3 ms·
One thing is neural networks are much smaller than human brains and most likely have far few overlapping redundant systems. If you had three separate neural net
by danieltillett 10y ago
One thing is neural networks are much smaller than human brains and most likely have far few overlapping redundant systems. If you had three separate neural networks that called on a consensus you might find it much harder to find adversarial inputs.
- pizza 10y agoThis reminds me of signal attenuation/gain/feedback (by neurotransmitter release [I want to say dopamine...]) due to error in the visual cortex.. Hopefully someone who's studied that might have something to share.
- emcq 10y agoI believe the parent was referring to having an ensemble of models with different trained networks can reduce variance and perhaps avoid issues like these. Attention, localized gain, etc would not have this effect, but they tend to allow a smaller network to perform more sophisticated tasks.
- pizza 10y agoLike so? https://en.wikipedia.org/wiki/Stochastic_resonance_(sensory_neurobiology)#Multi-unit_systems_of_model_neurons https://en.wikipedia.org/wiki/Stochastic_resonance_(sensory_...