dlagul / sdfvae Goto Github PK

Dear authors, how can we use this algorithm on data with other #KPIs (also named as n in the code)？Now this code will raise ValueError('Invalid Kpi numbers: {}, choose from the candidate set [16,24,38,48].'.format(self.n)) when it is tested on other datasets with different #KPI (such as 49, 19 ...). Is the algorithm only applicable on the dataset used in the paper?

Hi, I have great interest in your paper and I got a question when I running your code. What's the meaning of the code on lines of 65~66 in the evaluation.py.

**evaluation.py**
65  if current_timestamp >= start_dt and current_timestamp <= end_dt:
66      detected_anomaly_intervals_tmp[k] = ground_truth_anomaly_intervals[j]

Does this mean that the whole anomaly intervals are detected when you only find an anomaly point in an interval?

请问您在SMD这个数据集上进行的实验训练集是否为OmniAnomaly开放的SMD训练集呢？
请问您是否可以告知在该实验上您所设置的相关参数呢？我采用了您的默认参数，数据预处理后数据增强了很多倍，以我的实验设备（满足您论文中提出的实验设备能力）报错内存不够，所以想询问一下您所使用的实验参数。
感谢！