O'Reilly logo

Stay ahead with the world's most comprehensive technology and business learning platform.

With Safari, you learn the way you learn best. Get unlimited access to videos, live online training, learning paths, books, tutorials, and more.

Start Free Trial

No credit card required

Troubleshooting Python Deep Learning

Video Description

Practical solutions to your problems while building Deep Learning models using CNN, LSTM, scikit-learn, and NumPy

About This Video

  • Discover the limitless use of building any application using Deep Learning and ensure its issues aren’t a roadblock for your projects
  • Problems are addressed with practical yet unique solutions that are easy to understand and implement
  • Implement scikit-learn and NumPy, to resolve the common problems arising from Deep Learning models

In Detail

Building Deep Learning models with Python is a strenuous task and there are chances of getting stuck on specific tasks. When that happens, you usually end up searching for solutions and need to manually look for ways to resolve these problems. This wastes both time and effort, and may also lead to reduced performance of your Deep Learning system.

After carefully analyzing the most popular errors or problems that arise while working on Deep Learning models, we have identified the most usable models used for classification in this course and provided practical yet unique solutions to each problem that are easy to understand and implement.

You can either follow the entire course or directly jump into the section that covers a specific problem you’re facing. Some of the common yet important issues we cover include errors while building and training Deep Learning with neural networks, especially without a specific framework.

By the end of the course, you will be well-versed to tackle and troubleshoot any errors with your Deep Learning models.

Downloading the example code for this course: You can download the example code files for this course on GitHub at the following link: https://github.com/PacktPublishing/Troubleshooting-Python-Deep-Learning. If you require support please email: customercare@packt.com

Table of Contents

  1. Chapter 1 : Solutions to Convolutional Neural Network Problems – Part One
    1. The Course Overview 00:04:45
    2. Concatenate Two CNNs Correctly 00:06:01
    3. Splitting Trained Model 00:07:05
    4. Resolving fit_generator Errors 00:03:53
    5. Model Object Has No Attribute load_model Keras 00:01:31
    6. High val_acc, But Low Accuracy in Practice 00:05:49
    7. Error in Adding a Dense Layer 00:02:30
    8. Model with Multiple Output Errors 00:01:58
    9. Model That Uses Dropout Is Still Overfitting 00:04:20
  2. Chapter 2 : Solutions to Convolutional Neural Network Problems – Part Two
    1. When the Value Error Input 0 Is Incompatible with Layer conv2d_1 00:03:11
    2. Interpreting kernel_size Notation in CNNs 00:05:33
    3. Choosing Last Layer’s Activation Function in CNN 00:03:48
    4. Using Validation Accuracy 00:04:20
    5. Error When Using CNN to Classify Text 00:04:13
    6. Kernel Weight Initialization in CNN Model 00:02:40
    7. Common Problems When Using Pre-Trained CNN Models 00:03:50
    8. Shape Error When Training CIFAR-10 Dataset on CNN 00:04:13
  3. Chapter 3 : Solutions to Recurrent Neural Network Problems
    1. Building an RNN Model in Keras 00:03:48
    2. Wrong Input: ValueError – Error When Checking Input 00:06:19
    3. Correct Text Preparation for Machine Translation 00:05:13
    4. Handling Invalid Input Shape Error 00:03:48
    5. Mapping Series of Vectors to a Single Vector 00:03:30
    6. Resolving a Bad Output from RNN While Generating a Simple Sequence 00:02:58
    7. Preparing Data Correctly for Time Series Prediction 00:04:52
    8. How to Enable Stateful RNN? 00:03:32
  4. Chapter 4 : Solutions to LSTM Recurrent Neural Networks Problems
    1. Stacking Multiple LSTM in Keras TypeError: Call() Got an Unexpected Keyword Argument 'return_sequences' 00:04:19
    2. Working with Different Lengths of Input and Output Sequences 00:07:25
    3. How to Use Stacked LSTMs? 00:03:08
    4. Using CNN-LSTM for Time Series Prediction 00:04:12
    5. Solving LSTM Underfitting on Time Series Problem 00:02:57
    6. Using LSTM for Multi-Value Prediction 00:02:22
    7. How To Do Text Classification with LSTM? 00:05:40
    8. Data Preparation for Seq2Seq Learning 00:03:38
  5. Chapter 5 : Troubleshooting Models with scikit-learn
    1. LabelBinarizer Returns Vector When There Are Two Classes 00:03:42
    2. Handling Missing Values 00:06:35
    3. Evaluating Deep Learning Models Using Additional Metrics 00:04:09
    4. Fixing Warning Messages 00:05:13
    5. Generating Test Datasets 00:02:58
    6. Normalizing and Standardizing the Data 00:03:30
    7. Preparing Text for Use with Deep Learning Models 00:03:57
  6. Chapter 6 : Solving NumPy Problems
    1. Converting a 2D Matrix to a One-Hot Encoded Matrix 00:04:21
    2. Reshaping a 2D NumPy Array to 3D Array 00:02:24
    3. Fix load.npy Error in Python3 00:02:42
    4. Turn ND Matrix into 1D Vector 00:05:12