3 ms·
Since adding noise is a critical part in many neural network training models, it is entirely possible that an at-hand source of randomness is highly useful. Dr
by exoesquitur 8y ago
Since adding noise is a critical part in many neural network training models, it is entirely possible that an at-hand source of randomness is highly useful.
Drugs that reduce the accessibility of this randomness by biasing the signal pathways (or something) might be an interesting research vector.
- maze-le 8y ago>> Since adding noise is a critical part in many neural network training models Our brains do not work exactly like neural network models. Neural Networks are a model, on how nervous systems could work. It doesn't follow, that nature has implemented them the same way. There is no need for the human brain to produce noise with some arbitrary process. There is more than enough noise in the human brain... Every single neuron is a powerful noise-filter[0], that seemingly can generate signals when the "correct" noise-signal is filtered. >> Drugs that reduce the accessibility of this randomness by biasing the signal pathways The randomness you are talking about is way the brain operates under normal circumstances[1]... so no, it wouldn't be beneficial, since we cannot discern signal from noise in any meaningful way [2]. [0]: https://en.wikipedia.org/wiki/Neuronal_noise https://en.wikipedia.org/wiki/Neuronal_noise [1]: https://ieeexplore.ieee.org/document/486289 https://ieeexplore.ieee.org/document/486289 [2]: not even on the level of the single neuron -- except for very few types of neurons [pyramid neurons] in a rat model -- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1456039/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1456039/
- rusk 8y agoIANAP but my understanding of a lot of psychiatric drugs is that they do just this, they dampen brain activity which may be analogous to what you're describing.