1
$\begingroup$

What are linear and non-linear machine learning algorithms? How to compare and select the right one for a given use case of classification problems? Which are the key performance parameters to select a model?

$\endgroup$

closed as too broad by SmallChess, TwinPenguins, Stephen Rauch, Aditya, Toros91 Apr 17 '18 at 6:03

Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. Avoid asking multiple distinct questions at once. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 4
    $\begingroup$ Welcome to DS.SE! Your question seems quite broad, could you sharpen it to contain only one well defined one? $\endgroup$ – mico Apr 15 '18 at 13:16
0
$\begingroup$

Actually, there are many linear and non-linear machine learning algorithms. Selecting a right algorithm highly depends on your data-set and the nature of your data. for instance, in machine-learning era, it was customary to estimate functions by assigning a typical model to the problem and reducing the error by predicting the appropriate coefficients using a cost function for regression tasks. In such cases, you should have bias-variance trade-off. That means you should not fit the data nor miss it. you should find a good estimate, a good model that has generalization capability. For doing all of these you have to choose features which help you describe the problem better for making a model. In deep-learning, we usually do not have this trade-off. If you increase the size of your training set, you can almost be sure that you can have better results. In machine-learning, you can always be sure that by making complex non-linear models, you overfit your data while using complex deep-learning models does not necessarily mean that if you employ generalization techniques which avoid overfitting. For choosing a right model, it is customary to use a simple linear model and make it complicated step by step. Problems which their inputs have numerous features, you can not see and visualize the data-set to check whether they are linearly separable in that space or not. Consequently, beginning with a simple model and making it complex step by step is a logical solution. For measuring performance there are different solutions that all depend on your goal. Take a look at here.

$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.