Does anyone know any significant shortcomings of neural networks? I know they are prone to:
>overfitting
>biased results if training data is biased
>huge amount of data required in order to do anything significant
>long training times
>requires powerful hardware for training in reasonable amount of time
>hardly any analytical approaches work for optimization of weights - all initial values have to be guessed, learning the right values takes time
what else?
>overfitting
>biased results if training data is biased
>huge amount of data required in order to do anything significant
>long training times
>requires powerful hardware for training in reasonable amount of time
>hardly any analytical approaches work for optimization of weights - all initial values have to be guessed, learning the right values takes time
what else?
