I just want to ask some details about using population-based optimizations in signal processing. i want to use optimal weights for my signal processing data. i mean i define a set of normalized weights for each of signal's property, they will be used in training with NN. the purpose is to minimize the prediction or classification errors. so, i think the data should be used in each training - test process, every time with new weights and finally the final test should be provided. some questions are as below
1.is it true way?
2.is there any other alternative way?
3. is it useful way to raise the Efficiency?
Yes, you can use any search heuristic algorithm for optimization of signal processing objectives. But remember that these optimization techniques generally take lot of time for better results so it is normally not applicable in real-time or even online optimization objectives.
1. Is it true way?
hmmmmmmmm............. you can you but it is not a perfect way.
2. Is there any other alternative way?
Yes better to use mathematical optimization techniques.
3. Is it useful way to raise the Efficiency?
Yes optimization is always one of the better way to increase the efficiency by using optimal useage of resources etc.
You can start running the training with any weighting factors provided these are within the numerical range that can be sinthesised inside a neuron. As optimised weights derive from the learning process perhaps you should target efficient learning algorithms instead.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Process