Artificial Intelligence is all the rage and thanks to the third party component system in Delphi you can dive right in with a suite of components called Intelligence Lab from Mitov Software. You can do all kinds of fun things with neural networks like generate unique baby names, feed them romance novels, colorize black and white photos, and build chat bots. These components have been upgraded to support Delphi Berlin plus they work across Android and Windows with some form of support for IOS and OSX. The component suite consists of components with everything from classifiers, to converters, to filters, to generators, to timing components, to training components. Examples of the components are two neural network components with Backprop and RProp algorithms, a naive Bayesian classifier, and a Radial Basis Function Network component. These components could be used to build OCR applications, computer vision functionality, SPAM filters, AI, speech recognition, and decision making functionality. These components are free to download and use for non commercial use but for commercial uses there is a license fee. The included demos are a K-Nearest Neighbors (KNN) demo which demonstrates a machine learning algorithm for classifying objects. A Neural Network demo which demonstrates a neural network classifier. A Self-Organizing Map demo which demonstrates a type of neural network that uses unsupervised learning. A naive Bayes demo which demonstrates a Naive Bayes classifier. And a Face Recognition Demo which demonstrates training a Radial Basis Function Network to recognize faces, and then performing the actual recognition. These components should work with C++Builder as well.
Who’s going to pay almost $400 (over $1300 with source) for implementations of a few trivial machine learning algorithms in 2016? There are plenty of open source alternatives for all of these basic algorithms, for instance FANN for neural networks, which has a Delphi interface:
http://leenissen.dk/fann/wp/
I wish the Delphi community would stop trying to make money off of Delphi itself; it makes the cost of entry for Delphi enormous, especially in an era when almost all languages are open source.
Does libFANN run on Android/IOS/OSX? It’s C so I’m sure it could be compiled to it but I don’t see an existing implementation. Could easily cost $400+ to get libFANN working on all the platforms and get the headers in place.
No real problem compiling it for Android:
http://andre-pomp.blogspot.com/2014/10/tutorial-compiling-fann-for-android.html
Or iOS or OS X:
http://stackoverflow.com/questions/18704779/how-to-install-and-use-fann-with-xcode
Yes, it sounds simple. However, getting a C library compiled and setup for all platforms (Android, IOS32+IOS64, OSX, Win32, and Win64) and tied into working cross platform Object Pascal headers can be very time consuming. I know. I did the libSOX library for Android, IOS32+IOS64, OSX, Win32, and Win64. Maybe libSOX was more complicated because it had to access hardware.
http://www.fmxexpress.com/convert-sound-files-and-apply-audio-effects-in-delphi-xe8-firemonkey-on-android-and-ios/
You’re welcome to do the same for libFANN and I will happily write a post about it as well 🙂