Thursday, May 31, 2012

An experiment in open-source textbooks 2

To further put my money where my mouth is, in regards to my support for open source textbooks, I'm following up Monday's post by making the outline of my open source textbook, Intelligent Information Systems, available online. The outline is in PDF format, and is available at the following address:

Readers are encouraged to comment on the outline via the comments section of this blog - I want to hear your opinions!


  1. Looks good Mike!

    Could it be worth considering hybrid systems such as neuro-fuzzy systems and genetic-fuzzy systems?

    Applications of each IIS method are mentioned at the end of each section and there is also an entire section on applications. Would there be any benefit of moving all material on applications to the last chapter in each section?

    Perhaps a section on real-world case studies? Unsure if this is different to the current section on applications.

    A list of top journals, conferences, software and societies/organisations for each method would be nice and can help point readers to research material.

    Type-2 fuzzy logic systems could be worth a mention. Sections on evolutionary computing and neural systems both have multiple classes of these methods but the material on fuzzy in rule-based systems only has one class, type-1.

    Are SVMs classified as ANNs? I appreciate why they're included though. Perhaps a different name for that section would fit better with the inclusion of SVM.

    I wonder whether advantages and disadvantages of methods should be combined in a single part of a chapter. They are currently split. Combining them could present an opportunity for a discussion about weighing up the pros and cons because it is not always a straight forward decision.


    1. Steve, thanks very much for your comments. These are all good ideas - I will definitely include hybrid systems and Type-2 fuzzy logic systems. I also like your ideas about combining advantages / disadvantages into single sections. Lists of top journals / conferences / software could date quickly, but the advantage of an open-source textbook is that can be updated quickly as well.

      The applications sections at the end of each methods chapter are intended to be small case studies on applying the models, whereas the chapters in the applications section of the book are intended to be real-world case studies that can have multiple methods applied to them. I think I will expand this section: I can at least add a section on Eco-informatics / ecological modelling, from my own research experience.

      I'll post an updated outline in the next week.

      Thanks again,

  2. Hello Mike,

    Some statisticians have stated that neural networks have outperformed statistical models because neural models contain a larger number of parameters than most statistical methods. To my mind, neural network modellers have set aside this issue. Hence, I think that it would be useful to highlight the role of parameters in implementing neural models. For example, you could compare the role of ‘neural parameters’ in deploying a MLP with the role of ‘statistical parameters’ in implementing a multiple linear regression.

    The best of luck with your textbook.