Here is an answer from another forum to my question: MT: So now I know what HTM can definitely do - recognize outline drawings of objects as "cup," "dog" etc. And that IS impressive? No other object recognition system has been able to do this before, as he claims?
A: I don't know of any that works too well (and invariantly, including recognizing mirrored images & different orientation etc... This is fairly important for a system functioning in a dynamic world, and it is obviously a tough nut to crack) You can download that picture demonstration program and try it out yourself btw; http://www.numenta.com/about-numenta/technology/pictures-demo.php Try drawing some of the trained images, but have them mirrored for instance. The success rate is not mind blowing, but then the simulated network is (I reckon) very small. (We would need a huge network implemented in hardware before we can expect reasonable results in reasonable time) Note though that the success rate in a real world is not 100% for a human being either. AFAIK nothing more complex has been demonstrated with HTM. But it seems reasonable to assume that the method they are using can be used to recognize correlations between any spatial/temporal patterns (so that includes data from any sensory system... ...and indeed the human cortex is demonstrably able to recognize data from artificial sensory systems) And if you like to investigate more, the algorithm source is available: http://www.numenta.com/for-developers/software.php ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=fabd7936
