Machine Learning Adoption in Blockchain-Based Intelligent Manufacturing
Theoretical Basics, Applications, and Challenges
Om Prakash Jena, Sabyasachi Pramanik, Ahmed A. Elngar, Om Prakash Jena, Sabyasachi Pramanik, Ahmed A. Elngar
- 192 pages
- English
- ePUB (mobile friendly)
- Available on iOS & Android
Machine Learning Adoption in Blockchain-Based Intelligent Manufacturing
Theoretical Basics, Applications, and Challenges
Om Prakash Jena, Sabyasachi Pramanik, Ahmed A. Elngar, Om Prakash Jena, Sabyasachi Pramanik, Ahmed A. Elngar
About This Book
This book looks at industry change patterns and innovations (such as artificial intelligence, machine learning, big data analysis, and blockchain support and efficiency technology) that are speeding up industrial transformation, industrial infrastructure, biodiversity, and productivity.
This book focuses on real-world industrial applications and case studies to provide for a wider knowledge of intelligent manufacturing. It also offers insights into manufacturing, logistics, and supply chain, where systems have undergone an industrial transformation. It discusses current research of machine learning along with blockchain techniques that can fill the gap between research and industrial exposure. It goes on to cover the effects that the Fourth Industrial Revolution has on industrial infrastructures and looks at the current industry change patterns and innovations that are accelerating industrial transformation activities.
Researchers, scholars, and students from different countries will appreciate this book for its real-world applications and knowledge acquisition. This book targets manufacturers, industry owners, product developers, scientists, logistics, and supply chain engineers.
- Focuses on real-world industrial applications and case studies to provide for a wider knowledge of intelligent manufacturing
- Offers insights into manufacturing, logistics, and supply chain where systems have undergone an industrial transformation
- Discusses current research of machine learning along with blockchain techniques that can fill the gap between research and industrial exposure
- Covers the effects that the 4th Industrial Revolution has on industrial infrastructures
- Looks at industry change patterns and innovations that are speeding up industrial transformation activities
Frequently asked questions
Information
1 Integration of Big Data, Machine Learning, and Blockchain Technology
CONTENTS
1.1 INTRODUCTION
DL Algorithms | Summary |
---|---|
Convolutional Neural Networks (CNNs)(Albawi, Mohammed, and Al-Zawi 2017) | The input is usually an image but can be speech or any other data to be analyzed. The data passes through the convolutional layers that do most of the computations and extract the high-level features convolving the input with appropriate filters. Pooling layers are employed to reduce the dimensions. |
Recurrent Neural Networks (RNNs)(Mikolov et al. 2011) | Employed when the output depends on the current input and the previous outputs. Hence, each neuron is associated with a feedback loop. The hidden layers thus have a memory component associated with them where they store the output locally. The training is done based on an algorithm called backpropagation through time. |
Long Short-Term Memory (LSTM)(Tang et al. 2016) | Special RNNs that overcome the problem of long dependencies in RNNs. The usage of gatesâthe forget gate (decides which data to be kept or eliminated), read gate (regulates which neurons can read the content), write gate (to write information)âregulates the information. Like RNNs, backpropagation through a time algorithm is used to train the network. |
Generative Adversarial Networks (GANs)(Pan et al. 2019) | It comprises two neural networks that contest against each other: generative and discriminative. The generative network delivers new images/data based on the training data set, and the discriminative model assesses them. The job of the generative network is to create images that match the training data set, and the job of the discriminative network is to differentiate between the actual image and the forged one. |
AutoEncoders (AEs)(Roche et al. 2019) | AEs aim to reconstruct the input. They have the same number of layers in the input and output and in between are connected by a set of hidden layers. The data from the input layer are encoded, and a new representation is created, from which the decoding is done to reconstruct the original image to be produced as the output. |
Restricted Boltzmann Machine (RBM)(Zhang et al. 2018) | These are similar to AEs as they also reconstruct the input from latent variables. There are two types of units in an RBM: hidden units and visible units. The neurons in an RBM must produce a bipartite graph. Each visible unit must be connected to all hidden units and vice versa, but none should be connected to its kind. RBMs are stacked and converted into deep neural networks and serve as the building blocks of deep belief networks. |
- 1.2. Big data analytics
- 1.3. Big data and ML
- 1.4. ML and blockchain
- 1.5. Big data and blockchain
- 1.6. Big data, ML, and blockchain technology
- 1.7. Blockchain- and ML-based supply chain management
- 1.8. Benefits of MLâblockchain integration
- 1.9. Challenges faced in MLâblockchain integration
1.2 BIG DATA ANALYTICS
- Data acquisition: Users can obtain or create data by primary sources or secondary sources that are open and democratized. Some open sources of data include Google Public Data Explorer, Kaggle, data.gov, World Bank open data, and the UCI ML repository.
- Data storage: The storage of data is met with challenges, for example, the limited capabilities of the usersâ systems force the adoption of cloud services. Apache Hadoop framework provides HDFS (Hadoop Distributed File System) for distributed storage of data. Other databases include Cassandra (distributed and fault-tolerant), MongoDB (for flexible storage of data), and Neo4g (stores data as key-value pairs).
- Preprocessing the data: Preprocessing stage comprises the cleaning of data (filtering of noisy data, handling missing data), data transformation, and removing redundancies. Data are compressed, and dimensionality reduction is applied. The features in the data may not all be relevant; hence, a subset of the features is selected using co-relation analysis. Principal component analysis is used for feature extraction.
- Actual analysis: The analysis of the data can be exploratory or confirmatory. In the confirmatory analysis, a hypothesis is proposed, and the data sets are examined to see if the proposed hypothesis is correct or incorrect. Exploratory analysis is close to actual data mining, where the data sets are examined to reveal patterns, find correlations, or generate new information.
- Utilization of harnessed information: The data harnessed can be utilized to detect frauds, spam, detect diseases, identify business trends, marketing, language translation, improve verification systems, geotagging, personalized ads, and product recommendations.