Now a days we are training neural network with large amount of training data (Deep Learning). Can't we train other already established algorithms Deep? If yes than can we use the term deep learning? Is there any literature in that author did the comparison of deep neural network with other algorithms with the same amount of training data??
Please suggest.
Thank you.
I used the hidden Markov model (HMM) to train images. It includes a number of hidden states and the training can be iterated several times. Using the neural network for this purpose consumes more time than HMM.
You could train optimization algorithms with lots of data, it will indeed improve the convergence of the algorithms. But, I agree with Alrik Stegmaier
, it may be not "deep training" as concept of deep learning.For other global optimization algorithms, training will help to find better control parameter values of optimization algorithms. I think you could focus on designing "adaptive" algorithms that could adaptively find the optimal control parameters with lots of training data.
I have one paper to compare GA, RWBS, PSO and DE in the context of telecommunications. Please see it if you think you could gain somthing.
Article Evolutionary-Algorithm-Assisted Joint Channel Estimation and...
You could used the swarm optimization algorithms , which have ability to reach to global optimal model with given training dataset. An example of swarm intelligent are: ACO, PSO, ABC,...etc.
- Badges
- Science topic