Machine Learning and TensorFlow
Artificial Intelligence Present and Future
José Papo
Gerente de relações com startups e developers
Google América Latina
@josepapo
“Machine learning is a core,
transformative way by which we’re
re-thinking how we’re doing everything”
Sundar Pichai
CEO, Google
“Machine learning will cause every
successful huge IPO win in 5 years.”
Eric Schmidt
Executive Chairman, Alphabet
Basic Concepts
● Artificial General Intelligence
● Artificial Superintelligence
● Artificial Narrow Intelligence
Artificial Intelligence
Machine Learning (Narrow AI)
Deep Learning (ML on Steroids!!!)
What’s different now from 10 years ago?
WAY MORE
DATA
More
Compute
Better
Algorithms
Machine Learning at Google
Products using Machine Learning
TensorFlow
● Open source Machine
Learning library
● Especially useful for
Deep Learning
● For research and
production
● Apache 2.0 license
Raspberry
Pi
DatacentersYour laptop Android iOS
Portable & Scalable
A multidimensional array.
A graph of operations.
Data Flow Graphs
Computation is defined as a directed acyclic graph
(DAG) to optimize an objective function
● Graph is defined in high-level language (Python)
● Graph is compiled and optimized
● Graph is executed (in parts or fully) on available low
level devices (CPU, GPU)
● Data (tensors) flow through the graph
● TensorFlow can compute gradients automatically
Image source: Wikimedia
+ =
A Neural Algorithm of Artistic Style http://arxiv.org/abs/1508.06576
?
Image source: Wikimedia
+ =
A Neural Algorithm of Artistic Style http://arxiv.org/abs/1508.06576
Image source: Wikimedia
+ =
A Neural Algorithm of Artistic Style http://arxiv.org/abs/1508.06576
goo.gl/fyDxhC
Most popular ML open source project on GitHub
Cloud, Mobile, Machine Learning
Cloud Machine Learning APIs
See, Hear and Understand the world
Cloud
Natural Language
Cloud
Speech
Cloud
Translate
Cloud
Vision
Faces
Faces, facial landmarks, emotions
OCR
Read and extract text, with
support for > 10 languages
Label
Detect entities from furniture to
transportation
Logos
Identify product logos
Landmarks & Image Properties
Detect landmarks & dominant
color of image
Safe Search
Detect explicit content - adult,
violent, medical and spoof
Cloud Vision API
Confidential & ProprietaryGoogle Cloud Platform 34
Cloud Natural Language API
Extract sentence, identify parts of
speech and create dependency parse
trees for each sentence.
Identify entities and label by types such
as person, organization, location, events,
products and media.
Understand the overall sentiment of a
block of text.
Syntax Analysis Entity Recognition
Sentiment Analysis
Confidential & ProprietaryGoogle Cloud Platform 35
Cloud Speech API
Automatic Speech Recognition (ASR)
powered by deep learning neural
networking to power your
applications like voice search or
speech transcription.
Recognizes over 80
languages and variants
with an extensive
vocabulary.
Returns partial
recognition results
immediately, as they
become available.
Filter inappropriate
content in text results.
Audio input can be captured by an application’s
microphone or sent from a pre-recorded audio
file. Multiple audio file formats are supported,
including FLAC, AMR, PCMU and linear-16.
Handles noisy audio from many
environments without requiring
additional noise cancellation.
Audio files can be uploaded in the
request and, in future releases,
integrated with Google Cloud
Storage.
Automatic Speech Recognition Global Vocabulary Inappropriate Content
Filtering
Streaming Recognition
Real-time or Buffered Audio Support Noisy Audio Handling Integrated API
Mobile Vision API
Providing on-device vision for applications
Face API
faces, facial landmarks, eyes
open, smiling
Barcode API
1D and 2D barcodes
Text API
Latin-based text / structure
Common Mobile Vision API
Support for fast image and video on-device detection and tracking.
NEW!
Face API
Photo credit developers.google.com/vision
Text Detection
Latin based language
Understand text structure
Photo credit Getty Images
Barcode Detection
1D barcodes
EAN-13/8
UPC-A/E
Code-39/93/128
ITF
Codabar
2D barcodes
QR Code
Data Matrix
PDF-417
AZTEC
UPC
DataMatrix
QR Code
PDF 417
Video and image credit Google
Machine Learning Democratization
Use Cases in Latin America
ACESSO UNIVERSAL A MEDICINA DE QUALIDADE
Machine Learning
AGENDA
• Otimização do broadcast
• Otimização do processo billing
• Personal Cloud
Machine
Learning
operacion
al
comercia
l
usuário
s
3perspectiva
s
Otimização do broadcast
• Reduzir a quantidade de envio de mensagens de
estímulo mantendo a mesma taxa de retorno.
comercia
l Desafio
Proposta
• Identificar o comportamento ou características dos
usuários mais propensos a responder ao estímulo.
Otimização do broadcast
• Text Mining para tratamento das frases,
classificando-as, como por exemplo, pela ideia
transmitida.
• Análise de modelos preditivos para seleção dos
clientes mais propensos.
comercia
l
Processo de
análise
Otimização do broadcast
• Prever quem não irá responder a nossa oferta nos
dá a possibilidade de pensarmos em algo diferente
para este usuário e desta forma conhecê-lo um
pouco mais.
• Redução de média 40% nos envios de broadcast.
comercia
l
Resultad
o
Otimização do billing
• Aumentar o sucesso nas cobranças dos serviços
prestados.
Desafi
o
Propost
a
• Identificar os clientes mais propensos em
determinados horários.
operacion
al
Otimização do billing
• Tratamento e enriquecimento da base de dados com
BigQuery.
• Análise de modelos preditivos para criação de escore de
crédito.
Processo de
análise
operacion
al
Otimização do billing
Proposição de
uso
Otimização do billing
• Redução de custos com infraestrutura de TI, uso
mais inteligente de recursos.
• Melhora de 42% em média na acertividade do
billing.
Resultad
o
operacion
al
Personal Cloud
• Detectar objetos e faces dentro das fotos dos
usuários do Personal Cloud para possibilitar busca e
criação de álbuns de forma automática
Desafi
o
Propost
a
• Utilização da API do Google Cloud Vision.
usuário
s
Busca por tags e álbuns
automáticos
Otimização do broadcast
Processo de
análiseusuário
s pé
dedo
bolsa
óculo
s
praia
Don’t Think Outside The
Box, Think Like There is NO
BOX!
tensorflow.org
github.com/tensorflow
Want to learn more?
Udacity class on Deep Learning, goo.gl/iHssII
Guides, codelabs, videos
MNIST for Beginners, goo.gl/tx8R2b
TF Learn Quickstart, goo.gl/uiefRn
TensorFlow for Poets, goo.gl/bVjFIL
ML Recipes, goo.gl/KewA03
TensorFlow and Deep Learning without a PhD, goo.gl/pHeXe7
Learn More...
OBRIGADO!
slideshare.net/jpapo
José Papo
@josepapo

Machine learning and TensorFlow

  • 1.
    Machine Learning andTensorFlow Artificial Intelligence Present and Future José Papo Gerente de relações com startups e developers Google América Latina @josepapo
  • 3.
    “Machine learning isa core, transformative way by which we’re re-thinking how we’re doing everything” Sundar Pichai CEO, Google
  • 4.
    “Machine learning willcause every successful huge IPO win in 5 years.” Eric Schmidt Executive Chairman, Alphabet
  • 5.
  • 6.
    ● Artificial GeneralIntelligence ● Artificial Superintelligence ● Artificial Narrow Intelligence Artificial Intelligence
  • 7.
  • 8.
    Deep Learning (MLon Steroids!!!)
  • 10.
    What’s different nowfrom 10 years ago? WAY MORE DATA More Compute Better Algorithms
  • 12.
  • 13.
  • 19.
  • 21.
    ● Open sourceMachine Learning library ● Especially useful for Deep Learning ● For research and production ● Apache 2.0 license
  • 22.
  • 23.
    A multidimensional array. Agraph of operations.
  • 24.
    Data Flow Graphs Computationis defined as a directed acyclic graph (DAG) to optimize an objective function ● Graph is defined in high-level language (Python) ● Graph is compiled and optimized ● Graph is executed (in parts or fully) on available low level devices (CPU, GPU) ● Data (tensors) flow through the graph ● TensorFlow can compute gradients automatically
  • 26.
    Image source: Wikimedia += A Neural Algorithm of Artistic Style http://arxiv.org/abs/1508.06576 ?
  • 27.
    Image source: Wikimedia += A Neural Algorithm of Artistic Style http://arxiv.org/abs/1508.06576
  • 28.
    Image source: Wikimedia += A Neural Algorithm of Artistic Style http://arxiv.org/abs/1508.06576 goo.gl/fyDxhC
  • 29.
    Most popular MLopen source project on GitHub
  • 30.
  • 31.
    Cloud Machine LearningAPIs See, Hear and Understand the world
  • 32.
  • 33.
    Faces Faces, facial landmarks,emotions OCR Read and extract text, with support for > 10 languages Label Detect entities from furniture to transportation Logos Identify product logos Landmarks & Image Properties Detect landmarks & dominant color of image Safe Search Detect explicit content - adult, violent, medical and spoof Cloud Vision API
  • 34.
    Confidential & ProprietaryGoogleCloud Platform 34 Cloud Natural Language API Extract sentence, identify parts of speech and create dependency parse trees for each sentence. Identify entities and label by types such as person, organization, location, events, products and media. Understand the overall sentiment of a block of text. Syntax Analysis Entity Recognition Sentiment Analysis
  • 35.
    Confidential & ProprietaryGoogleCloud Platform 35 Cloud Speech API Automatic Speech Recognition (ASR) powered by deep learning neural networking to power your applications like voice search or speech transcription. Recognizes over 80 languages and variants with an extensive vocabulary. Returns partial recognition results immediately, as they become available. Filter inappropriate content in text results. Audio input can be captured by an application’s microphone or sent from a pre-recorded audio file. Multiple audio file formats are supported, including FLAC, AMR, PCMU and linear-16. Handles noisy audio from many environments without requiring additional noise cancellation. Audio files can be uploaded in the request and, in future releases, integrated with Google Cloud Storage. Automatic Speech Recognition Global Vocabulary Inappropriate Content Filtering Streaming Recognition Real-time or Buffered Audio Support Noisy Audio Handling Integrated API
  • 36.
    Mobile Vision API Providingon-device vision for applications
  • 37.
    Face API faces, faciallandmarks, eyes open, smiling Barcode API 1D and 2D barcodes Text API Latin-based text / structure Common Mobile Vision API Support for fast image and video on-device detection and tracking. NEW!
  • 38.
    Face API Photo creditdevelopers.google.com/vision
  • 39.
    Text Detection Latin basedlanguage Understand text structure Photo credit Getty Images
  • 40.
    Barcode Detection 1D barcodes EAN-13/8 UPC-A/E Code-39/93/128 ITF Codabar 2Dbarcodes QR Code Data Matrix PDF-417 AZTEC UPC DataMatrix QR Code PDF 417 Video and image credit Google
  • 41.
  • 58.
    ACESSO UNIVERSAL AMEDICINA DE QUALIDADE
  • 59.
  • 60.
    AGENDA • Otimização dobroadcast • Otimização do processo billing • Personal Cloud Machine Learning
  • 61.
  • 62.
    Otimização do broadcast •Reduzir a quantidade de envio de mensagens de estímulo mantendo a mesma taxa de retorno. comercia l Desafio Proposta • Identificar o comportamento ou características dos usuários mais propensos a responder ao estímulo.
  • 63.
    Otimização do broadcast •Text Mining para tratamento das frases, classificando-as, como por exemplo, pela ideia transmitida. • Análise de modelos preditivos para seleção dos clientes mais propensos. comercia l Processo de análise
  • 64.
    Otimização do broadcast •Prever quem não irá responder a nossa oferta nos dá a possibilidade de pensarmos em algo diferente para este usuário e desta forma conhecê-lo um pouco mais. • Redução de média 40% nos envios de broadcast. comercia l Resultad o
  • 65.
    Otimização do billing •Aumentar o sucesso nas cobranças dos serviços prestados. Desafi o Propost a • Identificar os clientes mais propensos em determinados horários. operacion al
  • 66.
    Otimização do billing •Tratamento e enriquecimento da base de dados com BigQuery. • Análise de modelos preditivos para criação de escore de crédito. Processo de análise operacion al
  • 67.
  • 68.
    Otimização do billing •Redução de custos com infraestrutura de TI, uso mais inteligente de recursos. • Melhora de 42% em média na acertividade do billing. Resultad o operacion al
  • 69.
    Personal Cloud • Detectarobjetos e faces dentro das fotos dos usuários do Personal Cloud para possibilitar busca e criação de álbuns de forma automática Desafi o Propost a • Utilização da API do Google Cloud Vision. usuário s Busca por tags e álbuns automáticos
  • 70.
    Otimização do broadcast Processode análiseusuário s pé dedo bolsa óculo s praia
  • 71.
    Don’t Think OutsideThe Box, Think Like There is NO BOX!
  • 93.
    tensorflow.org github.com/tensorflow Want to learnmore? Udacity class on Deep Learning, goo.gl/iHssII Guides, codelabs, videos MNIST for Beginners, goo.gl/tx8R2b TF Learn Quickstart, goo.gl/uiefRn TensorFlow for Poets, goo.gl/bVjFIL ML Recipes, goo.gl/KewA03 TensorFlow and Deep Learning without a PhD, goo.gl/pHeXe7 Learn More...
  • 94.