I.       Overview:

This training focuses on advanced machine learning topics using Google Cloud Platform where you will get hands-on experience optimizing, deploying, and scaling production ML models of various types in hands-on labs. This training picks up where “Machine Learning on Google Cloud Platform” left off and teaches you how to build scalable, accurate, and production-ready models for structured data, image data, time-series, and natural language text. It ends with a module on building recommendation systems.

II.     Duration: 05 days

III. Objectives

• Implement the various flavors of production ML systems—static, dynamic, and continuous training; static and dynamic inference; and batch and online processing.
• Solve an ML problem by building an end-to-end pipeline, going from data exploration, preprocessing, feature engineering, model building, hyperparameter tuning, deployment, and serving.
• Develop a range of image classification models from simple linear models to high-performing convolutional neural networks (CNNs) with batch normalization, augmentation, and transfer learning.
• Forecast time-series values using CNNs, recurrent neural networks (RNNs), and LSTMs.
• Apply ML to natural language text using CNNs, RNNs, LSTMs, reusable word embeddings, and encoder-decoder generative models.
• Implement content-based, collaborative, hybrid, and neural recommendation models in TensorFlow.

IV. Intended Audience

• Data Engineers and programmers interested in learning how to apply machine learning in practice.
• Anyone interested in learning how to leverage machine learning in their enterprise.

V.    Prerequisites

To get the most out of this training, participants should have:

• Knowledge of machine learning and TensorFlow to the level covered in Machine Learning on Google Cloud Platform coursework.
• Experience coding in Python.
• Knowledge of basic statistics.
•  Knowledge of SQL and cloud computing (helpful).

VI.  Outline

Module 1: Machine Learning on Google Cloud Platform

•  Effective ML.
•  Fully managed ML.

Module 2: Explore the Data

•  Exploring the dataset.
•  BigQuery.
•  BigQuery and AI Platform Notebooks.

Module 3: Creating the dataset

•  Creating a dataset.
• Module 4: Build the Model
•  Build the model.

Module 5: Operationalize the model

•  Operationalizing the model.
•  Cloud AI Platform.
•  Train and deploy with Cloud AI Platform.
•  BigQuery ML.
•  Deploying and Predicting with Cloud AI Platform.

Module 6: Architecting Production ML Systems

•  The Components of an ML System.
•  The Components of an ML System: Data Analysis and Validation.
•  The Components of an ML System: Data Transformation + Trainer.
•  The Components of an ML System: Tuner + Model Evaluation and Validation.
•  The Components of an ML System: Serving.
•  The Components of an ML System: Orchestration + Workflow.
•  The Components of an ML System: Integrated Frontend + Storage.
•  Training Design Decisions.
•  Serving Design Decisions.
•  Designing from Scratch.

Module 7: Ingesting data for Cloud-based analytics and ML

•  Data On-Premise.
•  Large Datasets.
•  Data on Other Clouds.
•  Existing Databases.

Module 8: Designing Adaptable ML systems

•  Adapting to Data.
•  Changing Distributions.
•  Right and Wrong Decisions.
•  System Failure.
•  Mitigating Training-Serving Skew through Design.
•  Debugging a Production Model.

Module 9: Designing High-performance ML systems

•  Training.
•  Predictions.
•  Why distributed training?
•  Distributed training architectures.
•  Faster input pipelines.
•  Native TensorFlow Operations.
•  TensorFlow Records.
•  Parallel pipelines.
•  Data parallelism with All Reduce.
•  Parameter Server Approach.
•  Inference.

Module 10: Hybrid ML systems

•  Machine Learning on Hybrid Cloud.
•  KubeFlow.
•  Embedded Models.
•  TensorFlow Lite.
•  Optimizing for Mobile.

Module 11: Welcome to Image Understanding with TensorFlow on GCP

•  Images as Visual Data.
•  Structured vs Unstructured Data.

Module 12: Linear and DNN Models

•  Linear Models.
•  DNN Models Review.
•  Review: What is Dropout?

Module 13: Convolutional Neural Networks (CNNs)

•  Understanding Convolutions.
•  CNN Model Parameters.
•  Working with Pooling Layers.
•  Implementing CNNs with TensorFlow.

Module 14: Dealing with Data Scarcity

•  The Data Scarcity Problem.
•  Data Augmentation.
•  Transfer Learning.
•  No Data, No Problem.

Module 15: Going Deeper Faster

•  Batch Normalization.
•  Residual Networks.
•  Accelerators (CPU vs GPU, TPU).
•  TPU Estimator.
•  Neural Architecture Search.

Module 16: Pre-built ML Models for Image Classification

•  Pre-built ML Models.
•  Cloud Vision API.
•  AutoML Vision.
•  AutoML Architecture.

Module 17: Working with Sequences

•  Sequence data and models.
•  From sequences to inputs,
•  Modeling sequences with linear models.
•  Modeling sequences with DNNs.
•  Modeling sequences with CNNs.
•  The variable-length problem4m.

Module 18: Recurrent Neural Networks

•  Introducing Recurrent Neural Networks.
•  How RNNs represent the past.
•  The limits of what RNNs can represent.
•  The vanishing gradient problem.

Module 19: Dealing with Longer Sequences

•  LSTMs and GRUs.
•  RNNs in TensorFlow.
•  Deep RNNs.
•  Improving our Loss Function.
•  Working with Real Data.

Module 20: Text Classification

•  Working with Text.
•  Text Classification.
•  Selecting a Model.
•  Python vs Native TensorFlow.

Module 21: Reusable Embeddings

•  Historical methods of making word embeddings.
•  Modern methods of making word embeddings.
•  Introducing TensorFlow Hub.
•  Using TensorFlow Hub within an estimator.

Module 22: Recurrent Neural NetworksEncoder-Decoder Models

•  Introducing Encoder-Decoder Networks.
•  Attention Networks.
•  Training Encoder-Decoder Models with TensorFlow.
•  Introducing Tensor2Tensor.
•  AutoML Translation.
•  Dialogflow.

Module 23: Recommendation Systems Overview

•  Types of Recommendation Systems.
•  Content-Based or Collaborative.
•  Recommendation System Pitfalls.

Module 24:Content-Based Recommendation Systems

•  Content-Based Recommendation Systems.
•  Similarity Measures.
•  Building a User Vector.
•  Making Recommendations Using a User Vector.
•  Making Recommendations for Many Users.
•  Using Neural Networks for Content-Based Recommendation Systems.

Module 25:Collaborative Filtering Recommendation Systems

•  Types of User Feedback Data.
•  Embedding Users and Items.
•  Factorization Approaches.
•  The ALS Algorithm.
•  Preparing Input Data for ALS.
•  Creating Sparse Tensors For Efficient WALS Input.
•  Instantiating a WALS Estimator: From Input to Estimator.
•  Instantiating a WAL Estimator: Decoding TFRecords.
•  Instantiating a WALS Estimator: Recovering Keys.
•  Instantiating a WALS Estimator: Training and Prediction.
•  Issues with Collaborative Filtering.
•  Cold Starts.

Module 26:Neural Networks for Recommendation Systems

•  Hybrid Recommendation System.
•  Context-Aware Recommendation Systems.
•  Context-Aware Algorithms.
•  Contextual Postfiltering.
•  Modeling Using Context-Aware Algorithms.

Module 27:Building an End-to-End Recommendation System

•  Architecture Overview.
•  Cloud Composer Overview.
•  Cloud Composer: DAGs.
•  Cloud Composer: Operators for ML9.
•  Cloud Composer: Scheduling.
•  Cloud Composer: Triggering Workflows with Cloud Functions.
•  Cloud Composer: Monitoring and Logging.
• On demand
• Take this course on demand
• Upcoming classrooms
• There are no upcoming instructor-led sessions