about the interpretation of embeddings.csv #42
-
|
While examining the embeddings.csv file created by flexynesis, I had a question regarding the interpretation of the embeddings. Would it be more accurate to evaluate the embeddings of cell lines within themselves (i.e., examining how each cell line relates to its own embedding)? Or, would evaluating cell lines with similar embeddings (i.e., comparing cell lines with similar features) provide better insights into the patterns? Thank you. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
|
In this context embeddings are used to compare cell lines with each other. Each cell line is represented by a numerical vector (embedding). Such a numerical representation allows us to compute similarities/differences between different samples (cell lines). In our use case, as we are talking about predicting an outcome using these embeddings, the way the cell lines cluster/associate with each other should be driven by the target variable. This is the primary purpose of seeing if the embeddings actually reflect a clustering of samples based on what we trained the network to learn. |
Beta Was this translation helpful? Give feedback.
-
|
Thank you for your reply. |
Beta Was this translation helpful? Give feedback.
In this context embeddings are used to compare cell lines with each other. Each cell line is represented by a numerical vector (embedding). Such a numerical representation allows us to compute similarities/differences between different samples (cell lines). In our use case, as we are talking about predicting an outcome using these embeddings, the way the cell lines cluster/associate with each other should be driven by the target variable. This is the primary purpose of seeing if the embeddings actually reflect a clustering of samples based on what we trained the network to learn.