m/z and intensity embedding

[DawnJYe]

m/z and intensity are embedded to yield peak embedding by FloatEncoder in depthcharge,both are sinusoidal embedding.But in your paper,m/z and intensity are embedded separately,and intensity is embedded through a linear layer.So which way do you do it?

[wfondrie]

Hi @XiaojieGan - the published versions of Casanovo using a sinusoidal embedding for the m/z values and a linear projection for the intensity values. However, in more recent versions of the underlying library, Depthcharge, I've switched to using a second sinusoidal embedding as you noted.

#350 Is the PR upgrading to this new version of Depthcharge.

[DawnJYe]

Thanks for your answer.So you embedded both m/z and intensity by sinusoidal embedding in your implementation.I‘d like to take this approach,but I don't know it's effect.

[DawnJYe]

Hi @XiaojieGan - the published versions of Casanovo using a sinusoidal embedding for the m/z values and a linear projection for the intensity values. However, in more recent versions of the underlying library, Depthcharge, I've switched to using a second sinusoidal embedding as you noted.

#350 Is the PR upgrading to this new version of Depthcharge.

In your ablation experiment,there are a variety of approaches to be mentioned. "Sinusoidal I embedding",intensity is embedded by sinusoidal embedding. But "Standard Casanovo" using a linear projection for the intensity,how did you do it?