Page 1
Page 1
Modern deep learning for tabular data : Novel approaches to common modeling problems
Synthesizes and presents novel deep learning approaches to a seemingly unlikely domain - tabular data. Whether for finance, business, security, medicine, or countless other domain, deep learning can help mine and model complex patterns in tabular data - an incredibly ubiquitous form of structured data. Part I of the book offers a rigorous overview of machine learning principles, algorithms, and implementation skills relevant to holistically modeling and manipulating tabular data. Part II studies five dominant deep learning model designs - Artificial Neural Networks, Convolutional Neural Networks, Recurrent Neural Networks, Attention and Transformers, and Tree-Rooted Networks - through both their 'default' usage and their application to tabular data. Part III compounds the power of the previously covered methods by surveying strategies and techniques to supercharge deep learning systems: autoencoders, deep data generation, meta-optimization, multi-model arrangement, and neural network interpretability.
Information theory and machine learning
The recent successes of machine learning, especially regarding systems based on deep neural networks, have encouraged further research activities and raised a new set of challenges in understanding and designing complex machine learning algorithms. New applications require learning algorithms to be distributed, have transferable learning results, use computation resources efficiently, convergence quickly on online settings, have performance guarantees, satisfy fairness or privacy constraints, incorporate domain knowledge on model structures, etc. A new wave of developments in statistical learning theory and information theory has set out to address these challenges.
Hybrid Intelligent Systems : Analysis and Design
The objective of this edited volume is to offer a general view at the recent conceptual developments of Soft Computing (SC) regarded as a general methodology supporting the design of hybrid systems along with their diversified applications to modeling, simulation and control of non-linear dynamical systems. As of now, SC methodologies embrace neural networks, fuzzy logic, genetic algorithms and chaos theory. Each of these methodologies exhibits well delineated advantages and disadvantages. Interestingly, they have been found useful in solving a broad range of problems. However, many real-world complex problems require a prudent, carefully orchestrated integration of several of these methodologies to fully achieve the required efficiency, accuracy, and interpretability of the solutions. In this edited volume, an overview of SC methodologies, and their applications to modeling, simulation and control, will be given in an introductory paper by the Editors. Then, detailed methods for integrating the different SC methodologies in solving real-world problems will be given in the papers by the other authors in the book. The edited volume will cover a wide spectrum of applications including areas such as: robotic dynamic systems, non-linear plants, manufacturing systems, and time series prediction.
Explainable Artificial Intelligence : An Introduction to Interpretable Machine Learning
Offers a curriculum for introducing interpretability to machine learning at every stage. The authors provide compelling examples that a core teaching practice like leading interpretive discussions can be taught and learned by teachers and sustained effort. And what better way to strengthen the quality of AI and Machine learning outcomes. I hope that this book will become a primer for teachers, data Science educators, and ML developers, and together we practice the art of interpretive machine learning.
Classification and Modeling with Linguistic Information Granules : Advanced Approaches to Linguistic Data Mining
Many approaches have already been proposed for classification and modeling in the literature. These approaches are usually based on mathematical mod els. Computer systems can easily handle mathematical models even when they are complicated and nonlinear (e.g., neural networks). On the other hand, it is not always easy for human users to intuitively understand mathe matical models even when they are simple and linear. This is because human information processing is based mainly on linguistic knowledge while com puter systems are designed to handle symbolic and numerical information. A large part of our daily communication is based on words. We learn from various media such as books, newspapers, magazines, TV, and the Inter net through words. We also communicate with others through words. While words play a central role in human information processing, linguistic models are not often used in the fields of classification and modeling. If there is no goal other than the maximization of accuracy in classification and modeling, mathematical models may always be preferred to linguistic models. On the other hand, linguistic models may be chosen if emphasis is placed on interpretability.
