Subutai Ahmad

We should clean up the unused variable. We don't want to randomly break ties - with learning turned off, a given input should lead always to the same result. numpy.argsort...

> Using a random tie breaker gets rid of this bias, and can be deterministic with a random seed. Yes, that's how it used to be before. But this has...

> In a 2000 column region, you'll see (slightly) more activity amongst columns 1900-2000 than for 1-100. With a random tie breaker, the "preferred" columns will be dispersed amongst the...

> how about using connection length as a tie breaker? @vova616 If I understand you correctly, this is already the case. The number of active connections into each column is...

I don't know that there is a single set of parameters that are optimal for everything. For global inhibition and typical streaming analytics scenarios, the parameters here are an excellent...

@arhik I think your confusion is exactly the point of this issue :-) At the end this should be a relatively simple PR. It is only to rename the classes...

@breznak Very interesting proposal! I think this should be very possible with some work. There are some classes that call some C++ routines such as random number generation, and faster...

@david-ragazzi Even if we have a pure Python version of NuPIC, we still need to have a python + C++ version - the latter will be _much_ faster than a...

> Yeah, but you mean only the Python binding to NuPIC C++, ok? Not a hybrid implementation.. Yeah, basically what @breznak said. We would have two build options: one that...

OPF is setup in an OO way to support different model features. Is it possible to implement this feature as a subclass of CLAModel or perhaps as a separate independent...