Machine Learning Engine for on-prem v20.0 release notes
Machine Learning Engine for on-prem v20.0 includes minor bug fixes and improvements.
Compatibility
Machine Learning Engine v20.0 is compatible with the Rome release.
Installation
You can access this update along with the download and installation instructions via the Machine Learning Engine Installation and Configuration for Self-Hosted Customers Knowledge Base (KB0782052).
For technical assistance and information, please visit ServiceNow Support: https://support.servicenow.com/support/
Release notes
| Problem category | Number | Short description |
|---|---|---|
| NLU Service | PRB1495427 | The Intent Discovery report analysis does not complete and it is stuck on the state of "Classifying" |
| NLU Service | PRB1479517 | Inconsistent inference results with the MxN vocabulary |
| NLU Service | PRB1498024 | NLU scores lower for intents containing the exact match to the utterance |
| NLU Service | PRB1496043 | Prediction results are not populated if the sources level is confirmed as software/hardware |
| NLU Service | PRB1491436 | The exact same utterance returns different results on the same NLU Model published |
| NLU Service | PRB1499630 | Exact match fails for utterance MCI Fault on Lloyds dataset |
| Predictive Intelligence - Server | PRB1488756 | Clustering updates on the trainer causes OOM errors and server instability issues |
| Predictive Intelligence - Server | PRB1458110 | Regression Framework Solution gives an error "NaN for the Symmetric Mean Absolute Percentage Error (SMAPE)" |
| Predictive Intelligence - Server | PRB1487079 | Classification training fails with an error "Data has less than 2 output classes" even though there are over 20 classes |
| Predictive Intelligence - Server | PRB1482553 | Agent Zero batch prediction API returns results only for the first input |