Advances, Systems and Applications
Notation | Description |
---|---|
\(\textbf{M}\) | DNN model |
\(\mathbf {M_{edge}}\) | edge part of DNN model |
\(\mathbf {M_{cloud}}\) | cloud part of DNN model |
\(L_i\) | the ith layer of model |
\(w_i\) | weight of the ith layer |
\(\widehat{w_i}\) | quantized weight of the ith layer |
\(a_i\) | activation of the ith layer |
\(\widehat{a_i}\) | quantized activation of the ith layer |
\(Q_b\) | quantization function with b bit precision |
\(\pi _i\) | quantization policy of the ith layer |
\(\Pi\) | quantization policy of DNN model |
\(B_i\) | the ith Block of model |
P | privacy evaluation function |
p | partition point of DNN model |
E | energy measurement of quantized model |
\(E_L\) | inference energy consumption of one layer |
\(E_T\) | energy consumption of network transmission |
\(\epsilon\) | quantization loss evaluation function |
\(E_b\) | energy consumption constraint |
\(\varepsilon _b\) | quantization loss constraint |