hands-on with google’s machine learning apis, 11/17/2016

Hands-On with Google’s Machine

Learning APIsStephen Wylie

11/17/2016

@SWebCEO+StephenWyliemrcity

About meWEBSITESwww.stev-o.usgoshtastic.blogspot.comwww.ledgoes.comwww.openbrite.com

[email protected]@openbrite.com

G+https://plus.google.com/u/1/+StephenWylie

TWITTER@SWebCEO

GITHUBmrcity

Senior Software Engineer at Capital One

Test/QA Lead for Auto Finance Innovation

Successful Kickstarter (1000% funded - BriteBlox)

Intel Innovator, DMS Member, Vintage computer collector/restorer/homebrewer, hackathoner

Civic Hacking Cmte Chair @ DMS@SWebCEO +StephenWylie #MachineLearning

http://www.stev-o.us/

https://goshtastic.blogspot.com/

http://www.ledgoes.com/

http://www.openbrite.com/

mailto:[email protected]

mailto:[email protected]

https://plus.google.com/u/1/+StephenWylie

https://plus.google.com/u/1/+StephenWylie

https://twitter.com/SWebCEO

https://twitter.com/SWebCEO

https://github.com/mrcity/

Tonight’s Mission

Touch on ML API offerings Explore Google’s RESTful ML APIs

Cloud Vision & Natural Language API Prediction API TensorFlow (Not a RESTful API but still cool)

Allocate time to play, develop ideas Have good conversations, network Find learning partner or group@SWebCEO +StephenWylie #MachineLearning

Before We Start…

Hopefully you followed instructions on https://github.com/mrcity/mlworkshop/

Get access to the APIsGet a REST Client Install TensorFlow@SWebCEO +StephenWylie #MachineLearning

https://github.com/mrcity/mlworkshop/

https://github.com/mrcity/mlworkshop/

What is Machine Learning?

@SWebCEO +StephenWylie #MachineLearning

Who’s using ML?

Chat bots? Self-driving cars? Pepper the robot? Uber for selfie ID? Document recognition & field extraction


Machine Learning Tools Ecosystem

APIs you interface with HP, Amazon, Microsoft, IBM, Google, Facebook’s Caffe on mobile

& Web Software you use

Orange (U of Ljubljana, Slovenia) Weka (U of Waikato, New Zealand)

Hardware you compile programs to run on nVidia GPUs with CUDA, DGX-1 supercomputer Can BTC hardware be used for ML?


Google’s ML APIs In Particular

Google Play Services Mobile Vision API

RESTful ML Services Cloud Vision API Cloud Natural Language API Prediction API

Local ML Services TensorFlow SyntaxNet


^ Pre-defined models

v User-defined models

Cloud Vision APIWhat does it look like to you?


Detect Faces, Parse Barcodes, Segment Text


Availability

Native AndroidNative

iOSRESTful

API

FACE API BARCODE API TEXT API

What do you see in that cloud?

Breaks down into more features than just FACE, BARCODE, and TEXT:

From https://cloud.google.com/vision/docs/requests-and-responses


Feature Type DescriptionLABEL_DETECTION Execute Image Content Analysis on the entire image and returnTEXT_DETECTION Perform Optical Character Recognition (OCR) on text within the

imageFACE_DETECTION Detect faces within the imageLANDMARK_DETECTION

Detect geographic landmarks within the image

LOGO_DETECTION Detect company logos within the imageSAFE_SEARCH_DETECTION

Determine image safe search properties on the image

IMAGE_PROPERTIES Compute a set of properties about the image (such as the image's dominant colors)

https://cloud.google.com/vision/docs/requests-and-responses

Cloud Vision APIs

Can simultaneously detect multiple featuresFeatures billed individually per use on imageNo Barcode feature as yetSimple JSON request/response formatSubmit image from Cloud Storage or in

Base64Returns 0 or more annotations by confidence@SWebCEO +StephenWylie #MachineLearning

Response types


Feature Returns

Label Description of the picture’s contentsConfidence score

Text, Logo Text contents or logo owner nameBounding polygon containing the text or logo[Logo only] Confidence score

Face Bounding polygon and rotational characteristics of the facePositions of various characteristics such as eyes, ears, lips, chin, foreheadConfidence score of exhibiting joy, sorrow, anger, or surprise

Landmark

Safe Search Likelihood of the image containing adult or violent content, that it was a spoof, or contains graphic medical imagery

Image properties

Dominant RGB colors within the image, ordered by fraction of pixels

Demo


Mobile Vision vs. Cloud Vision

Mobile Vision is for Native Android

Handles more data processingCan utilize camera videoTakes advantage of OpenGL

(& hardware?)@SWebCEO +StephenWylie #MachineLearning

Cloud Natural Language APIMaking computers speak human


Natural Language API: Analyze Any ASCIIParses text for parts of speechDiscovers entities like organizations, people,

locationsAnalyzes text sentimentUse Speech, Vision, Translate APIs upstreamWorks with English, Spanish, or JapaneseSentiment analysis only available for English@SWebCEO +StephenWylie #MachineLearning

Sample NL API Request

@SWebCEO +StephenWylie #MachineLearningFrom https://cloud.google.com/natural-language/docs/basics

<- Optional, can be guessed automatically

<- Not required for Sentiment Analysis queries

<- Optional, defaults to Entities

https://cloud.google.com/natural-language/docs/basics

Interpreting NL API Sentiment Responses


POLARITY

-1 1

MAGNITUDE

0 ∞

1

10

102

103

Sample analyzeSentiment response for the

Gettysburg Address:

{ “polarity”: 0.4, “magnitude”: 3.8}

Demo


Google Prediction APITo further their conquest for all knowledge past, present, and future


Making Predictions With Google

Build “trained” model or use “hosted” model Hosted models (all demos):

Language identifier Tag categorizer (as android, appengine, chrome, youtube) Sentiment predictor

Trained models: Submit attributes and labels for each example Need at least six examples Store examples in Cloud Storage


Don’t Model Trains; Train Your Model

Train API against dataprediction.trainedmodels.insert

Send prediction queryprediction.trainedmodels.predict

Update the modelprediction.trainedmodels.update

Other CRUD operations: list, get, delete@SWebCEO +StephenWylie #MachineLearning

Don’t Model Trains; Train Your Model

Insert query requires:idmodelTypestorageDataLocation

Don’t forget: poll for status updates


Demo


TensorFlowAll that Linear Algebra you slept through in college


About TensorFlow

Offline library for large-scale numerical computation

Think of a graph: Nodes represent

mathematical operations Edges represent tensors

flowing between them Excellent at building

deep neural networks@SWebCEO +StephenWylie #MachineLearning

𝑅𝑒𝐿𝑈𝑛 ≈ 𝑓 (𝑥 )

Tense About Tensors?

Think about MNIST handwritten digitsEach number is 28 pixels squaredThere are 10 numbers, 0-9


Tense About Tensors?

Define an input tensor of shape(any batch size, 784) x = tf.placeholder(tf.float32, shape=[None, 784])

Define a target output tensor of shape(any batch size, 10) y_ = tf.placeholder(tf.float32, shape=[None, 10])

Define weights matrix (784x10)and biases vector (10-D)


One-Hot: Cool To the Touch Load the input data

from tensorflow.examples.tutorials.mnist import input_datamnist = input_data.read_data_sets('MNIST_data', one_hot=True)

One-hot?! Think about encoding categorical features:

US = 0, UK = 1, India = 2, Canada = 3, … This implies ordinal properties and confuses learners Break encoding into Booleans:

This is where the 10-D target output tensor comes from


US = [1, 0, 0, 0]UK = [0, 1, 0, 0]Etc…

TensorFlow Data Structures - Placeholders Come from inputs prior to computation x (input picture as vector), y_ (one-hot

10-D classification vector)


[0, 0, 0, 0, 0, 0, 0, 0, …

0, 0, 0, 0, 0, 1, 1, 0, …

0, 0, 0, 0, 1, 1, 1, 0, …

0, 0, 0, 0, 1, 1, 0, 0, …

0, 0, 0, 1, 1, 1, 0, 0, …...

[0, 0, 0, 0, 1,0, 0, 0, 0, 0]

Input x y_

TensorFlow Data Structures – Variables Values (i.e. model parameters) inside

nodes in a graph Used and modified by learning process Need to be initialized with

init = tf.initialize_all_variables() W (weights to scale inputs by), b (bias to

add to scaled value)@SWebCEO +StephenWylie #MachineLearning

Training a Dragon, if the Dragon is a Model Your Simple Model:

y = tf.nn.softmax(tf.matmul(x, W) + b) Cross-entropy: distance between guess & correct answer

cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y), reduction_indices=[1]))

Gradient descent: minimize cross-entropytrain_step = tf.train.GradientDescentOptimizer(0.5).minimize(cross_entropy)

Learning rate: 0.5@SWebCEO +StephenWylie #MachineLearning

𝐻 𝑦 ′ (𝑦 )=−∑𝑖𝑦 𝑖′ log (𝑦 𝑖)

Dragon Get Wiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiings!

Start a Session Run initialize_all_variables Run training for 1000 steps

sess = tf.Session()sess.run(init)for i in range(1000): batch_xs, batch_ys = mnist.train.next_batch(100) sess.run(train_step, feed_dict={x: batch_xs, y_: batch_ys})

Expensive to use all training data at once! Pick 100 random samples each step


Test Flight Evaluation

Compare labels between guess y and correct y_

correct_prediction =tf.equal(tf.argmax(y,1), tf.argmax(y_,1))

Cast each Boolean result into either a 0 or 1, then average itaccuracy = tf.reduce_mean( tf.cast(correct_prediction, tf.float32))

Print the final figureprint(sess.run(accuracy, feed_dict={x: mnist.test.images, y_: mnist.test.labels}))


Demo


Future Talks, Other Talks

Follow me if you want to hear these! Build a Neural Network in Python with NumPy Build a Neural Network with nVidia CUDA

Elsewhere, Mapmaking with Google Maps API, Polymer,

and Firebase The Process Of Arcade Game ROM Hacking


More Resources

Google’s “Googly Eyes” Android app [Mobile Vision API]https://github.com/googlesamples/android-vision/tree/master/visionSamples/googly-eyes

Quick, Draw! Google classification API for sketcheshttps://quickdraw.withgoogle.com/


https://github.com/googlesamples/android-vision/tree/master/visionSamples/googly-eyes

https://github.com/googlesamples/android-vision/tree/master/visionSamples/googly-eyes

https://quickdraw.withgoogle.com/

https://quickdraw.withgoogle.com/

Thank You@SWebCEO +StephenWylie #MachineLearning

hands-on with google’s machine learning apis, 11/17/2016

Data & Analytics