Fig. 9: A detailed schematic of the proposed wave-based ELM architecture.
From: Electromagnetic wave-based extreme deep learning with nonlinear time-Floquet entanglement
On the left, the input side includes a signal generator (SG) with variable attenuators (VOA) that encode the information. A Voltage-controlled phase shifter (VCP) is used to change the phase of the modulation depending on the voltage applied, which is fixed for each input and does not change in time. AA denotes the array of source antennas. Vp is the required voltage to control the VCP (\({V}_{p}=\gamma {\bar{\zeta }}^{{{{{{\mathrm{in}}}}}}}\)), where \({\bar{\zeta }}^{{{{{{\mathrm{in}}}}}}}\) and γ are the mean of input vector and a scaling factor, respectively.
