You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
1. **Cosine Similarity**: The cosine similarity between the centroids of synthetic and training samples.
4725
4817
2. **Discriminator AUC**: The AUC of a discriminative model to distinguish between synthetic and training samples.
4726
-
4727
-
The SentenceTransformer model [all-MiniLM-L6-v2](https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2) is
4818
+
4819
+
The Model2Vec model [potion-base-8M](https://huggingface.co/minishlab/potion-base-8M) is
4728
4820
used to compute the embeddings of a string-ified representation of individual records. In case of sequential data
4729
4821
the records, that belong to the same group, are being concatenated. We then calculate the cosine similarity
4730
4822
between the centroids of the provided datasets within the embedding space.
@@ -4796,6 +4888,12 @@ components:
4796
4888
format: "double"
4797
4889
minimum: 0.0
4798
4890
maximum: 1.0
4891
+
imsTrnHol:
4892
+
description: "Share of training samples that are identical to a holdout sample. Serves as a reference for `ims_training`."
4893
+
type: "number"
4894
+
format: "double"
4895
+
minimum: 0.0
4896
+
maximum: 1.0
4799
4897
dcrTraining:
4800
4898
description: "Average L2 nearest-neighbor distance between synthetic and training samples."
4801
4899
type: "number"
@@ -4806,6 +4904,11 @@ components:
4806
4904
type: "number"
4807
4905
format: "double"
4808
4906
minimum: 0.0
4907
+
dcrTrnHol:
4908
+
description: "Average L2 nearest-neighbor distance between training and holdout samples. Serves as a reference for `dcr_training`."
4909
+
type: "number"
4910
+
format: "double"
4911
+
minimum: 0.0
4809
4912
dcrShare:
4810
4913
description: "Share of synthetic samples that are closer to a training sample than to a holdout sample. This should not be significantly larger than 50%."
0 commit comments