calclavia / htm-tensorflow Goto Github PK

you seems that implement 'The HTM SPatial Pooler- a neocorical algorithm for noline sparse distributed coding' but i can't find Neighbor(i)={j | ||y(i)-y(j)|| < theta, j=|i} ----(4) and (6) (7)
I cant' find part (4), (6), (7)

Was this work derived from the HTM paper? Would be great to have this work under Apache 2.0 or MIT! See discussion under https://discourse.numenta.org/t/a-word-on-licensing/2023

Hi,

I took a look of the mnist.py and found that was a type of offline supervised learning by using spacial pooling, so I would like to know if there are some examples/ways to support unsupervised online learning as NuPIC did by using your tensorflow implementation?

Regards,
Ming

@calclavia

I run mnist.py, error happens as follows
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcublas.so.8.0 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcudnn.so.5 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcufft.so.8.0 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:135] successfully opened CUDA library libcurand.so.8.0 locally
Using TensorFlow backend.
Traceback (most recent call last):
File "mnist.py", line 111, in
main()
File "mnist.py", line 43, in main
model = HTMModel()
File "mnist.py", line 29, in init
pooler = SpatialPooler(htm_units, lr=1e-2)
File "/home/wanghz/htm-tensorflow/layers/spatial_pooler.py", line 26, in init
super().init(**kwargs)
TypeError: super() takes at least 1 argument (0 given)

Could you please tell me how to fix it? Thanks

I'm new to computational neuroscience research and am having tough time understanding which Numenta's research papers you used. Can you please give their reference in Readme? That would be really helpful.

Thanks for implementation btw!

@calclavia I read spatial_pooler.py, and cannot understand line 95, delta = tf.einsum('ij,ik,jk->jk', x_shifted, y, self.connection) / batch_size. what is the mechanism of computing delta p based on x, y and self.connection? Could you give me some clues or hints? Thanks