Developing AI Applications

Name: Developing AI Applications | An Introduction
Brand: Packt Publishing
Price: 35.99 EUR
Availability: OnlineOnly

An Introduction

Metin Karatas(Autor*in)

Packt Publishing

1. Auflage

Erschienen am 18. Juni 2025

405 Seiten

E-Book

ePUB mit Adobe-DRM

Systemvoraussetzungen

978-1-80602-248-9 (ISBN)

35,99 €inkl. 7% MwSt.

Systemvoraussetzungen

für ePUB mit Adobe-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

Weitere Details

Inhalt

Cover
Contents
1: Introduction
1.1 What Does This Book Offer?
1.2 What Is Artificial Intelligence?
1.3 The History of AI: A Brief Overview
1.4 Development Tools Used in This Book
1.4.1 Python
1.4.2 Jupyter Notebook
1.4.3 KNIME
1.4.4 ChatGPT and GPT-4
1.4.5 DALL-E 2 or DALL-E 3
2: Installation
2.1 Anaconda Distribution
2.1.1 Windows and macOS
2.1.2 Linux
2.1.3 Configuration and Test
2.2 KNIME
2.2.1 Installation
2.2.2 Configuration
2.2.3 Test
3: Artificial Neural Networks
3.1 Classification
3.2 The Recipe
3.2.1 Data Preparation
3.2.2 Building Up the AI
3.2.3 Training the AI
3.2.4 Testing the AI
3.2.5 Using AI
3.3 Building ANNs
3.4 Structure of an Artificial Neuron
3.5 Feed Forward
3.6 Back Propagation
3.7 Updating the Weights
3.8 ANN for Classification
3.9 Hyperparameters and Overfitting
3.10 Dealing with Nonnumerical Data
3.11 Dealing with Data Gaps
3.11.1 Filling Empty Cells with Data
3.11.2 Removing Rows with Empty Cells
3.12 Correlation versus Causality
3.13 Standardization of the Data
3.14 Regression
3.15 Deployment
3.15.1 Training, Testing, and Saving
3.15.2 Using the ANN Model
3.16 Exercises
3.16.1 Exercise 1: Hyperparameter Optimization for Classification
3.16.2 Exercise 2: Hyperparameter Optimization for Regression
3.16.3 Exercise 3: ANN for Classification
3.16.4 Exercise 4: ANN for Regression
4: Decision Trees
4.1 Simple Decision Trees
4.1.1 Decision Tree Classifier
4.1.2 Decision Tree Regressor
4.1.3 Decision Forests
4.1.4 Random Forest Classifier
4.1.5 Random Forest Regressor
4.2 Boosting
4.2.1 Gradient Boosting
4.2.2 XGBoost Classifier
4.2.3 Automatic Hyperparameter Setting Using GridSearchCV
4.3 XGBoost Regressor
4.4 Deployment
4.5 Decision Trees Using Orange
4.6 Exercises
4.6.1 Exercise 1: XGBoost for Classification
4.6.2 Exercise 2: XGBoost for Regression
4.6.3 Exercise 3: Automatic Hyperparameter Optimization
5: Convolutional Layers and Images
5.1 Simple Image Classification
5.2 Hyperparameter Optimization Using Early Stopping and KerasTuner
5.3 Convolutional Neural Network
5.4 Image Classification Using CIFAR-10
5.5 Using Pretrained Networks
5.6 Exercises
5.6.1 Exercise 1: Hyperparameter Optimization for CIFAR-10
5.6.2 Exercise 2: Pretrained VGG19 Model
6: Transfer Learning
6.1 How It Works
6.2 Exercises
6.2.1 Exercise 1: Rock-Paper-Scissors
6.2.2 Exercise 2: Human or Horse
7: Anomaly Detection
7.1 Unbalanced Data
7.2 Resampling
7.3 Autoencoders
7.4 Exercises
7.4.1 Exercise 1: Anomaly Detection Using XGBoost and Upsampling
7.4.2 Exercise 2: Anomaly Detection Using an Autoencoder
8: Text Classification
8.1 Embedding Layer
8.2 GlobalAveragePooling1D Layer
8.3 Text Vectorization
8.4 Analysis of the Relationships
8.5 Classifying Large Amounts of Data
8.6 Exercises
8.6.1 Exercise 1: Hyperparameter Optimization
8.6.2 Exercise 2: Text Classification
8.6.3 Exercise 3: Text Classification Using Upsampling
9: Cluster Analysis
9.1 Graphical Analysis of the Data
9.2 The k-Means Clustering Algorithm
9.3 The Finished Program
9.4 Exercises
9.4.1 Exercise 1: Grouping of Diamonds
9.4.2 Exercise 2: Grouping of Mushrooms
10: AutoKeras
10.1 Classification
10.2 Regression
10.3 Image Classification
10.4 Text Classification
10.5 Exercises
10.5.1 Exercise 1: Classification
10.5.2 Exercise 2: Regression
10.5.3 Exercise 3: Image Classification
10.5.4 Exercise 4: Text Classification
11: Visual Programming Using KNIME
11.1 Simple ANNs
11.1.1 Classification
11.1.2 Classification Using Python Node
11.1.3 Regression
11.1.4 Regression Using Python Node
11.2 XGBoost
11.2.1 Classification
11.2.2 Deployment
11.2.3 Regression
11.3 Image Classification Using a Pretrained Model
11.3.1 Image Classification Using Keras Node
11.3.2 Image Classification Using Python Node
11.4 Transfer Learning
11.4.1 Transfer Learning Using Keras Node
11.4.2 Transfer Learning Using Python Node
11.5 Autoencoder
11.5.1 Autoencoder with Keras Node
11.5.2 Autoencoder with Python Node
11.6 Text Classification
11.6.1 Text Classification with Keras Node
11.6.2 Text Classification with Python Node
11.7 AutoML
11.7.1 Installation
11.7.2 Classification
11.8 Cluster Analysis
11.8.1 Manual Cluster Setting
11.8.2 Cluster Setting with a Loop
11.9 Time Series Analysis
11.9.1 Recurrent Neural Networks
11.9.2 Long Short-Term Memory
11.9.3 Prediction of Energy Consumption (Next Hour) Using Keras Node
11.9.4 Prediction of Energy Consumption (Next Hour) Using Python Node
11.9.5 Prediction of Energy Consumption (Next 500 Hours) Using Keras Node
11.9.6 Prediction of Energy Consumption (Next 500 Hours) Using Python Node
11.10 Text Generation
11.10.1 Data Preparation
11.10.2 Trainings
11.10.3 Generation
11.11 Further Information on KNIME
11.12 Exercises
11.12.1 Exercise 1: XGBoost for Classification, Mushrooms
11.12.2 Exercise 2: XGBoost for Regression, Diamonds
11.12.3 Exercise 3: Image Classification Using InceptionV3
11.12.4 Exercise 4: Transfer Learning, Horses or Humans
11.12.5 Exercise 5: Anomaly Detection Using an Autoencoder, ECG
11.12.6 Exercise 6: Text Classification
11.12.7 Exercise 7: AutoML for Regression
11.12.8 Exercise 8: Cluster Analysis
11.12.9 Exercise 9: Time Series Analysis
11.12.10 Exercise 10: Text Generation
12: Reinforcement Learning
12.1 Q-Learning
12.2 Python Knowledge Required for the Game
12.2.1 Lists
12.2.2 Branches
12.2.3 Loops
12.2.4 Random Choice
12.2.5 Functions
12.3 Trainings
12.4 Test
12.5 Outlook
12.6 Exercises
12.6.1 Exercise 1: Hyperparameters
12.6.2 Exercise 2: Expansion of the Game
13: Genetic Algorithms
13.1 The Algorithm
13.1.1 Start Generation
13.1.2 Selection
13.1.3 Reproduction
13.1.4 Mutation
13.1.5 New Generation
13.2 Example of a Sorted List
13.3 Example of Equation Systems
13.4 Real-Life Sample Application
13.5 Exercises
13.5.1 Exercise 1: Hyperparameter Optimization
13.5.2 Exercise 2: System of Equations
14: ChatGPT and GPT-4
14.1 Prompt Engineering
14.1.1 Generating Content
14.1.2 Programming
14.1.3 Analyzing and Summarizing
14.1.4 Final Questions for ChatGPT
14.2 The ChatGPT Programming Interface
14.2.1 Application Programming Interface Key and First Program
14.2.2 Parameters
14.2.3 Input Filters
14.2.4 Roles
14.2.5 Memory
14.2.6 User Profiles
14.2.7 Playground
14.2.8 Speech to Text
14.3 Exercise 1: Math Support
15: DALL-E and Successor Models
15.1 DALL-E 2
15.1.1 Prompt Engineering
15.1.2 Editing Generated Images
15.2 DALL-E 3
15.3 Programming Interface
15.3.1 Image Creation
15.3.2 Image Variations
15.3.3 Image Processing
15.4 Exercise 1: DALL-E API with Moderation
16: Outlook
Appendices
A: Exercise Solutions
A.1 Chapter 3
Exercise 1: Hyperparameter Optimization for Classification
Exercise 2: Hyperparameter Optimization for Regression
Exercise 3: ANN for Classification
Exercise 4: ANN for Regression
A.2 Chapter 4
Exercise 1: XGBoost for Classification
Exercise 2: XGBoost for Regression
Exercise 3: Automatic Hyperparameter Optimization
A.3 Chapter 6
Exercise 1: Rock-Paper-Scissors
Exercise 2, Part 1: Human or Horse, Training and Testing
Exercise 2, Part 2: Human or Horse, Application
A.4 Chapter 7
Exercise 1: Anomaly Detection Using XGBoost and Upsampling
Exercise 2: Anomaly Detection Using an Autoencoder
A.5 Chapter 8
Exercise 1: Hyperparameter Optimization
Exercise 2: Text Classification
Exercise 3: Text Classification Using Upsampling
A.6 Chapter 9
Exercise 1: Grouping of Diamonds
Exercise 2: Grouping of Mushrooms
A.7 Chapter 10
Exercise 1: Classification
Exercise 2: Regression
Exercise 3: Image Classification
Exercise 4: Text Classification
A.8 Chapter 11
Exercise 1: XGBoost for Classification, Mushrooms
Exercise 2: XGBoost for Regression, Diamonds
Exercise 3: Image Classification Using InceptionV3
Exercise 4: Transfer Learning, "Human or Horse"
Exercise 5: Anomaly Detection Using an Autoencoder: ECG
Exercise 6: Text Classification
Exercise 7: AutoML for Regression
Exercise 8: Cluster Analysis
Exercise 9: Time Series Analysis
Exercise 10: Text Generation
A.9 Chapter 12
Exercise 1: Hyperparameters
Exercise 2: Expansion of the Game
A.10 Chapter 13
Exercise 1: Hyperparameter Optimization
Exercise 2: System of Equations
A.11 Chapter 14
Exercise 1: Math Support
A.12 Chapter 15
Exercise 1: DALL-E API with Moderation
B: References
C: The Author
Index

1.4 Development Tools Used in This Book

Many programming languages, libraries, and tools are available for developing AIs. Here is a brief introduction to the selection used in this book. Nothing is installed or programmed yet.

1.4.1 Python

The Python programming language is both very easy to learn and very powerful. Even during development, great importance was attached to readability, platform independence, and simple structures. Even if you program on Windows, the program can then be run on a Linux computer or macOS. These advantages have led to it becoming one of the most popular programming languages of all. In the world of data science and AI, Python has become the de facto standard. As already mentioned, this book isn't intended as an introduction to the Python programming language. For us, it's only a means to an end. The following Python statements are intended to illustrate the simplicity. We won't start programming until the installation has been completed in the next chapter.

The standard "Hello World" example looks like this:

print("Hello World!")

The print function can be used to output data to the console, in this case, "Hello World!". We can also save this character combination (string data type) in a variable first and then output it:

output = "Hello World!"
print(output)

Data Types

The most elementary data types in Python are as follows:

Strings (character strings), which are written in single or double quotation marks (e.g., "Hello World!")
Integers (e.g., 72)
Floats (floating-point numbers, e.g., 3.14)

Certain operations are possible depending on the data type. String data such as "Hello" and "World" can be combined to form "Hello World". Integer or float data can be offset against each other, for example, 3 × 8.

We'll only introduce other data types when they are required to solve a specific task.

The same variable can also be reused for an integer (integer data type) or floating-point number (float data type). Python variables can store objects of any type:

output = "Hello World!"
print(output)

output = 42
print(output)

output = 42.2
print(output)

Listing 1.1 Variables with Python

Here, "Hello World!", 42, and 42.2 are output one after the other. Now let's take a look at a supposedly more complicated program:

# Setting the age to 20
age = 20
# Query whether the person is of legal age
if age >= 18:
print("You are of legal age!")
else:
print("You are not yet of legal age!")

Listing 1.2 If-Else Structure with Python

The # character introduces comments that can help when reading the code, and these lines aren't executed. The number 20 is stored in the age variable. You'll then be asked whether the age is greater than or equal to 18. If that is the case (as in this example), the corresponding output is generated. The else branch is skipped, and the program is ended. Let's clarify the following questions:

What effect would it have if the value 16 were stored in the age variable? Answer: The if branch would be skipped, and the else branch would be executed. The output would read "You are not yet of legal age!"
In some countries you're only of legal age at 21. What changes are necessary to this program? Answer: In the if query, the comparison is made with the number 21 instead of 18.
Can the change to the age of majority query be described as "programming" or as a "configuration"? You can argue about that if you have nothing better to do.

As you can see, it's quite easy to trace the source code in Python.

1.4.2 Jupyter Notebook

We need software into which we can type the Python commands. This software is used for programming, execution, and testing. Software of this type is called an integrated development environment (IDE). There are numerous very good IDEs available for Python. However, simple Jupyter Notebooks (open source) have established themselves in data science and AI development. The program is programmed in the browser and usually consists of a single file. To install and configure Python and Jupyter Notebook (and many other tools if required), we'll use the free Anaconda distribution, which was developed for this purpose. It's available for Windows, Linux and macOS. Using Anaconda you can set up simple development environments locally on a PC or centrally on a server. The advantage of the server version is that you can easily program in Jupyter Notebook from multiple clients via a browser. No further installation is necessary on these clients. However, if instead of using your own server, you want to use one provided by Anaconda in the cloud, you'll have to pay for it.

Development Environment

An IDE provides important tools that support you in programming. This includes the automatic recognition and completion of commands, auto-correction, and helpful hints in the event of programming errors.

1.4.3 KNIME

The KNIME Analytics Platform software takes a completely different approach in that programming is carried out using graphical blocks, which are dragged and dropped into the workspace, linked together, and configured. This is also known as visual programming.

KNIME is based on Eclipse, a widely used development environment for various programming languages. It's open source, free of charge, and platform-independent.

The following example shows a program that reads the contents of an Excel file and plots the numerical values in a graph (see Figure 1.1).

You can see the output of the Line Chart module in Figure 1.2. For this purpose, right-click on the block after executing it, and then click on View: Line Chart.

If the data source isn't an Excel file, but instead a database or comma-separated values (CSV) file, for example, you only need to replace the Excel Reader module.

Figure 1.1 Loading an Excel File and Plotting the Contents

Figure 1.2 Plotting the Contents

1.4.4 ChatGPT and GPT-4

Let's just ask ChatGPT (or its successor GPT-4) itself who or what it is:

User: Who or what are you? Explain it to me as if I were 12 years old.

Answer: Of course! Imagine I'm like an extremely advanced computer program that can answer questions. You can think of me a bit like an interactive encyclopedia or a digital interlocutor who knows a lot of information.

I am not really a person and have no thoughts or feelings of my own. I am a tool created by humans to answer questions and provide information based on what was known until 2021.

You can try out this chatbot from OpenAI at https://chat.openai.com. We'll take a closer look at this impressive tool later on in this book.

1.4.5 DALL-E 2 or DALL-E 3

DALL-E 2 and its successor DALL-E 3 are based on the ChatGPT model, which has been adapted for text-to-image synthesis. The name is derived from Salvador Dali (important surreal artist) and WALL-E (Pixar film robot).

You can use this tool (https://labs.openai.com) to generate images from textual descriptions (see Figure 1.3).

User: The image of a programmer sitting on an airplane and programming on a laptop.

Figure 1.3 Image Generated by DALL-E 3

...

Systemvoraussetzungen

Als PDF speichern Als Link merken