You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
It would be neat to keep a list around a list of "simple problems where neural nets can't learn a generalizable solution, even with lots of training data". There are many of them, and they could serve as a benchmark for future progress.https://twitter.com/GaryMarcus/status/957685727969660929 …
The gist of it is this: neural nets do *pattern recognition*, which achieves *local generalization* (which works great for supervised perception). But many simple problems require some (small) amount of abstract modeling, which modern neural nets can't learn
Can Capsule Networks form part of a solution to address the need for abstract modeling?
As far as I can tell, no. Capsules address limitations of CNNs but do not address the overarching limitations of deep learning.
Here's one: train a dense net, an RNN, and an OptNet to evaluate y=x^2 far from the training data (scalar and digit representations). Then try y=sqrt(x). Wildly different generalization regions.
Applies to most humans too.
The biggest problem with deep neural nets is that they cannot see things that they have not been trained on, a serious flaw. The brain, by contrast, can instantly see complex objects it has never seen before. It does not even learn complex patterns. It learns to see the world.
Actually it seems like the brain classifies new objects or scenes into kind of a friend or degrees of foeness default bucket.
If you frame the abstract modeling problem as a classification problem and collect some data, then why can't modern neural nets learn.
Pattern recognition IS some kind of generalization, although it is stuck within the spatial domain for the nets. They need context as a meta-layer and do the "pattern-recognition" then. Perhaps, multimodal input could help, just like in humans...
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
