Slide 1

Slide 1 text

No content

Slide 2

Slide 2 text

Getting started with ML using Tensorflow’s high-level APIs Akash Tandon Data Engineering @ SocialCops

Slide 3

Slide 3 text

XKCD comics

Slide 4

Slide 4 text

Moving beyond the buzzwords Onalytica.com

Slide 5

Slide 5 text

Machine learning in a nutshell - What is ML? - What exactly is a model again? - Types of ML algorithms - When to use it?

Slide 6

Slide 6 text

Machine learning in a nutshell "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.” - Tom Mitchell

Slide 7

Slide 7 text

ML programming model Source

Slide 8

Slide 8 text

ML in the wild - Health industry - Agriculture - E-commerce - .. and the list goes on!

Slide 9

Slide 9 text

7 steps of ML - Gathering data - Data preparation - Choosing a model - Training - Evaluation - Parameter tuning - Prediction We’ll better understand these steps through a code example further ahead.

Slide 10

Slide 10 text

Why Tensorflow? - History - Concept of graphs - Support of end-to-end workflow including deployment - Vibrant community and diverse ecosystem including support for multiple languages including Python, JS, and R

Slide 11

Slide 11 text

What’s with the name? - Tensor + Flow - Tensors - Graphs and dataflow

Slide 12

Slide 12 text

Tensors ~ data representation source

Slide 13

Slide 13 text

Dataflow graphs source

Slide 14

Slide 14 text

Hello world of ML - Image classification using MNIST dataset - We’ll use Google’s CoLab for this.

Slide 15

Slide 15 text

Time to code!

Slide 16

Slide 16 text

I built my model, but what now?

Slide 17

Slide 17 text

Enter Tensorflow serving - Flexible, high-performance serving system for machine learning models, designed for production environments. - Export your pre-trained model and serve them to consumers/customers.

Slide 18

Slide 18 text

Diverse ecosystem - Move across languages and re-use your models - Train your model in R or Python - Serve them in production using Javascript using tensorflow.js - Visualize processes using Tensorboard - Serve models using tf-serving - Make music and art using Tensorflow Magenta - … and a lot more!

Slide 19

Slide 19 text

ML resources - https://ai.google/education/ - https://developers.google.com/machine-learning/crash-c ourse/ - https://github.com/tensorflow/workshops - https://www.coursera.org/learn/machine-learning

Slide 20

Slide 20 text

Email: [email protected] Twitter: @AkashTandon Github: @analyticalmonk Feel free to reach out

Slide 21

Slide 21 text

That’s all, folks!