Pipeline Automated Search always giving Autoencoder as detection algorithm

[sagarshankar]

trafficstars

I was running an automated search pipeline on multiple time series datasets and I am always seeing that the best pipeline consists of the Autoencoder detection algorithm? All the examples from the Jupyter notebooks have Autoencoder in their pipelines. I check the brute_force_search.py and uncommented some of the other detection algorithms that exist but I still keep seeing Autoencoder only in the best pipeline results. Why is this the case regardless of the dataset?

[sagarshankar]

How can I disable certain detection algorithm primitives like Autoencoder since that is the only pipeline result I get?

[lhenry15]

At this moment, you may want to look into the brute_force_search.py to configure the search space. We are now working on a JSON file-based interface for user to manually create search space and will be released in the following two months. Thanks.

[sagarshankar]

I tried commenting some of the detection algorithms like Autoencoder (pyod.ae) and others in the brute_force_search.py. But it still provided the autoencoder as the best pipeline regardless of the dataset used?