3 ms·
This seems to imply the features lernt by neural networks are very different from the features humans use to distinguish the same objects because they are affec
by danbruc 10y ago
This seems to imply the features lernt by neural networks are very different from the features humans use to distinguish the same objects because they are affected by distortions that do almost not interfere with features used by humans at all.
- danieltillett 10y agoOne 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_...
- nabla9 10y agoHuman vision is not a single snapshot. We have two 'cameras' and they scan the image they are looking at by jumping around the image 20–200 ms intervals. The perceived image is integration of many of these jumps and its' constantly changing.
- spacehacker 10y agoI don't think it is necessarily as easy as that. See for example this discussion: https://www.reddit.com/r/MachineLearning/comments/4e2c3q/151106292_foveationbased_mechanisms_alleviate/ https://www.reddit.com/r/MachineLearning/comments/4e2c3q/151...