However, finding the right deep learning architecture for your program can be challenging.
Each architecture has its strengths and weaknesses.
It’s free, every week, in your inbox.
The first part of a NAS strategy is to define the search space for the target neural internet.
The basic component of any deep learning model is the neural layer.
it’s possible for you to determine the number and jot down of layers to explore.
The more elements you add to the search space, the more versatility you get.
More advanced architectures usually have several branches of layers and other elements.
These types of architectures are more difficult to explore with NAS because they have more moving parts.
In this case, the NAS algorithm can optimize small blocks separately and then use them in combination.
Therefore, a neural architecture search algorithm also needs a search strategy.
The search strategy determines how the NAS algorithm experiments with different neural networks.
There are other techniques that speed up the search process.
Another strategy is to frame neural architecture search as areinforcement learning problem.
Other search strategies includeevolutionary algorithmsandMonte Carlo tree search.
Obviously, doing full training on each neural connection takes a long time and requires very large computational resources.
Transfer learning is applicable when the source and destination model have compatible architectures.
you’ve got the option to read the original articlehere.
