I could not quite get the reason behind this condition of dropping of anomalies. The comment says dropping of initial anomalies but if my batch size is say 64 and error window is 64x30, this line of code limits identified anoms only in the last part of the window, [1856:1920] and ignores the rest.
i_anom = i_anom[i_anom >= len(e_s) - self.config.batch_size]
https://github.com/khundman/telemanom/blob/26831a05d47857e194a7725fd982d5dea5402dd4/telemanom/errors.py#L359C13-L360C1
I could not quite get the reason behind this condition of dropping of anomalies. The comment says dropping of initial anomalies but if my batch size is say 64 and error window is 64x30, this line of code limits identified anoms only in the last part of the window, [1856:1920] and ignores the rest.
i_anom = i_anom[i_anom >= len(e_s) - self.config.batch_size]https://github.com/khundman/telemanom/blob/26831a05d47857e194a7725fd982d5dea5402dd4/telemanom/errors.py#L359C13-L360C1