Amazon Lookout for Equipment Update: This release adds a field exposing model quality to read APIs for models. It also adds a model quality field to the API response when creating an inference scheduler.

Author: AWS

.changes/next-release/feature-AmazonLookoutforEquipment-9831e1a.json

@@ -0,0 +1,6 @@
+{
+    "type": "feature",
+    "category": "Amazon Lookout for Equipment",
+    "contributor": "",
+    "description": "This release adds a field exposing model quality to read APIs for models. It also adds a model quality field to the API response when creating an inference scheduler."
+}

services/lookoutequipment/src/main/resources/codegen-resources/service-2.json

@@ -1076,6 +1076,10 @@
         "Status":{
           "shape":"InferenceSchedulerStatus",
           "documentation":"<p>Indicates the status of the <code>CreateInferenceScheduler</code> operation. </p>"
+        },
+        "ModelQuality":{
+          "shape":"ModelQuality",
+          "documentation":"<p>Provides a quality assessment for a model that uses labels. If Lookout for Equipment determines that the model quality is poor based on training metrics, the value is <code>POOR_QUALITY_DETECTED</code>. Otherwise, the value is <code>QUALITY_THRESHOLD_MET</code>. </p> <p>If the model is unlabeled, the model quality can't be assessed and the value of <code>ModelQuality</code> is <code>CANNOT_DETERMINE_QUALITY</code>. In this situation, you can get a model quality assessment by adding labels to the input dataset and retraining the model.</p> <p>For information about using labels with your models, see <a href=\"https://docs.aws.amazon.com/lookout-for-equipment/latest/ug/understanding-labeling.html\">Understanding labeling</a>.</p> <p>For information about improving the quality of a model, see <a href=\"https://docs.aws.amazon.com/lookout-for-equipment/latest/ug/best-practices.html\">Best practices with Amazon Lookout for Equipment</a>.</p>"
         }
       }
     },
@@ -2029,6 +2033,10 @@
         "ModelDiagnosticsOutputConfiguration":{
           "shape":"ModelDiagnosticsOutputConfiguration",
           "documentation":"<p>Configuration information for the model's pointwise model diagnostics.</p>"
+        },
+        "ModelQuality":{
+          "shape":"ModelQuality",
+          "documentation":"<p>Provides a quality assessment for a model that uses labels. If Lookout for Equipment determines that the model quality is poor based on training metrics, the value is <code>POOR_QUALITY_DETECTED</code>. Otherwise, the value is <code>QUALITY_THRESHOLD_MET</code>.</p> <p>If the model is unlabeled, the model quality can't be assessed and the value of <code>ModelQuality</code> is <code>CANNOT_DETERMINE_QUALITY</code>. In this situation, you can get a model quality assessment by adding labels to the input dataset and retraining the model.</p> <p>For information about using labels with your models, see <a href=\"https://docs.aws.amazon.com/lookout-for-equipment/latest/ug/understanding-labeling.html\">Understanding labeling</a>.</p> <p>For information about improving the quality of a model, see <a href=\"https://docs.aws.amazon.com/lookout-for-equipment/latest/ug/best-practices.html\">Best practices with Amazon Lookout for Equipment</a>.</p>"
         }
       }
     },
@@ -2181,6 +2189,10 @@
         "ModelDiagnosticsResultsObject":{
           "shape":"S3Object",
           "documentation":"<p>The Amazon S3 output prefix for where Lookout for Equipment saves the pointwise model diagnostics for the model version.</p>"
+        },
+        "ModelQuality":{
+          "shape":"ModelQuality",
+          "documentation":"<p>Provides a quality assessment for a model that uses labels. If Lookout for Equipment determines that the model quality is poor based on training metrics, the value is <code>POOR_QUALITY_DETECTED</code>. Otherwise, the value is <code>QUALITY_THRESHOLD_MET</code>.</p> <p>If the model is unlabeled, the model quality can't be assessed and the value of <code>ModelQuality</code> is <code>CANNOT_DETERMINE_QUALITY</code>. In this situation, you can get a model quality assessment by adding labels to the input dataset and retraining the model.</p> <p>For information about using labels with your models, see <a href=\"https://docs.aws.amazon.com/lookout-for-equipment/latest/ug/understanding-labeling.html\">Understanding labeling</a>.</p> <p>For information about improving the quality of a model, see <a href=\"https://docs.aws.amazon.com/lookout-for-equipment/latest/ug/best-practices.html\">Best practices with Amazon Lookout for Equipment</a>.</p>"
         }
       }
     },
@@ -3522,6 +3534,14 @@
         "MANUAL"
       ]
     },
+    "ModelQuality":{
+      "type":"string",
+      "enum":[
+        "QUALITY_THRESHOLD_MET",
+        "CANNOT_DETERMINE_QUALITY",
+        "POOR_QUALITY_DETECTED"
+      ]
+    },
     "ModelStatus":{
       "type":"string",
       "enum":[
@@ -3590,7 +3610,11 @@
           "shape":"RetrainingSchedulerStatus",
           "documentation":"<p>Indicates the status of the retraining scheduler. </p>"
         },
-        "ModelDiagnosticsOutputConfiguration":{"shape":"ModelDiagnosticsOutputConfiguration"}
+        "ModelDiagnosticsOutputConfiguration":{"shape":"ModelDiagnosticsOutputConfiguration"},
+        "ModelQuality":{
+          "shape":"ModelQuality",
+          "documentation":"<p>Provides a quality assessment for a model that uses labels. If Lookout for Equipment determines that the model quality is poor based on training metrics, the value is <code>POOR_QUALITY_DETECTED</code>. Otherwise, the value is <code>QUALITY_THRESHOLD_MET</code>.</p> <p>If the model is unlabeled, the model quality can't be assessed and the value of <code>ModelQuality</code> is <code>CANNOT_DETERMINE_QUALITY</code>. In this situation, you can get a model quality assessment by adding labels to the input dataset and retraining the model.</p> <p>For information about using labels with your models, see <a href=\"https://docs.aws.amazon.com/lookout-for-equipment/latest/ug/understanding-labeling.html\">Understanding labeling</a>.</p> <p>For information about improving the quality of a model, see <a href=\"https://docs.aws.amazon.com/lookout-for-equipment/latest/ug/best-practices.html\">Best practices with Amazon Lookout for Equipment</a>.</p>"
+        }
       },
       "documentation":"<p>Provides information about the specified machine learning model, including dataset and model names and ARNs, as well as status. </p>"
     },
@@ -3656,6 +3680,10 @@
         "SourceType":{
           "shape":"ModelVersionSourceType",
           "documentation":"<p>Indicates how this model version was generated.</p>"
+        },
+        "ModelQuality":{
+          "shape":"ModelQuality",
+          "documentation":"<p>Provides a quality assessment for a model that uses labels. If Lookout for Equipment determines that the model quality is poor based on training metrics, the value is <code>POOR_QUALITY_DETECTED</code>. Otherwise, the value is <code>QUALITY_THRESHOLD_MET</code>. </p> <p>If the model is unlabeled, the model quality can't be assessed and the value of <code>ModelQuality</code> is <code>CANNOT_DETERMINE_QUALITY</code>. In this situation, you can get a model quality assessment by adding labels to the input dataset and retraining the model.</p> <p>For information about improving the quality of a model, see <a href=\"https://docs.aws.amazon.com/lookout-for-equipment/latest/ug/best-practices.html\">Best practices with Amazon Lookout for Equipment</a>.</p>"
         }
       },
       "documentation":"<p>Contains information about the specific model version.</p>"