Model especially sensitive to number of max_iter

[YingxiaoKong]

Hello Raghavendra,

Thank you for sharing your code! The normal data are generated from a gaussian distribution with mean as 0 and std as 1 and the sample size is 1000. The anomalous data are generated from a gaussian distribution with mean as 5 and std as 1 and the sample is 20. I tried it with different numbers of max_iter and the results are shown below:

The gradient errors for all cases are all very small (around 1.9e-8), but from the figure you can find totally different patterns of the anomaly scores. For me I feel that num_iter 10 and num_iter 50 provide quite convincing results. But this is a very simple dataset and I can visualize them. Given a multivariate dataset when the data is not easy to visualize and thus evaluate the model performance, is there any other way to do it?