Session IV: Getting Started with TensorFlow

The developers were introduced to the three ways on how to get started with different Machine Learning frameworks. These include using ready to use APIs, using existing models, and retraining it with developers data. They were also introduced to how they can use tensorflow to develop algorithms.This is a general datatype that hold numbers and what it means for them to flow.

I explained how the tensors are being computed and how they flow through a graph of operations such as addition and multiplication. The Graphs are defined in high level language such as Python, are compiled and then optimized. They are then executed on available low level CPUs and GPUs.We also advise them that they could accelerate TensorFlow performance on Intel Processors.

I introduced TensorFlow as the most popular open source library for machine intelligence on github. It is used by researchers, data scientist and developers. It is capable of large scale ML and production.

I then gave a practical example of how an image is recognized using deep neural network. Instead of writing rules to recognize images, developers can write algorithms that find patterns in data.

This example classifiers images as cats or dogs. The input to the network is the raw pixels and the output of the network is a prediction for whether the image is a cat or dog. The layers are composed of artificial neurons. Each neuron has a weight. Initially they are random and the network won’t classify images. We adjust the weights using an algorithm and backpropagation.

We also showed them examples of where machine learning is used in real life. This include:

. Faces, facial landmarks, emotions

. Label detection

. Logo detection

. Text Detection

We then gave them useful resources where they can learn more. These include the many great tutorials at the tensorflow.org and TensorFlow for Poets codelabs.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.