Cell death


Hi everyone,

Can you help me about cell death?
I dont understand much about what is cell death in brain, and how spatial pooler can implement a cell death. as i understand, if all connections in receptive field of a column are connected, it means that this column is death.
please correct me if i am wrong.

Thank you


Hi Rodi,

I’m not a biologist, but are you talking about the death of neurons due to trauma/Parkinson’s/Alzheimer’s, that sort of thing?

You’re describing some sort of death by overconnection, would you have a link to the biological detail of this?


Thank you.
actually, I read paper about HTM. and I see some simulation about cell death.
i dont understand this concept, how to implement it in HTM.
Can you tell me more about it?


Cells die all of the time in The human brain. Some times its random, other times the result of a blow the the head. And people dont even notice because the brain keeps working.

The spatial pooler paper simulates some cells randomly dying to show that the SP keeps working. All good models of the brain should be able to survive a few cells dying, since real brains can too.

I dont know the details of how they modeled cell death, other than: it was random and the cells died after the network was trained.


There is some discussion of the general SP operation and the part of the paper regarding failures on this thread:


No, I don’t think that this is true. It does mean that the cell will respond to any pattern that turns on the bits in it’s receptive field. This would be a highly unusual learned pattern and the “sparse activation” part of HTM theory should prevent this from happening.


One of the spatial pooler papers showed tolerance to simulated cell death. I this is what @Rodi is referring to. It is not a part of the Spatial Pooling process, it is just something that SP handles well.