Skip to main content
. 2021 Jul 31;21(15):5218. doi: 10.3390/s21155218

Table 2.

Memory footprint of the multi-layer perceptron (MLP) TinyML implementation on the wearable device.

Layers\Neurons per Layer 2 5 10 15
1 Peripherals: 20,812 B
TinyML: 2490 B
Total: 72.2%		 Peripherals: 20,812 B
TinyML: 2550 B
Total: 72.4%		 Peripherals: 20,812 B
TinyML: 2650 B
Total: 72.7%		 Peripherals: 20,812 B
TinyML: 2750 B
Total: 73%
2 Peripherals: 20,812 B
TinyML: 2514 B
Total: 72.3%		 Peripherals: 20,812 B
TinyML: 2586 B
Total: 72.5%		 Peripherals: 20,812 B
TinyML: 2706 B
Total: 72.9%		 Peripherals: 20,812 B
TinyML: 2826 B
Total: 73.2%
3 Peripherals: 20,812 B
TinyML: 2538 B
Total: 72.3%		 Peripherals: 20,812 B
TinyML: 2622 B
Total: 72.5%		 Peripherals: 20,812 B
TinyML: 2762 B
Total: 73%		 Peripherals: 20,812 B
TinyML: 2902 B
Total: 73.5%
4 Peripherals: 20,812 B
TinyML: 2562 B
Total: 72.4%		 Peripherals: 20,812 B
TinyML: 2658 B
Total: 72.7%		 Peripherals: 20,812 B
TinyML: 2818 B
Total: 73.2%		 Peripherals: 20,812 BTinyML: 2978 BTotal: 73.7%
5 Peripherals: 20,812 B
TinyML: 2586 B
Total: 72.5%		 Peripherals: 20,812 B
TinyML: 2694 B
Total: 72.8%		 Peripherals: 20,812 B
TinyML: 2874 B
Total: 73.4%		 Peripherals: 20,812 B
TinyML: 3054 B
Total: 74%