Table 2.

Model architecture of PPG Encoders in CLIP.

Layer	PPG embedding			Linear projection
	Type	Output	($N_f$, $K_s$,s)	Layer	Output
1	Input	(1250, 1)	–	GELU	(256)
2	Conv1D	(1241, 8)	(64, 10, 1)	Dense	(256)
3	MaxPooling1D	(620, 8)	–	Dropout	(256)
4	Conv1D	(616, 8)	(32, 5, 1)	ADD	(256)
5	MaxPooling1D	(308, 8)	–
6	Conv1D	(306, 16)	(16, 3, 1)
7	Conv1D	(304, 32)	(8, 3, 1)
8	Flatten	(9728)	–

Here, $N_f$, $K_s$, and s denote the number of filters, kernel size, and stride respectively.