Getting started with
TensorFlow
Gustavo García
@anarchyco
Machine Learning is everywhere
TensorFlow is an open source library for numerical
computation, specializing in machine learning
applications
TensorFlow Hello World
Machine Learning Pseudocode
labeled_data = ... // f.e. X: square_meters -> y: house_prices
model = ... // f.e. y = B * x + A
// Training
do N times: // epochs
for d in labeled_data: // usually in batches
y_predicted = model.predict(d)
model.adjust(y_predicted - y, learning_rate)
// Inference
model.predict(unseen_data)
Deep Neural Network with TensorFlow
Deep Neural Network with TensorFlow
Deep Neural Network with TensorFlow
Deep Neural Network with TensorFlow
Deep Neural Network with TensorFlow
Deep Neural Network with TensorFlow
Real world TensorFlow
1. Find a model to the type of problem you want to solve
2. Obtain as much good data as possible
3. Train the model (or retrain)
4. Save the model and deploy it
Demo
The End

Getting started with TensorFlow