PyCon Pune 2017

Tensorflow is a leading open source machine learning library in python which uses data flow graphs for numerical computation. While the nodes represent mathematical operations, the edges of graph represent tensors. These tensors flow through the data flow graph and based on the operations, the values of the gradients are calculated.

My talk will cover the underlying algorithms and ideas which were used to create this library. I will also try to bring out contrasts with respect to other popular libraries.

After the talk, you will know the basic constructs of Tensorflow and how they work together to deliver the awesomeness.

Outline

Tensorflow - A brief overview
Computation Graph
Basic Constructs (Variables, Kernels, Tensors, Operations, Session)
Execution for Single and Multiple Devices (Node Placement Algorithm)
Data vs Model Parallel Training
Queues, Control Flow and Partial Execution
Automatic Gradient Computation
Tensorflow vs Others
A short Demo

Speaker bio

I am a Software Developer at Grofers who has spent past one year developing and managing RESTful API’s for various inventory management systems. Besides, I am really interested in machine learning (deep learning to be specific). I am a huge fan of Pandas and currently I am exploring Tensorflow.