<< 1 >>
Rating: 
Summary: A Commentary on Kevin Swingler's Applying Neural Networks
Review: Applying Neural Networks not only is a good review of the types of neural networks and and excellent discussion of how to design and implement them. It not only teaches how to select the type of neural net to use. What I loved most about this book was that it discussed with insightful, vivid details how to plan for, conceptualize, and prepare the neural net project long before selecting the actual type of network. It tells how and why to make the inquiries and choices you must make starting very early and at each stage of project development. For example, it discusses how to prepare data, how to choose data types, how to scale it, how to collect it, validation of it, data quality checking, and encoding it. Data quality and preparation are important keys to neural network success, like ingredients-preparation in cooking. Swingler shows why in an easy-to-understand manner. The book also discusses how to select project variables, outlier removal, the tradeoffs involved in network parameter selections, building training and test data, how to analyze outputs and errors, how to set stop-training criteria (and a host of other thresholds), how to visualize training data and error distributions in 2D and 3D, what derivatives are and what they mean, how to do project maintenance, how to adapt the network to external changes, and total project management. Some very good examples of neural network projects illustrate how various researchers implemented these choices. This book will tell you how to make some excellent choices in the design and running of a neural network project, as well as teach you why you are selecting between the alternatives. It is the only true,in-depth neural network methodology book I have found.
Rating:
Summary: A Commentary on Kevin Swingler's Applying Neural Networks
Review: Applying Neural Networks not only is a good review of the types of neural networks and and excellent discussion of how to design and implement them. It not only teaches how to select the type of neural net to use. What I loved most about this book was that it discussed with insightful, vivid details how to plan for, conceptualize, and prepare the neural net project long before selecting the actual type of network. It tells how and why to make the inquiries and choices you must make starting very early and at each stage of project development. For example, it discusses how to prepare data, how to choose data types, how to scale it, how to collect it, validation of it, data quality checking, and encoding it. Data quality and preparation are important keys to neural network success, like ingredients-preparation in cooking. Swingler shows why in an easy-to-understand manner. The book also discusses how to select project variables, outlier removal, the tradeoffs involved in network parameter selections, building training and test data, how to analyze outputs and errors, how to set stop-training criteria (and a host of other thresholds), how to visualize training data and error distributions in 2D and 3D, what derivatives are and what they mean, how to do project maintenance, how to adapt the network to external changes, and total project management. Some very good examples of neural network projects illustrate how various researchers implemented these choices. This book will tell you how to make some excellent choices in the design and running of a neural network project, as well as teach you why you are selecting between the alternatives. It is the only true,in-depth neural network methodology book I have found.
Rating: 
Summary: A Request to the Book-Reviewers
Review: I haven't read this book and I don't know much about NNS, but willing to start. I found the reviewers of this book did an excellent job. Of course, every book has some limitations. Those pointing out the limitations are requested to recommend alternative books, especially for the beginners. Please ignore the star rating by me because I have to fill this to post my view.
Rating: 
Summary: Not the deepest book on the subject
Review: This book is a fairly easy read. About no mathematical blobs thrown around, and still it contains a great deal of information. While you will find truly deep books on neural networks, at least this is a book you will have a reasonable chance from start to finish.. And you will probably end up understanding most of it. Of course , it is an advantage to understand the backpropagation algorithm before buying this book (and also understand the math behind it), but it should contain all needed information. But be prepared to look at the references if you are going to implement a specific "not very standard" algorithm. Most of the papers are on the internet, so it shouldnt be a problem. The book only talks about feedforward and recurrent ANNs, using gradient descent seach ( Like backpropagation). It does not cover any unsupervised learning or GA training algorith. But if your field is supervised learning, this is a helpfull book for you. I havent looked at the software, and probably wont. If you want to truly understand ANN, implement the algorithms yourself.
Rating:
Summary: Not the deepest book on the subject
Review: This book reads like a doctoral thesis. The neural network theory presented is quite complete, if difficult to wade through. Having "practical" in its title, I expected far better examples on the accompanying disk. However, the source code came with no make files and no sample data. Many syntax errors quickly became apparent when I tried to incorporate the code into a project (unmatched parentheses, use of undeclared variables, etc.). Once I fixed those, bugs in the code began to surface, such as closing the output file after calling "return" and other serious bugs. It is clear that the code has never been actually tested. To summarize, if you already know something about neural networks and want to get deeper into the theory and formulas, this may be the book for you. But it certainly will NOT get you started writing an NN application without considerable effort and additional research.
Rating: 
Summary: Much more theoretical than practical
Review: This book reads like a doctoral thesis. The neural network theory presented is quite complete, if difficult to wade through. Having "practical" in its title, I expected far better examples on the accompanying disk. However, the source code came with no make files and no sample data. Many syntax errors quickly became apparent when I tried to incorporate the code into a project (unmatched parentheses, use of undeclared variables, etc.). Once I fixed those, bugs in the code began to surface, such as closing the output file after calling "return" and other serious bugs. It is clear that the code has never been actually tested. To summarize, if you already know something about neural networks and want to get deeper into the theory and formulas, this may be the book for you. But it certainly will NOT get you started writing an NN application without considerable effort and additional research.
Rating:
Summary: Much more theoretical than practical
Review: This book reads like a doctoral thesis. The neural network theory presented is quite complete, if difficult to wade through. Having "practical" in its title, I expected far better examples on the accompanying disk. However, the source code came with no make files and no sample data. Many syntax errors quickly became apparent when I tried to incorporate the code into a project (unmatched parentheses, use of undeclared variables, etc.). Once I fixed those, bugs in the code began to surface, such as closing the output file after calling "return" and other serious bugs. It is clear that the code has never been actually tested. To summarize, if you already know something about neural networks and want to get deeper into the theory and formulas, this may be the book for you. But it certainly will NOT get you started writing an NN application without considerable effort and additional research.
<< 1 >>
|
