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
Hello there, firstly thanks for making this amazing and easy to use tool!
I was actually going through the documentation to better understand how Tangram works and was a bit confused about the explanation in the "Tangram under the hood" page in the documentation website (https://tangram-sc.readthedocs.io/en/latest/working.html).
I think there may be two typos in the website, which I crosschecked by going into the explanation of the workings in the Tangram manuscript.
As we can see in the image below, the shape of matrix M is said to be voxels by cells, which would make the matrix multiplication $M^{T}S$ invalid as shape of S is cells by genes. Secondly, in the loss function, we seem to have (positive) sum of cosine similarities, which is not what we would want as that is exactly the quantity we would like to maximize, not minimize.
Below image highlights the correct statements in the manuscript's methods.
Regards,
Shashank Katiyar
The text was updated successfully, but these errors were encountered:
Hello there, firstly thanks for making this amazing and easy to use tool!
I was actually going through the documentation to better understand how Tangram works and was a bit confused about the explanation in the "Tangram under the hood" page in the documentation website (https://tangram-sc.readthedocs.io/en/latest/working.html).
I think there may be two typos in the website, which I crosschecked by going into the explanation of the workings in the Tangram manuscript.
As we can see in the image below, the shape of matrix M is said to be voxels by cells, which would make the matrix multiplication$M^{T}S$ invalid as shape of S is cells by genes. Secondly, in the loss function, we seem to have (positive) sum of cosine similarities, which is not what we would want as that is exactly the quantity we would like to maximize, not minimize.
Below image highlights the correct statements in the manuscript's methods.
Regards,
Shashank Katiyar
The text was updated successfully, but these errors were encountered: