemsXavier Amatriain
ย 
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...Xavier Amatriain
ย 
Staying Shallow & Lean in a Deep Learning World
Staying Shallow & Lean in a Deep Learning WorldStaying Shallow & Lean in a Deep Learning World
Staying Shallow & Lean in a Deep Learning WorldXavier Amatriain
ย 
AI hype or reality
AI  hype or realityAI  hype or reality
AI hype or realityAwantik Das
ย 
General introduction to AI ML DL DS
General introduction to AI ML DL DSGeneral introduction to AI ML DL DS
General introduction to AI ML DL DSRoopesh Kohad
ย 
Production-Ready BIG ML Workflows - from zero to hero
Production-Ready BIG ML Workflows - from zero to heroProduction-Ready BIG ML Workflows - from zero to hero
Production-Ready BIG ML Workflows - from zero to heroDaniel Marcous
ย 
Machine Learning: Artificial Intelligence isn't just a Science Fiction topic
Machine Learning: Artificial Intelligence isn't just a Science Fiction topicMachine Learning: Artificial Intelligence isn't just a Science Fiction topic
Machine Learning: Artificial Intelligence isn't just a Science Fiction topicRaรบl Garreta
ย 
L15.pptx
L15.pptxL15.pptx
L15.pptxImonBennett
ย 
Machine Learning: Inteligencia Artificial no es sรณlo un tema de Ciencia Ficci...
Machine Learning: Inteligencia Artificial no es sรณlo un tema de Ciencia Ficci...Machine Learning: Inteligencia Artificial no es sรณlo un tema de Ciencia Ficci...
Machine Learning: Inteligencia Artificial no es sรณlo un tema de Ciencia Ficci....NET Conf UY
ย 
ML Platform Q1 Meetup: Airbnb's End-to-End Machine Learning Infrastructure
ML Platform Q1 Meetup: Airbnb's End-to-End Machine Learning InfrastructureML Platform Q1 Meetup: Airbnb's End-to-End Machine Learning Infrastructure
ML Platform Q1 Meetup: Airbnb's End-to-End Machine Learning InfrastructureFei Chen
ย 
Day 2 (Lecture 5): A Practitioner's Perspective on Building Machine Product i...
Day 2 (Lecture 5): A Practitioner's Perspective on Building Machine Product i...Day 2 (Lecture 5): A Practitioner's Perspective on Building Machine Product i...
Day 2 (Lecture 5): A Practitioner's Perspective on Building Machine Product i...Aseda Owusua Addai-Deseh
ย 
Open source ml systems that need to be built
Open source ml systems that need to be builtOpen source ml systems that need to be built
Open source ml systems that need to be builtNikhil Garg
ย 
Nikhil Garg, Engineering Manager, Quora at MLconf SF 2016
Nikhil Garg, Engineering Manager, Quora at MLconf SF 2016Nikhil Garg, Engineering Manager, Quora at MLconf SF 2016
Nikhil Garg, Engineering Manager, Quora at MLconf SF 2016MLconf
ย 
Building A Machine Learning Platform At Quora (1)
Building A Machine Learning Platform At Quora (1)Building A Machine Learning Platform At Quora (1)
Building A Machine Learning Platform At Quora (1)Nikhil Garg
ย 
Prototyping Workshop - Wireframes, Mockups, Prototypes
Prototyping Workshop - Wireframes, Mockups, PrototypesPrototyping Workshop - Wireframes, Mockups, Prototypes
Prototyping Workshop - Wireframes, Mockups, PrototypesMarta Soncodi
ย 
Data science
Data scienceData science
Data sciencePurna Chander
ย 
Apache Cassandra Lunch #54: Machine Learning with Spark + Cassandra Part 2
Apache Cassandra Lunch #54: Machine Learning with Spark + Cassandra Part 2Apache Cassandra Lunch #54: Machine Learning with Spark + Cassandra Part 2
Apache Cassandra Lunch #54: Machine Learning with Spark + Cassandra Part 2Anant Corporation
ย 
Persian MNIST in 5 Minutes
Persian MNIST in 5 MinutesPersian MNIST in 5 Minutes
Persian MNIST in 5 MinutesShahriar Yazdipour
ย 

Similar to 10 more lessons learned from building Machine Learning systems (20)

Strata 2016 - Lessons Learned from building real-life Machine Learning Systems
Strata 2016 -  Lessons Learned from building real-life Machine Learning SystemsStrata 2016 -  Lessons Learned from building real-life Machine Learning Systems
Strata 2016 - Lessons Learned from building real-life Machine Learning Systems
ย 
Scaling Recommendations at Quora (RecSys talk 9/16/2016)
Scaling Recommendations at Quora (RecSys talk 9/16/2016)Scaling Recommendations at Quora (RecSys talk 9/16/2016)
Scaling Recommendations at Quora (RecSys talk 9/16/2016)
ย 
BIG2016- Lessons Learned from building real-life user-focused Big Data systems
BIG2016- Lessons Learned from building real-life user-focused Big Data systemsBIG2016- Lessons Learned from building real-life user-focused Big Data systems
BIG2016- Lessons Learned from building real-life user-focused Big Data systems
ย 
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
Recsys 2016 tutorial: Lessons learned from building real-life recommender sys...
ย 
Staying Shallow & Lean in a Deep Learning World
Staying Shallow & Lean in a Deep Learning WorldStaying Shallow & Lean in a Deep Learning World
Staying Shallow & Lean in a Deep Learning World
ย 
AI hype or reality
AI  hype or realityAI  hype or reality
AI hype or reality
ย 
General introduction to AI ML DL DS
General introduction to AI ML DL DSGeneral introduction to AI ML DL DS
General introduction to AI ML DL DS
ย 
Production-Ready BIG ML Workflows - from zero to hero
Production-Ready BIG ML Workflows - from zero to heroProduction-Ready BIG ML Workflows - from zero to hero
Production-Ready BIG ML Workflows - from zero to hero
ย 
Machine Learning: Artificial Intelligence isn't just a Science Fiction topic
Machine Learning: Artificial Intelligence isn't just a Science Fiction topicMachine Learning: Artificial Intelligence isn't just a Science Fiction topic
Machine Learning: Artificial Intelligence isn't just a Science Fiction topic
ย 
L15.pptx
L15.pptxL15.pptx
L15.pptx
ย 
Machine Learning: Inteligencia Artificial no es sรณlo un tema de Ciencia Ficci...
Machine Learning: Inteligencia Artificial no es sรณlo un tema de Ciencia Ficci...Machine Learning: Inteligencia Artificial no es sรณlo un tema de Ciencia Ficci...
Machine Learning: Inteligencia Artificial no es sรณlo un tema de Ciencia Ficci...
ย 
ML Platform Q1 Meetup: Airbnb's End-to-End Machine Learning Infrastructure
ML Platform Q1 Meetup: Airbnb's End-to-End Machine Learning InfrastructureML Platform Q1 Meetup: Airbnb's End-to-End Machine Learning Infrastructure
ML Platform Q1 Meetup: Airbnb's End-to-End Machine Learning Infrastructure
ย 
Day 2 (Lecture 5): A Practitioner's Perspective on Building Machine Product i...
Day 2 (Lecture 5): A Practitioner's Perspective on Building Machine Product i...Day 2 (Lecture 5): A Practitioner's Perspective on Building Machine Product i...
Day 2 (Lecture 5): A Practitioner's Perspective on Building Machine Product i...
ย 
Open source ml systems that need to be built
Open source ml systems that need to be builtOpen source ml systems that need to be built
Open source ml systems that need to be built
ย 
Nikhil Garg, Engineering Manager, Quora at MLconf SF 2016
Nikhil Garg, Engineering Manager, Quora at MLconf SF 2016Nikhil Garg, Engineering Manager, Quora at MLconf SF 2016
Nikhil Garg, Engineering Manager, Quora at MLconf SF 2016
ย 
Building A Machine Learning Platform At Quora (1)
Building A Machine Learning Platform At Quora (1)Building A Machine Learning Platform At Quora (1)
Building A Machine Learning Platform At Quora (1)
ย 
Prototyping Workshop - Wireframes, Mockups, Prototypes
Prototyping Workshop - Wireframes, Mockups, PrototypesPrototyping Workshop - Wireframes, Mockups, Prototypes
Prototyping Workshop - Wireframes, Mockups, Prototypes
ย 
Data science
Data scienceData science
Data science
ย 
Apache Cassandra Lunch #54: Machine Learning with Spark + Cassandra Part 2
Apache Cassandra Lunch #54: Machine Learning with Spark + Cassandra Part 2Apache Cassandra Lunch #54: Machine Learning with Spark + Cassandra Part 2
Apache Cassandra Lunch #54: Machine Learning with Spark + Cassandra Part 2
ย 
Persian MNIST in 5 Minutes
Persian MNIST in 5 MinutesPersian MNIST in 5 Minutes
Persian MNIST in 5 Minutes
ย 

More from Xavier Amatriain

Data/AI driven product development: from video streaming to telehealth
Data/AI driven product development: from video streaming to telehealthData/AI driven product development: from video streaming to telehealth
Data/AI driven product development: from video streaming to telehealthXavier Amatriain
ย 
AI-driven product innovation: from Recommender Systems to COVID-19
AI-driven product innovation: from Recommender Systems to COVID-19AI-driven product innovation: from Recommender Systems to COVID-19
AI-driven product innovation: from Recommender Systems to COVID-19Xavier Amatriain
ย 
AI for COVID-19 - Q42020 update
AI for COVID-19 - Q42020 updateAI for COVID-19 - Q42020 update
AI for COVID-19 - Q42020 updateXavier Amatriain
ย 
AI for COVID-19: An online virtual care approach
AI for COVID-19: An online virtual care approachAI for COVID-19: An online virtual care approach
AI for COVID-19: An online virtual care approachXavier Amatriain
ย 
Lessons learned from building practical deep learning systems
Lessons learned from building practical deep learning systemsLessons learned from building practical deep learning systems
Lessons learned from building practical deep learning systemsXavier Amatriain
ย 
AI for healthcare: Scaling Access and Quality of Care for Everyone
AI for healthcare: Scaling Access and Quality of Care for EveryoneAI for healthcare: Scaling Access and Quality of Care for Everyone
AI for healthcare: Scaling Access and Quality of Care for EveryoneXavier Amatriain
ย 
Towards online universal quality healthcare through AI
Towards online universal quality healthcare through AITowards online universal quality healthcare through AI
Towards online universal quality healthcare through AIXavier Amatriain
ย 
From one to zero: Going smaller as a growth strategy
From one to zero: Going smaller as a growth strategyFrom one to zero: Going smaller as a growth strategy
From one to zero: Going smaller as a growth strategyXavier Amatriain
ย 
Learning to speak medicine
Learning to speak medicineLearning to speak medicine
Learning to speak medicineXavier Amatriain
ย 
ML to cure the world
ML to cure the worldML to cure the world
ML to cure the worldXavier Amatriain
ย 
Recommender Systems In Industry
Recommender Systems In IndustryRecommender Systems In Industry
Recommender Systems In IndustryXavier Amatriain
ย 
Medical advice as a Recommender System
Medical advice as a Recommender SystemMedical advice as a Recommender System
Medical advice as a Recommender SystemXavier Amatriain
ย 
Past present and future of Recommender Systems: an Industry Perspective
Past present and future of Recommender Systems: an Industry PerspectivePast present and future of Recommender Systems: an Industry Perspective
Past present and future of Recommender Systems: an Industry PerspectiveXavier Amatriain
ย 
Machine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora ExampleMachine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora ExampleXavier Amatriain
ย 
Past, present, and future of Recommender Systems: an industry perspective
Past, present, and future of Recommender Systems: an industry perspectivePast, present, and future of Recommender Systems: an industry perspective
Past, present, and future of Recommender Systems: an industry perspectiveXavier Amatriain
ย 
Barcelona ML Meetup - Lessons Learned
Barcelona ML Meetup - Lessons LearnedBarcelona ML Meetup - Lessons Learned
Barcelona ML Meetup - Lessons LearnedXavier Amatriain
ย 
10 more lessons learned from building Machine Learning systems - MLConf
10 more lessons learned from building Machine Learning systems - MLConf10 more lessons learned from building Machine Learning systems - MLConf
10 more lessons learned from building Machine Learning systems - MLConfXavier Amatriain
ย 
Machine Learning to Grow the World's Knowledge
Machine Learning to Grow  the World's KnowledgeMachine Learning to Grow  the World's Knowledge
Machine Learning to Grow the World's KnowledgeXavier Amatriain
ย 
MLConf Seattle 2015 - ML@Quora
MLConf Seattle 2015 - ML@QuoraMLConf Seattle 2015 - ML@Quora
MLConf Seattle 2015 - ML@QuoraXavier Amatriain
ย 
Lean DevOps - Lessons Learned from Innovation-driven Companies
Lean DevOps - Lessons Learned from Innovation-driven CompaniesLean DevOps - Lessons Learned from Innovation-driven Companies
Lean DevOps - Lessons Learned from Innovation-driven CompaniesXavier Amatriain
ย 

More from Xavier Amatriain (20)

Data/AI driven product development: from video streaming to telehealth
Data/AI driven product development: from video streaming to telehealthData/AI driven product development: from video streaming to telehealth
Data/AI driven product development: from video streaming to telehealth
ย 
AI-driven product innovation: from Recommender Systems to COVID-19
AI-driven product innovation: from Recommender Systems to COVID-19AI-driven product innovation: from Recommender Systems to COVID-19
AI-driven product innovation: from Recommender Systems to COVID-19
ย 
AI for COVID-19 - Q42020 update
AI for COVID-19 - Q42020 updateAI for COVID-19 - Q42020 update
AI for COVID-19 - Q42020 update
ย 
AI for COVID-19: An online virtual care approach
AI for COVID-19: An online virtual care approachAI for COVID-19: An online virtual care approach
AI for COVID-19: An online virtual care approach
ย 
Lessons learned from building practical deep learning systems
Lessons learned from building practical deep learning systemsLessons learned from building practical deep learning systems
Lessons learned from building practical deep learning systems
ย 
AI for healthcare: Scaling Access and Quality of Care for Everyone
AI for healthcare: Scaling Access and Quality of Care for EveryoneAI for healthcare: Scaling Access and Quality of Care for Everyone
AI for healthcare: Scaling Access and Quality of Care for Everyone
ย 
Towards online universal quality healthcare through AI
Towards online universal quality healthcare through AITowards online universal quality healthcare through AI
Towards online universal quality healthcare through AI
ย 
From one to zero: Going smaller as a growth strategy
From one to zero: Going smaller as a growth strategyFrom one to zero: Going smaller as a growth strategy
From one to zero: Going smaller as a growth strategy
ย 
Learning to speak medicine
Learning to speak medicineLearning to speak medicine
Learning to speak medicine
ย 
ML to cure the world
ML to cure the worldML to cure the world
ML to cure the world
ย 
Recommender Systems In Industry
Recommender Systems In IndustryRecommender Systems In Industry
Recommender Systems In Industry
ย 
Medical advice as a Recommender System
Medical advice as a Recommender SystemMedical advice as a Recommender System
Medical advice as a Recommender System
ย 
Past present and future of Recommender Systems: an Industry Perspective
Past present and future of Recommender Systems: an Industry PerspectivePast present and future of Recommender Systems: an Industry Perspective
Past present and future of Recommender Systems: an Industry Perspective
ย 
Machine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora ExampleMachine Learning for Q&A Sites: The Quora Example
Machine Learning for Q&A Sites: The Quora Example
ย 
Past, present, and future of Recommender Systems: an industry perspective
Past, present, and future of Recommender Systems: an industry perspectivePast, present, and future of Recommender Systems: an industry perspective
Past, present, and future of Recommender Systems: an industry perspective
ย 
Barcelona ML Meetup - Lessons Learned
Barcelona ML Meetup - Lessons LearnedBarcelona ML Meetup - Lessons Learned
Barcelona ML Meetup - Lessons Learned
ย 
10 more lessons learned from building Machine Learning systems - MLConf
10 more lessons learned from building Machine Learning systems - MLConf10 more lessons learned from building Machine Learning systems - MLConf
10 more lessons learned from building Machine Learning systems - MLConf
ย 
Machine Learning to Grow the World's Knowledge
Machine Learning to Grow  the World's KnowledgeMachine Learning to Grow  the World's Knowledge
Machine Learning to Grow the World's Knowledge
ย 
MLConf Seattle 2015 - ML@Quora
MLConf Seattle 2015 - ML@QuoraMLConf Seattle 2015 - ML@Quora
MLConf Seattle 2015 - ML@Quora
ย 
Lean DevOps - Lessons Learned from Innovation-driven Companies
Lean DevOps - Lessons Learned from Innovation-driven CompaniesLean DevOps - Lessons Learned from Innovation-driven Companies
Lean DevOps - Lessons Learned from Innovation-driven Companies
ย 

Recently uploaded

Double Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueDouble Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueBhangaleSonal
ย 
Top Rated Pune Call Girls Budhwar Peth โŸŸ 6297143586 โŸŸ Call Me For Genuine Se...
Top Rated  Pune Call Girls Budhwar Peth โŸŸ 6297143586 โŸŸ Call Me For Genuine Se...Top Rated  Pune Call Girls Budhwar Peth โŸŸ 6297143586 โŸŸ Call Me For Genuine Se...
Top Rated Pune Call Girls Budhwar Peth โŸŸ 6297143586 โŸŸ Call Me For Genuine Se...Call Girls in Nagpur High Profile
ย 
Call Now โ‰ฝ 9953056974 โ‰ผ๐Ÿ” Call Girls In New Ashok Nagar โ‰ผ๐Ÿ” Delhi door step de...
Call Now โ‰ฝ 9953056974 โ‰ผ๐Ÿ” Call Girls In New Ashok Nagar  โ‰ผ๐Ÿ” Delhi door step de...Call Now โ‰ฝ 9953056974 โ‰ผ๐Ÿ” Call Girls In New Ashok Nagar  โ‰ผ๐Ÿ” Delhi door step de...
Call Now โ‰ฝ 9953056974 โ‰ผ๐Ÿ” Call Girls In New Ashok Nagar โ‰ผ๐Ÿ” Delhi door step de...9953056974 Low Rate Call Girls In Saket, Delhi NCR
ย 
Bhosari ( Call Girls ) Pune 6297143586 Hot Model With Sexy Bhabi Ready For ...
Bhosari ( Call Girls ) Pune  6297143586  Hot Model With Sexy Bhabi Ready For ...Bhosari ( Call Girls ) Pune  6297143586  Hot Model With Sexy Bhabi Ready For ...
Bhosari ( Call Girls ) Pune 6297143586 Hot Model With Sexy Bhabi Ready For ...tanu pandey
ย 
Online banking management system project.pdf
Online banking management system project.pdfOnline banking management system project.pdf
Online banking management system project.pdfKamal Acharya
ย 
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced LoadsFEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced LoadsArindam Chakraborty, Ph.D., P.E. (CA, TX)
ย 
Call Girls in Ramesh Nagar Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service
Call Girls in Ramesh Nagar Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort ServiceCall Girls in Ramesh Nagar Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service
Call Girls in Ramesh Nagar Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service9953056974 Low Rate Call Girls In Saket, Delhi NCR
ย 
Call Girls in Netaji Nagar, Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service
Call Girls in Netaji Nagar, Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort ServiceCall Girls in Netaji Nagar, Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service
Call Girls in Netaji Nagar, Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service9953056974 Low Rate Call Girls In Saket, Delhi NCR
ย 
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...roncy bisnoi
ย 
KubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghlyKubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghlysanyuktamishra911
ย 
Unit 1 - Soil Classification and Compaction.pdf
Unit 1 - Soil Classification and Compaction.pdfUnit 1 - Soil Classification and Compaction.pdf
Unit 1 - Soil Classification and Compaction.pdfRagavanV2
ย 
Design For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the startDesign For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the startQuintin Balsdon
ย 
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance BookingCall Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance Bookingroncy bisnoi
ย 
NFPA 5000 2024 standard .
NFPA 5000 2024 standard                                  .NFPA 5000 2024 standard                                  .
NFPA 5000 2024 standard .DerechoLaboralIndivi
ย 
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...Call Girls in Nagpur High Profile
ย 
Thermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - VThermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - VDineshKumar4165
ย 
VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...
VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...
VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...SUHANI PANDEY
ย 
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 BookingVIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Bookingdharasingh5698
ย 
data_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdfdata_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdfJiananWang21
ย 

Recently uploaded (20)

Double Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueDouble Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torque
ย 
Top Rated Pune Call Girls Budhwar Peth โŸŸ 6297143586 โŸŸ Call Me For Genuine Se...
Top Rated  Pune Call Girls Budhwar Peth โŸŸ 6297143586 โŸŸ Call Me For Genuine Se...Top Rated  Pune Call Girls Budhwar Peth โŸŸ 6297143586 โŸŸ Call Me For Genuine Se...
Top Rated Pune Call Girls Budhwar Peth โŸŸ 6297143586 โŸŸ Call Me For Genuine Se...
ย 
Call Now โ‰ฝ 9953056974 โ‰ผ๐Ÿ” Call Girls In New Ashok Nagar โ‰ผ๐Ÿ” Delhi door step de...
Call Now โ‰ฝ 9953056974 โ‰ผ๐Ÿ” Call Girls In New Ashok Nagar  โ‰ผ๐Ÿ” Delhi door step de...Call Now โ‰ฝ 9953056974 โ‰ผ๐Ÿ” Call Girls In New Ashok Nagar  โ‰ผ๐Ÿ” Delhi door step de...
Call Now โ‰ฝ 9953056974 โ‰ผ๐Ÿ” Call Girls In New Ashok Nagar โ‰ผ๐Ÿ” Delhi door step de...
ย 
Bhosari ( Call Girls ) Pune 6297143586 Hot Model With Sexy Bhabi Ready For ...
Bhosari ( Call Girls ) Pune  6297143586  Hot Model With Sexy Bhabi Ready For ...Bhosari ( Call Girls ) Pune  6297143586  Hot Model With Sexy Bhabi Ready For ...
Bhosari ( Call Girls ) Pune 6297143586 Hot Model With Sexy Bhabi Ready For ...
ย 
Online banking management system project.pdf
Online banking management system project.pdfOnline banking management system project.pdf
Online banking management system project.pdf
ย 
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced LoadsFEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
ย 
Call Girls in Ramesh Nagar Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service
Call Girls in Ramesh Nagar Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort ServiceCall Girls in Ramesh Nagar Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service
Call Girls in Ramesh Nagar Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service
ย 
Call Girls in Netaji Nagar, Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service
Call Girls in Netaji Nagar, Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort ServiceCall Girls in Netaji Nagar, Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service
Call Girls in Netaji Nagar, Delhi ๐Ÿ’ฏ Call Us ๐Ÿ”9953056974 ๐Ÿ” Escort Service
ย 
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
ย 
KubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghlyKubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghly
ย 
Unit 1 - Soil Classification and Compaction.pdf
Unit 1 - Soil Classification and Compaction.pdfUnit 1 - Soil Classification and Compaction.pdf
Unit 1 - Soil Classification and Compaction.pdf
ย 
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak HamilCara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
ย 
Design For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the startDesign For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the start
ย 
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance BookingCall Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
ย 
NFPA 5000 2024 standard .
NFPA 5000 2024 standard                                  .NFPA 5000 2024 standard                                  .
NFPA 5000 2024 standard .
ย 
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...
ย 
Thermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - VThermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - V
ย 
VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...
VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...
VIP Model Call Girls Kothrud ( Pune ) Call ON 8005736733 Starting From 5K to ...
ย 
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 BookingVIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
ย 
data_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdfdata_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdf
ย 

10 more lessons learned from building Machine Learning systems

  • 1. 10MoreLessons Learned from building real-life Machine Learning Systems Xavier Amatriain (@xamat) 10/13/2015
  • 3. Our Mission โ€œTo share and grow the worldโ€™s knowledgeโ€ โ— Millions of questions & answers โ— Millions of users โ— Thousands of topics โ— ...
  • 4. Demand What we care about Quality Relevance
  • 5. Lots of data relations
  • 6. ML Applications @ Quora โ— Answer ranking โ— Feed ranking โ— Topic recommendations โ— User recommendations โ— Email digest โ— Ask2Answer โ— Duplicate Questions โ— Related Questions โ— Spam/moderation โ— Trending now โ— ...
  • 7. Models โ— Logistic Regression โ— Elastic Nets โ— Gradient Boosted Decision Trees โ— Random Forests โ— (Deep) Neural Networks โ— LambdaMART โ— Matrix Factorization โ— LDA โ— ...
  • 10. Implicit vs. Explicit โ— Many have acknowledged that implicit feedback is more useful โ— Is implicit feedback really always more useful? โ— If so, why?
  • 11. โ— Implicit data is (usually): โ—‹ More dense, and available for all users โ—‹ Better representative of user behavior vs. user reflection โ—‹ More related to final objective function โ—‹ Better correlated with AB test results โ— E.g. Rating vs watching Implicit vs. Explicit
  • 12. โ— However โ—‹ It is not always the case that direct implicit feedback correlates well with long-term retention โ—‹ E.g. clickbait โ— Solution: โ—‹ Combine different forms of implicit + explicit to better represent long-term goal Implicit vs. Explicit
  • 14. Training a model โ— Model will learn according to: โ—‹ Training data (e.g. implicit and explicit) โ—‹ Target function (e.g. probability of user reading an answer) โ—‹ Metric (e.g. precision vs. recall) โ— Example 1 (made up): โ—‹ Optimize probability of a user going to the cinema to watch a movie and rate it โ€œhighlyโ€ by using purchase history and previous ratings. Use NDCG of the ranking as final metric using only movies rated 4 or higher as positives.
  • 15. Example 2 - Quoraโ€™s feed โ— Training data = implicit + explicit โ— Target function: Value of showing a story to a user ~ weighted sum of actions: v = โˆ‘a va 1{ya = 1} โ—‹ predict probabilities for each action, then compute expected value: v_pred = E[ V | x ] = โˆ‘a va p(a | x) โ— Metric: any ranking metric
  • 17. Supervised/Unsupervised Learning โ— Unsupervised learning as dimensionality reduction โ— Unsupervised learning as feature engineering โ— The โ€œmagicโ€ behind combining unsupervised/supervised learning โ—‹ E.g.1 clustering + knn โ—‹ E.g.2 Matrix Factorization โ–  MF can be interpreted as โ— Unsupervised: โ—‹ Dimensionality Reduction a la PCA โ—‹ Clustering (e.g. NMF) โ— Supervised โ—‹ Labeled targets ~ regression
  • 18. Supervised/Unsupervised Learning โ— One of the โ€œtricksโ€ in Deep Learning is how it combines unsupervised/supervised learning โ—‹ E.g. Stacked Autoencoders โ—‹ E.g. training of convolutional nets
  • 20. Ensembles โ— Netflix Prize was won by an ensemble โ—‹ Initially Bellkor was using GDBTs โ—‹ BigChaos introduced ANN-based ensemble โ— Most practical applications of ML run an ensemble โ—‹ Why wouldnโ€™t you? โ—‹ At least as good as the best of your methods โ—‹ Can add completely different approaches (e. g. CF and content-based) โ—‹ You can use many different models at the ensemble layer: LR, GDBTs, RFs, ANNs...
  • 21. Ensembles & Feature Engineering โ— Ensembles are the way to turn any model into a feature! โ— E.g. Donโ€™t know if the way to go is to use Factorization Machines, Tensor Factorization, or RNNs? โ—‹ Treat each model as a โ€œfeatureโ€ โ—‹ Feed them into an ensemble
  • 22. The Master Algorithm? It definitely is an ensemble!
  • 24. Outputs will be inputs โ— Ensembles turn any model into a feature โ—‹ Thatโ€™s great! โ—‹ That can be a mess! โ— Make sure the output of your model is ready to accept data dependencies โ—‹ E.g. can you easily change the distribution of the value without affecting all other models depending on it? โ— Avoid feedback loops โ— Can you treat your ML infrastructure as you would your software one?
  • 25. ML vs Software โ— Can you treat your ML infrastructure as you would your software one? โ—‹ Yes and No โ— You should apply best Software Engineering practices (e.g. encapsulation, abstraction, cohesion, low couplingโ€ฆ) โ— However, Design Patterns for Machine Learning software are not well known/documented
  • 27. Feature Engineering โ— Main properties of a well-behaved ML feature โ—‹ Reusable โ—‹ Transformable โ—‹ Interpretable โ—‹ Reliable โ— Reusability: You should be able to reuse features in different models, applications, and teams โ— Transformability: Besides directly reusing a feature, it should be easy to use a transformation of it (e.g. log(f), max(f), โˆ‘ft over a time windowโ€ฆ)
  • 28. Feature Engineering โ— Main properties of a well-behaved ML feature โ—‹ Reusable โ—‹ Transformable โ—‹ Interpretable โ—‹ Reliable โ— Interpretability: In order to do any of the previous, you need to be able to understand the meaning of features and interpret their values. โ— Reliability: It should be easy to monitor and detect bugs/issues in features
  • 29. Feature Engineering Example - Quora Answer Ranking What is a good Quora answer? โ€ข truthful โ€ข reusable โ€ข provides explanation โ€ข well formatted โ€ข ...
  • 30. Feature Engineering Example - Quora Answer Ranking How are those dimensions translated into features? โ€ข Features that relate to the answer quality itself โ€ข Interaction features (upvotes/downvotes, clicks, commentsโ€ฆ) โ€ข User features (e.g. expertise in topic)
  • 32. Machine Learning Infrastructure โ— Whenever you develop any ML infrastructure, you need to target two different modes: โ—‹ Mode 1: ML experimentation โ–  Flexibility โ–  Easy-to-use โ–  Reusability โ—‹ Mode 2: ML production โ–  All of the above + performance & scalability โ— Ideally you want the two modes to be as similar as possible โ— How to combine them?
  • 33. Machine Learning Infrastructure: Experimentation & Production โ— Option 1: โ—‹ Favor experimentation and only invest in productionizing once something shows results โ—‹ E.g. Have ML researchers use R and then ask Engineers to implement things in production when they work โ— Option 2: โ—‹ Favor production and have โ€œresearchersโ€ struggle to figure out how to run experiments โ—‹ E.g. Implement highly optimized C++ code and have ML researchers experiment only through data available in logs/DB
  • 34. Machine Learning Infrastructure: Experimentation & Production โ— Option 1: โ—‹ Favor experimentation and only invest in productionazing once something shows results โ—‹ E.g. Have ML researchers use R and then ask Engineers to implement things in production when they work โ— Option 2: โ—‹ Favor production and have โ€œresearchersโ€ struggle to figure out how to run experiments โ—‹ E.g. Implement highly optimized C++ code and have ML researchers experiment only through data available in logs/DB
  • 35. โ— Good intermediate options: โ—‹ Have ML โ€œresearchersโ€ experiment on iPython Notebooks using Python tools (scikit-learn, Theanoโ€ฆ). Use same tools in production whenever possible, implement optimized versions only when needed. โ—‹ Implement abstraction layers on top of optimized implementations so they can be accessed from regular/friendly experimentation tools Machine Learning Infrastructure: Experimentation & Production
  • 37. Model debuggability โ— Value of a model = value it brings to the product โ— Product owners/stakeholders have expectations on the product โ— It is important to answer questions to why did something fail โ— Bridge gap between product design and ML algos โ— Model debuggability is so important it can determine: โ—‹ Particular model to use โ—‹ Features to rely on โ—‹ Implementation of tools
  • 38. Model debuggability โ— E.g. Why am I seeing or not seeing this on my homepage feed?
  • 40. Distributing ML โ— Most of what people do in practice can fit into a multi- core machine โ—‹ Smart data sampling โ—‹ Offline schemes โ—‹ Efficient parallel code โ— Dangers of โ€œeasyโ€ distributed approaches such as Hadoop/Spark โ— Do you care about costs? How about latencies?
  • 41. Distributing ML โ— Example of optimizing computations to fit them into one machine โ—‹ Spark implementation: 6 hours, 15 machines โ—‹ Developer time: 4 days โ—‹ C++ implementation: 10 minutes, 1 machine โ— Most practical applications of Big Data can fit into a (multicore) implementation
  • 43. Data Scientists and ML Engineers โ— We all know the definition of a Data Scientist โ— Where do Data Scientists fit in an organization? โ—‹ Many companies struggling with this โ— Valuable to have strong DS who can bring value from the data โ— Strong DS with solid engineering skills are unicorns and finding them is not scalable โ—‹ DS need engineers to bring things to production โ—‹ Engineers have enough on their plate to be willing to โ€œproductionizeโ€ cool DS projects
  • 44. The data-driven ML innovation funnel Data Research ML Exploration - Product Design AB Testing
  • 45. Data Scientists and ML Engineers โ— Solution: โ—‹ (1) Define different parts of the innovation funnel โ–  Part 1. Data research & hypothesis building -> Data Science โ–  Part 2. ML solution building & implementation -> ML Engineering โ–  Part 3. Online experimentation, AB Testing analysis-> Data Science โ—‹ (2) Broaden the definition of ML Engineers to include from coding experts with high-level ML knowledge to ML experts with good software skills Data Research ML Solution AB Testing Data Science Data Science ML Engineering
  • 47. โ— Make sure you teach your model what you want it to learn โ— Ensembles and the combination of supervised/unsupervised techniques are key in many ML applications โ— Important to focus on feature engineering โ— Be thoughtful about โ—‹ your ML infrastructure/tools โ—‹ about organizing your teams