DeepLearning.AI TensorFlow Developer
Programa especializado

Horario: Sábados de 14 a 18
Comienzo: Sábado 6 de febrero
Curso online
Duración: 4 meses
Carga horaria: 48 horas en 16 clases de cuatro horas


Arancel: matrícula AR$ 4.000 y cuatro cuotas de AR$ 4.000
Para comenzar, deben estar pagas la matrícula y las cuotas 1 y 2
En marzo se abona la cuota 3

Precio general: pago único de U$S 500

DeepLearning.AI TensorFlow Developer - PROGRAMA ESPECIALIZADO

Descubrí las herramientas para desarrolladores de software utilizadas para construir aplicaciones de Inteligencia Artificial con TensorFlow. TensorFlow es un Framework de Machine Learning muy popular, desarrollado por Google.

En esta especialización de cuatro cursos explorarás excitantes oportunidades para aplicaciones de IA.
Comenzarás desarrollando una comprensión general de cómo construir y entrenar redes neuronales.
Mejorarás el rendimiento de la red usando convoluciones mientras la entrenas para identificar imágenes del mundo real.

Le enseñarás a las máquinas cómo entender, analizar y responder a preguntas de la gente con los sistemas de Procesamiento de Lenguaje Natural.Aprenderás a procesar texto, a representar oraciones como vectores y a ingresar datos a una red neuronal. ¡Incluso entrenarás a una IA para crear poesía original!

La Inteligencia Artificial ya está transformando industrias a los largo del mundo. Después de terminar esta especialización, serás capaz de aplicar tus nuevas habilidades a un amplio rango de problemas y proyectos.

QUÉ APRENDERÁS

- Mejores prácticas en TensorFlow

- Manejar imágenes del mundo real y explorar estrategias para prevenir el sobreajuste incluyendo data
  augmentation y   dropout.

- Construir sistemas de NLP con TensorFlow.

- Aplicar RNNs, GRUs y LSTMs entrenándolas usando repositorios de texto.


CURSO 1 Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning

CURSO 2 Convolutional Neural Networks in TensorFlow

CURSO 3 Natural Language Processing in TensorFlow

CURSO 4 Sequences, Time Series and Prediction


PROGRAMA

CURSO 1 Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning

SEMANA 1 - UN NUEVO PARADIGMA DE PROGRAMACIÓN

- A primer in machine learning
- The ‘Hello World’ of neural networks
- From rules to data
- Working through ‘Hello World’ in TensorFlow and Python
- Try it for yourself
- Your First Neural Network
- Introduction to Google Colaboratory

SEMANA 2 - INTRODUCCIÓN A COMPUTER VISION
 
- Introduction to computer vision
- Exploring how to use data
- Writing code to load training data
- The structure of Fashion MNIST data
- Coding a Computer Vision Neural Network
- See how it's done
- Get hands-on with computer vision
- Callbacks to control training
- See how to implement Callbacks
- Walk through a notebook with Callbacks
- Implement a Deep Neural Network to recognize handwritten digits

SEMANA 3 - MEJORANDO LA VISIÓN CON REDES NEURONALES CONVOLUCIONALES

-  ¿What are convolutions and pooling?
-  Coding convolutions and pooling layers
-  Implementing convolutional layers
-  Learn more about convolutions
-  Implementing pooling layers
-  Getting hands-on, your first ConvNet
-  Improving the Fashion classifier with convolutions
-  Try it for yourself
-  Walking through convolutions
-  Experiment with filters and pools
-  Improving DNN Performance using Convolutions

SEMANA 4 - USANDO IMÁGENES DEL MUNDO REAL

-  Explore an impactful, real-world solution
-  Understanding ImageGenerator
-  Designing the neural network
-  Defining a ConvNet to use complex images
-  Train the ConvNet with ImageGenerator
-  Training the ConvNet with fit_generator
-  Exploring the solution
-  Walking through developing a ConvNet
-  Training the neural network
-  Walking through training the ConvNet with fit_generator
-  Experiment with the horse or human classifier
-  Adding automatic validation to test accuracy
-  Get hands-on and use validation
-  Exploring the impact of compressing images
-  Get Hands-on with compacted images
-  Exercise - Handling Complex Images

CURSO 2 Convolutional Neural Networks in TensorFlow

SEMANA 1 -  EXPLORANDO UN DATASET MÁS GRANDE

- The cats vs dogs dataset
- Training with the cats vs. dogs dataset
- Looking at the notebook
- Working through the notebook
- What you'll see next
- Fixing through cropping
- Visualizing the effect of the convolutions
- Looking at accuracy and loss
- What have we seen so far?
- Exercise- ¡Attempt the cats vs. dogs Kaggle challenge!

SEMANA 2 - DATA AUGMENTATION: UNA TÉCNICA PARA MITIGAR EL OVERFITTING

- Augmentation
- Image Augmentation
- Introducing augmentation
- Start Coding...
- Coding augmentation with ImageDataGenerator
- Looking at the notebook
- Demonstrating overfitting in cats vs. dogs
- The impact of augmentation on Cats vs. Dogs
- Adding augmentation to cats vs. dogs
- ¡Try it for yourself!
- Exploring augmentation with horses vs. humans
- ¿What have we seen so far?
- Exercise- Full cats vs. dogs using augmentation


SEMANA 3 - TRANSFER LEARNING

- Understanding transfer learning: the concepts
- ¡Start coding!
- Coding transfer learning from the inception mode
- Adding your DNN
- Coding your own model with transferred features
- ¡Using dropouts!
- Exploring dropouts
- Applying Transfer Learning to Cats v Dogs
- Exploring Transfer Learning with Inception
- ¿What have we seen so far?
- Exercise 3 - Horses vs. humans using Transfer Learning
 
SEMANA 4 - CLASIFICACIÓN CON MÚLTIPLES CLASES

- Moving from binary to multi-class classification
- Introducing the Rock-Paper-Scissors dataset
- Explore multi-class with Rock Paper Scissors dataset
- ¡Check out the code!
- Train a classifier with Rock Paper Scissors
- Try testing the classifier
- Test the Rock Paper Scissors classifier
- ¿What have we seen so far?
- Exercise 4 - Multi-class classifier


CURSO 3 Natural Language Processing in TensorFlow

SEMANA 1 -  ANÁLISIS DE SENTIMIENTOS EN TEXTOS

- Introduction
- Word based encodings
- Using APIs
- Notebook for lesson 1
- Text to sequence
- Looking more at the Tokenizer
- Padding
- Notebook for lesson 2
- ¿Sarcasm, really?
- Working with the Tokenizer
- News headlines dataset for sarcasm detection
- Notebook for lesson 3
- Exercise 1- Explore the BBC news archive

SEMANA 2 - WORD EMBEDDINGS

- Introduction
- The IMBD dataset  
- IMDB reviews dataset
- Looking into the details
- ¿How can we use vectors?
- Try it yourself
- More into the details
- Notebook for lesson 1
- ¿Remember the sarcasm dataset?
- Building a classifier for the sarcasm dataset
- Let’s talk about the loss function
- Pre-tokenized datasets
- TensorFlow datasets
- Diving into the code (part 1)
- Subwords text encoder
- Diving into the code (part 2)
- Notebook for lesson 3
- Exercise 2- BBC news archive


SEMANA 3 - MODELOS SECUENCIALES

- Introduction
- Link to Andrew's sequence modeling course
- More info on LSTMs
- Implementing LSTMs in code
- Accuracy and loss
- A word from Laurence
- Looking into the code
- Using a convolutional network
- Going back to the IMDB dataset
- Tips from Laurence
- Exploring different sequence models
- Exercise 3- Exploring overfitting in NLP


SEMANA 4 - MODELOS SECUENCIALES Y LITERATURA

- Introduction
- Looking into the code
- Training the data
- More on training the data
- Notebook for lesson 1
- Finding what the next word should be
- Example
- Predicting a word
- ¡LecciónPoetry!
- Link to Laurence's poetry
- Looking into the code
- ¡Laurence the poet!
- Your next task
- Link to generating text using a character-based RNN
- Exercise 4- Using LSTMs, see if you can write Shakespeare!

CURSO 4 Sequences, Time Series and Prediction

SEMANA 1 -  SECUENCIAS Y PREDICCIONES

- Introduction
- Sequences and Prediction
- Time series examples
- Machine learning applied to time series
- Common patterns in time series
- Introduction to time series
- Introduction to time series notebook
- Train, validation and test sets
- Metrics for evaluating performance
- Moving average and differencing
- Trailing versus centered windows
- Forecasting
- Forecasting notebook
- Exercise 1 - Create and predict synthetic data


SEMANA 2 - REDES NEURONALES PROFUNDAS PARA SERIES DE TIEMPO

- Preparing features and labels
- Preparing features and labels notebook
- Sequence bias
- Feeding windowed dataset into neural network
- Single layer neural network
- Machine learning on time windows
- Prediction
- More on single layer neural network
- Single layer neural network notebook
- Deep neural network training, tuning and prediction
- Deep neural network
- Deep neural network notebook
- Create Synthetic Data and predict with a DNN
- Exercise 2 - Predict with a DNN

SEMANA 3 - REDES NEURONALES RECURRENTES PARA SERIES DE TIEMPO

- Conceptual overview
- Shape of the inputs to the RNN
- Outputting a sequence
- Lambda layers
- Adjusting the learning rate dynamically
- More info on Huber loss
- RNN
- RNN notebook
- LSTM
- Link to the LSTM lesson
- Coding LSTMs
- More on LSTM
- LSTM notebook
- Exercise 3 - Mean Absolute Error

SEMANA 4 - SERIES DE TIEMPO - DATOS DEL MUNDO REAL

- Convolutions
- Convolutional neural networks course
- Bi-directional LSTMs
- More on batch sizing
- LSTM
- LSTM notebook
- Real data - sunspots
- Train and tune the model
- Prediction
- Sunspots
- Sunspots notebook
- Combining our tools for analysis
- Exercise 4 - Sunspots

Congratulations!
TensorFlow in practice has come to an end
Specialization wrap up - A conversation with Andrew Ng