Machine Learning Introduction for Digital Business Leaders

Preview:

Citation preview

Machine Learning Introduction for Digital Business Leaders

Sudha Jamthe

Barcelona Technology School, Barcelona, Oct 2016

1IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Instructor: SUDHA JAMTHE

2

The IoT Show (YouTube)

Past : Mobile, Social Growth Leader

STEM & Regional Economic Development –Turkey/Czech/Barcelona

CEO IoTDisruption, Technology Futurist

IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

MACHINE LEARNING IS A PART OF AI

3

Machine Learning

VRAR

Voice to Text

Text to Speech

Facial Recognition

Computer Vision

Natural Language Processing

IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Copyright: Sudha Jamthe

IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Machine Learning teaches computers to develop programs with large

volume of data and drive decisions without human intervention.

4IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

What can a computer do with Machine Learning?

1. Tell the self driving car to stop2. This newsfeed is spam3. This photo is the face of Sudha Jamthe4. Sell this stock now5. I recommend this movie on Netflix 6. I can confirm if this cell looks like cancer cell

or not

5IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

My Machine Learning Dog

6IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Recommendation Engines

Amazon uses ML for video recommendation

7IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Facial Recognition – Google Photos

8IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Computer Vision

Take a series of 2D pictures to create a 3D analysise.g. Civilmaps, Robots, Wearable Pills

9IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Health Informatics

10IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Image: Stanford Student research on Cancer from Tobacco usage

Facebook – Spam Detection

• Facebook Spam Detection

• Credit Card company – Fraud Detection

• Pathogen in Blood Samples

11IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

12

WHICH ONE IS AMY MY ASSISTANT?

CLUE: Amy is a botIoTDisruptions.co http://bit.ly/IoTAIBook @sujamthe

Language Translations

Live Demo (Try this on your computers)http://www.ibm.com/watson/developerclo

ud/services-catalog.html

Speech to Text https://speech-to-text-demo.mybluemix.net/http://visual-recognition-demo.mybluemix.net/

https://cloud.google.com/products/machine-learning/https://cloud.google.com/speech/

13IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

GOOGLE DEEPMIND TO PREDICT KIDNEY PROBLEMS (DEEP LEARNING ALGORITHM)

14

Image source: Digital Trends

IoTDisruptions.com http://bit.ly/IoTAIBook @sujamtheIoTDisruptions.co http://bit.ly/IoTAIBook @sujamthe

AUTONOMOUS VEHICLES

15

Image: Scania

IoTDisruptions.co http://bit.ly/IoTAIBook @sujamthe

16

ROBOTS MAKING MACHINES

IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Marketing

• Customer Segmentation

• Churn Prediction

• Customer Analytics

17IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Machine Learning APIs

• api.ai, Cogito, DataSift, iSpeech, Microsoft Project Oxford, Mozscape, and OpenCalais.

• Amazon, Google, IBM Watson, and Microsoft

• Amazon AWS serviceshttps://aws.amazon.com/machine-learning/

18IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Machine Learning Jobs

IoTDisruptions.co http://bit.ly/IoTAIBook @sujamthe

19

Who Does Machine Learning Today?

• http://datam.i2r.a-star.edu.sg/dad/dadMLlab.htm

• https://www.ml.cmu.edu/people/phd-students.html

IoTDisruptions.co http://bit.ly/IoTAIBook @sujamthe

20

ML Job Roles in Companies Digital Transformation

ML Researcher ML Engineer -> Data Scientist

ML Product Manager/Business Roles (EVOLVING)

21IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Machine Learning Jobs (6.6K jobs)

• Facebook ML AI/ ML Program Manager • Siemens - Senior Innovator – (Machine Learning / Deep Learning)• Chief Machine Learning Scientist - $5m compensation package (FinTech)• Twitter -Product Manager, Core Ads: Ad Supply Management• SAP Frontend Developer for Machine Learning development• GroupOn Director, Business Operations & Analytics• Apple Sr. Infrastructure Engineer - Applied Machine Learning• Machine Learning Deployment Engineer, Professional Services, Google Cloud• Sr. Product Manager in LinkedIn, Splunk, Walmart, Cisco, Amazon• Computational Biologist, 23andMe• SAP, Director, Platform Strategy• Visa, Senior Design Technologist• CTO, XiQ• Sr. Product Manager, Capital One Bank Labs• Business and Market Strategist, Enlitic

22IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

How does Machine Learning Work?

24IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

How does Machine Learning Work?

1. Supervised

– Get Training Data, Define factors, Remove biases

2. Unsupervised Learning aka Deep Mining

- No need to define factors, train same architecture for generic problem solving.

25IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Machine Learning Works by Building Data Models (a geometric shape of

data*)

image: Techemregence Gaussian mixture model

Training Data & Factors

Predict Outcomes to make decisions

26IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Machine Learning Platform(TensorFlow, ScikitLearn, Theano, Spark Mlib, H20)

The right model(decision tree, nearest neighbor, neural net, support vector machine,

ensemble of multiple models)

A learning procedure to fit the data (linear regression, gradient descent, genetic algorithms, bagging,

boosting, and other model-specific methods)

Right data and fit(bias and variance, overfitting and underfitting, missing data,

data leakage)

Deploying a ML Service into a Product(Identity, Data Governance, Algorithmic accuracy/

Compliance for industry)

Building Products with Machine Learning

© Sudha Jamthe 27IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Building Products with Machine Learning

28IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Pick a Machine Learning Platform(TensorFlow, ScikitLearn, Theano, Spark Mlib, H20)

Develop the right stats model (decision tree, nearest neighbor, neural net, support vector machine) and Fit the data using right learning procedure ( linear regression, gradient descent, boosting, nagging)

Develop Training Data & Fit(bias and variance, overfitting and underfitting, missing data,

data leakage)

Defining a ML Service into a Product(Identity, Data Governance, Algorithmic accuracy/

Compliance for industry)

© Sudha Jamthe

ML Engineer

ML Researcher

ML Product Manager

AI COMPANIES LANDSCAPE

29IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

30

INDUSTRIES IMPACTED BY MACHINE LEARNING

IoTDisruptions.co http://bit.ly/IoTAIBook @sujamthe

Questions?

http://slideshare.com/sujamthe

Email me sujamthe@gmail.com

Twitter @sujamthe

IoT Disruptions 2020http://bit.ly/IoTAIBook

31IoTDisruptions.com http://bit.ly/IoTAIBook @sujamthe

Recommended