Table 1 The performance of our proposed IDNN framework.

From: Space-efficient optical computing with an integrated chip diffractive neural network

Type

The no. modulators

Matrix dimension

Area (10 × 10 neurons)

Power consumption (10 × 10 neurons)

MZI-VMM10

16

4 × 4

5.2 mm2

------

MZI-VMM7

9

4 × 4

6 mm2

3 W

MZI-VMM13

15

4 × 4

4.8 mm2

7 W

Our work

10

10 × 10

0.53 mm2

17.5 mW

Back to article page