Machine learning has become a hot topic in the last few years. However, the identification of a perfect ML algorithm for a particular application is still not easy. Therefore, I have two questions,
1) How to choose the best ML algorithm?
2) Is ML always beneficial?
I think it is hard to define any perfect ML algorithm for all sort of applications. Choosing a ML algorithm for your problem depends on your dataset, training time and also hardware resources you have. But google's EfficientNet(https://ai.googleblog.com/2019/05/efficientnet-improving-accuracy-and.html) recently claimed state of the art result on 5 of the 8 mostly used dataset. Maybe you can give it a try in your problem. Anil Kamboj
It is hard to decide the best ML because it depend on the problem you are trying to solve and the target you want to get such as time consuming in training and testing. I think there is a program may help you to decide the best ML when you insert your input and try it on several ML to see the best results . the program called Weka https://sourceforge.net/projects/weka/
Anil Kamboj
Both these publications can help you understanding the ML rules, best of luck.
Anything that can be represented as a step by step procedure and formulated into an algorithm can be done using ML. Beneficial is a subjective term, iet us say it could be done more effectively (better accuracy and precision) and more efficiently (using less time and computational resources like memory) if the optimal algorithm is chosen.
As mentioned above the choice between ML algorithms for non-computer scientists depends on the task and expected results. Also and most important is the size of your dataset. And the biggest limitation would be the “black box” problem where we cannot judge the obtained results especially less famous algorithm. Second limitation is the optimum structure.
- Badges
- Science topic