Let me be the first to say thank you for releasing your source and starting this project!
May your gift gain wings and fly far on the spirit of generosity.
May it uplift us who have sought to understand ourselves.
May it prove truly useful and ultimately benefit those who have less.
-Dave Dumaresq
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
You are very welcome :)
By the way, if you or anyone else would like to join the further development of this project, please let me know. There are a lot of things to do in improving and researching this technology further!
Saulius
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Great job you've done! I'm looking at the code now. I would like to help with the project, but will need to get myself up to speed with C++. I did some 'C' coding back in the 1980's. Even some neural network stuff. Also, I'll need to get up to speed with Bayesian networks, but I have a degree in math, so that shouldn't take long. I'm on my 3rd reading of 'On Intelligence' now. Tremendous work! Most exciting ideas that I've seen since I discovered neural nets.
Which version of BCB did you use? I've tried to build an executable but keep getting a message that says it can't load Chartfx3.hpp. I've downloaded Borland's 30 day trial Builder evaluation package. I'm an educator so I can get the student/educator version for about $100 through our college bookstore. Are you using the educator version?
Thanks,
Gary
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
that would be great if you are going to join the project! I do share the excitement about the theory in "On Intelligence". I can help with C++ if any section of the code is not clear.
The code does compile on Borland C++ Builder 6.0 Enterprise, and I am pretty sure it would compile on C++ Builder 5.0 Professional. In fact, no "extra fancy" features of C++ Builder are used in this project, so it would be safe to say that it should compile on any version of C++ Builder with minimal modifications, if any.
ChartFX is not an essential component of the program, it is only used for graphical representation of the results (bar chart). You probably downloaded the Builder 2006 and I have not tried to use it for this project. However, I found this solution to your specific situation: http://support.borland.com/entry.jspa?externalID=3890&categoryID=135
See if it helps.
Saulius
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I'd also like to thank you for releasing this software. However, it will be easier to attract other developers if there were a library separate from the Windows-only GUI. As it is, it's very hard to understand just what exactly it does, and I can't easily access it from a dynamic language like Python or Ruby to play around with it.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Chris,
I see your point and this is a good task. Two possibilities:
1. Separate GUI part of the code from the algorithmic core ("engine"), so that people could easily write their own GUI for their OS and compiler environments. This has largely been done already. If you look at the code, GUI-related functionality is concentrated in just two files: MainFrm.cpp and InOut.cpp. The rest of the code uses calls to InOut.cpp to output the results. So one could adapt the code to any platform just by rewriting the code for GUI output in InOut.cpp and the GUI event handlers in MainFrm.cpp, with minimal changes to all other files.
2. The idea of making a library accessible from other languages. It is probably too early to implement this. The whole point of the project right now is to improve the algorithms themselves, and this requires constant modification of all project "engine" files. Although a great deal of basic experimentation could be done just by playing with the constants in file Constants.h, without changing anything else.
Saulius
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Yes, I am using Builder 2006. Your suggestion worked. Everything seems to compile, now.
After discovering that the package couldn't handle the space in the pathname in the project file called "Memory-prediction framework", and having to add 'cdiroutl.cpp' to the project because the linker couldn't open 'cdiroutl.obj', I've run up against the following now: "[Linker Fatal Error] Fatal: Unable to open file 'EXTCTRLS.OBJ'" This obj file doesn't exist on my computer. I see a EXTCTRLS.HPP which I've tried 'adding to the project' but that doesn't seeem to help.
I would download the Enterprise version 6, but the Borland website is either broken or they won't download trial validation keys for version 6, anymore.
I'll keep fiddling with version 2006 for a while. Our bookstore is out of the student editions anyway.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Whenever i try to open any one of the demo apps, i get a error stating that "outline index
not found". How do i get the apps to open without any problems being created ?
In addition to the post "Numenta has opened the floodgates", i guess that its true, but
thats going to bring a lot of people here to this project too. A lot of interesting things
are happening at once in this sphere of research after a long gap. :)
- Girish
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Girish,
I tried and could not reproduce the error here. Looks like other people are running the programs fine. Please make sure you extract the programs to a directory with Test images and Training images as subdirectories. Also, the file "UniqueL1VectorsGroups.txt" must be in the same folder as the programs. What Windows version are you using?
You also may want to try compiling the source using Turbo C++ and see if that produces the same error.
Yes, these kinds of projects bring popularity to the research field as a whole, so hopefully this open-source version will also get more exposure.
Saulius
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Yes, Neocortex and NuPic are quite similar. They use the same principles of Bayesian belief propagation network. Mechanisms for remembering and recognizing spatial patterns are not exactly the same, but yield similar results. Neocortex does not yet include temporal grouping, unlike NuPic. The recognition performance results over the testing image set are quite comparable in NuPic and Neocortex, according to the testing that I have done.
Saulius
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Let me be the first to say thank you for releasing your source and starting this project!
May your gift gain wings and fly far on the spirit of generosity.
May it uplift us who have sought to understand ourselves.
May it prove truly useful and ultimately benefit those who have less.
-Dave Dumaresq
You are very welcome :)
By the way, if you or anyone else would like to join the further development of this project, please let me know. There are a lot of things to do in improving and researching this technology further!
Saulius
Saulius,
Great job you've done! I'm looking at the code now. I would like to help with the project, but will need to get myself up to speed with C++. I did some 'C' coding back in the 1980's. Even some neural network stuff. Also, I'll need to get up to speed with Bayesian networks, but I have a degree in math, so that shouldn't take long. I'm on my 3rd reading of 'On Intelligence' now. Tremendous work! Most exciting ideas that I've seen since I discovered neural nets.
Which version of BCB did you use? I've tried to build an executable but keep getting a message that says it can't load Chartfx3.hpp. I've downloaded Borland's 30 day trial Builder evaluation package. I'm an educator so I can get the student/educator version for about $100 through our college bookstore. Are you using the educator version?
Thanks,
Gary
Gary,
that would be great if you are going to join the project! I do share the excitement about the theory in "On Intelligence". I can help with C++ if any section of the code is not clear.
The code does compile on Borland C++ Builder 6.0 Enterprise, and I am pretty sure it would compile on C++ Builder 5.0 Professional. In fact, no "extra fancy" features of C++ Builder are used in this project, so it would be safe to say that it should compile on any version of C++ Builder with minimal modifications, if any.
ChartFX is not an essential component of the program, it is only used for graphical representation of the results (bar chart). You probably downloaded the Builder 2006 and I have not tried to use it for this project. However, I found this solution to your specific situation:
http://support.borland.com/entry.jspa?externalID=3890&categoryID=135
See if it helps.
Saulius
I'd also like to thank you for releasing this software. However, it will be easier to attract other developers if there were a library separate from the Windows-only GUI. As it is, it's very hard to understand just what exactly it does, and I can't easily access it from a dynamic language like Python or Ruby to play around with it.
Chris,
I see your point and this is a good task. Two possibilities:
1. Separate GUI part of the code from the algorithmic core ("engine"), so that people could easily write their own GUI for their OS and compiler environments. This has largely been done already. If you look at the code, GUI-related functionality is concentrated in just two files: MainFrm.cpp and InOut.cpp. The rest of the code uses calls to InOut.cpp to output the results. So one could adapt the code to any platform just by rewriting the code for GUI output in InOut.cpp and the GUI event handlers in MainFrm.cpp, with minimal changes to all other files.
2. The idea of making a library accessible from other languages. It is probably too early to implement this. The whole point of the project right now is to improve the algorithms themselves, and this requires constant modification of all project "engine" files. Although a great deal of basic experimentation could be done just by playing with the constants in file Constants.h, without changing anything else.
Saulius
Yes, I am using Builder 2006. Your suggestion worked. Everything seems to compile, now.
After discovering that the package couldn't handle the space in the pathname in the project file called "Memory-prediction framework", and having to add 'cdiroutl.cpp' to the project because the linker couldn't open 'cdiroutl.obj', I've run up against the following now: "[Linker Fatal Error] Fatal: Unable to open file 'EXTCTRLS.OBJ'" This obj file doesn't exist on my computer. I see a EXTCTRLS.HPP which I've tried 'adding to the project' but that doesn't seeem to help.
I would download the Enterprise version 6, but the Borland website is either broken or they won't download trial validation keys for version 6, anymore.
I'll keep fiddling with version 2006 for a while. Our bookstore is out of the student editions anyway.
See the continuation of compiling discussion in the Help Forum, "Still no joy" thread.
Hi Saulius,
Whenever i try to open any one of the demo apps, i get a error stating that "outline index
not found". How do i get the apps to open without any problems being created ?
In addition to the post "Numenta has opened the floodgates", i guess that its true, but
thats going to bring a lot of people here to this project too. A lot of interesting things
are happening at once in this sphere of research after a long gap. :)
- Girish
Hi Girish,
I tried and could not reproduce the error here. Looks like other people are running the programs fine. Please make sure you extract the programs to a directory with Test images and Training images as subdirectories. Also, the file "UniqueL1VectorsGroups.txt" must be in the same folder as the programs. What Windows version are you using?
You also may want to try compiling the source using Turbo C++ and see if that produces the same error.
Yes, these kinds of projects bring popularity to the research field as a whole, so hopefully this open-source version will also get more exposure.
Saulius
Hello.
My name is Chris, AI skills and interested in HTM theory. I'm testing NuPIC technology.
Before testing purposes, let me know if this neocortex framework is similar to NuPIC technology?
Thank you.
Chris.
Hello Chris,
Yes, Neocortex and NuPic are quite similar. They use the same principles of Bayesian belief propagation network. Mechanisms for remembering and recognizing spatial patterns are not exactly the same, but yield similar results. Neocortex does not yet include temporal grouping, unlike NuPic. The recognition performance results over the testing image set are quite comparable in NuPic and Neocortex, according to the testing that I have done.
Saulius