Andrew Maguire

Step 1 - "By Type" has been done here: https://github.com/netdata/netdata/pull/15856

@ktsaou @vkalintiris @MrZammler @thiagoftsm based on [this comment](https://github.com/netdata/netdata-cloud/issues/793#issuecomment-1521965435) this issue maps out a potetnail approach to adding a new similarity method to weights endpoint on agent.

@vkalintiris fyi

@vkalintiris or @ktsaou I'd like to see if we could maybe add this new method to the weights endpoint in next few weeks or so. This would put in place...

## Some initial discussion `ml-demo-ml-enabled` (green line) vs `ml-demo-ml-enabled-newconf` (red line) under the new defaults the anomaly rate itself should tend to be lower. here is a typical profile of...

## Q: if we introduce a true anomaly to both systems with the anomaly rate on the node with the new defaults actually react enough to truly unseen patterns? ###...

## reduction in false positives under the new defaults we should expect to see less temporary bumps in the anomaly rate when nothing that unusual actually happened. for example in...

closing in favor of this (https://github.com/netdata/netdata/pull/15093) since a few things have changed since i opened this original draft PR. All findings and discussion above still stand and are relevant.

@vkalintiris @stelfrag sorry - i messed up other version of this PR - this is a clean one.

> The problem with new UI is that is has too many tunables and any change that you make in any chart (grouping, aggregation, chart type etc) is not preserved...