1. What are the contributions in "Programmable arti cial intelligence machine for wave sensing and communications" ?
Here, the authors propose a programmable artificial intelligence machine ( PAIM ) that can execute various intellectual tasks by realizing hierarchical connections of 20 brain neurons via a multi-layer digital-coding metasurface array.. The authors experimentally show that PAIM can handle various deep-learning tasks for wave sensing, including image classifications, mobile communication coder-decoder, and real-time multi-beam focusing.. In particular, the authors propose a reinforcement learning algorithm for on-site learning and discrete optimization algorithm for digital coding, making PAIM have autonomous intelligence ability and perform self-learning tasks without the support of extra computer.. Here, the authors propose a programmable and on-site trainable artificial intelligence machine ( PAIM ) using an array of information metasurfaces for wave sensing and communications, in which the multi-layer metasurfaces act as the programmable physical layers of DNN.. The authors design their PAIM to be a real-time re-trainable system, whose parameters could be set in digital to realize alive artificial neurons.. In the physical layer, PAIM could hierarchically manipulate the energy distribution of transmitted EM waves by a five-layer information metasurface array, from which the amplitude of transmitted wave through each meta-atom could be enhanced or attenuated by controlling the 70 value of digital parameters ( Fig. 1a ).. Here, the reprogrammable interconnection architectures in PAIM is the fundamental and essential factor to simulate the alive artificial neurons.. Therefore, the forward propagation model ( see Supplementary Materials Note 3 ‘ Forward Propagation Model ’ ) of PAIM can be regarded as a fully-connected network ( Supplementary Fig. S2 ).. However, compared with the traditional fully-connected network constructed by real numbers, the PAIM parameters have complex values and the trainable parts are complex-valued transmission coefficients of the metaatoms.. The traditional error 90 back-propagation method could be used to train the PAIM parameters.. Meanwhile, owing to the fast parameter-switching ability and direct feedback from receivers ( Supplementary Fig. S1 ), their PAIM enables self-learning capability by using the data gained from the direct interaction with environment, and does not need any prior knowledge.. To verify the powerful capabilities of PAIM, the authors firstly use it to deal with two image classification tasks: oil painting style ( Fig. 1e, g, Supplementary Fig. S4 ) and handwritten digit ( Supplementary Fig. S3 ) classifications.. In the first classification task with two kinds of oil paintings ( portraiture and landscape painting ), the authors simulate a PAIM with 6-layer metasurfaces, each of which consists of 25×25 programmable meta-atoms.. The input image ( Fig. 1e ) is grayed and reshaped to 25×25 pixels ( corresponding to the size of metasurface ) ( Fig. 1f ), and then inputted to PAIM by configuring the first-layer metasurface, in which the transmission coefficient of each meta-atom is set as the corresponding pixel value of the image.. At the end of PAIM, the authors assign 2 receivers to get the 2 kinds of oil paintings.. More details on the network design 120 and recognition results are provided in Supplementary Materials.. For demonstrating the versatility of their PAIM in real world, the authors design and fabricate a PAIM sample with five-layer programmable metasurfaces controlled by five FPGA modules, and each layer consists of 8×8 meta-atoms ( Figs. 1a, 3b ).. The support structure of the PAIM sample is presented in Supplementary Fig. S7, in which the first layer ( i. e., the input layer ) is illuminated by microwaves at 5.. 130 To test the real experimental performance of PAIM in image classification, the authors design two imaginative cases.. As mentioned above, the first-layer metasurface of PAIM acts as a digital-to-analog converter to convert the input image into the corresponding spatial distribution of EM waves.. The remaining four layers act as a recognizer, and several receiving antennas are put at the end of PAIM.. More details on the discrete optimization algorithm and calibration approach are provided in the Supplementary Materials.. The pixels belonging to pattern parts and background are allocated with different bias voltages of metaatoms in the first PAIM layer.. The authors down-sample the original prop image into an 8×8 pixel matrix and use different bias voltages of meta-atoms in the first PAIM layer to represent different pixel values.. The bias-voltage configurations of layers 2-5 ( corresponding to the recognition part of PAIM ) for this case are shown in Fig. 2I.. Besides the image classification, the authors further use PAIM for mobile communication codec, which can perform coding and decoding tasks in Code Division Multiple Access ( CDMA ) scheme, and transmit four kinds of orthogonal user codes simultaneously or separately in one channel.. As shown in Fig. 3a, the first-layer PAIM 160 metasurface is set as an encoder, on which each meta-atom sequentially corresponds to one bit in the binary number string.. The authors put four receiving antennas at the end of PAIM, and each antenna represents a user code.. The authors use to represent the four user codes, and to represent the receiving energies of the corresponding antennas.. The remaining four-layer PAIM metasurfaces are trained as a decoder.. When is transmitted by the first layer, the values of would be, in which the function f represents the linear forward propagation function of PAIM, and the term low indicates that the receiving energy is much less than that of high.. The experimental results verify the feasibility of PAIM in wireless communications ( Fig. 3c-f, Supplementary Fig. S11 ).. Compared with the traditional CDMA scheme, their PAIM performs coding and decoding using space dimension instead of time dimension, and realizes the Open System Interconnection ( OSI ) reference model in the physical layer instead of link layer ( Supplementary Fig. S12 ).. On the other hand, the strong capability of processing distributed space EM waves makes PAIM a good candidate to realize space division multiplexing and thus increases the channel capacity.. The 190 decoding function of PAIM is operated as a dependent system and is able to deal with signals from distributed communication base stations ( Supplementary Fig. S13 ).. Finally, the authors turn their PAIM into a dynamic multi-beam focusing lens, which could focus the EM energy on multiple points with arbitrary positions.. Different from the aforementioned cases, in which the training process is executed on a computer in advance, here the authors directly make on-site training to PAIM using reinforce-learning method in real time, which completely overcomes the limitation to require priori knowledge in the previous optical DNN platforms.. Benefit from the real-time programmable ability of PAIM, the authors can train the parameters by continuously interacting with 200 { } 1 2 3 4,,, C C C C { } 1 2 3 4,,, E E E E 1 C { } 1 2 3 4,,, E E E E { } 1 ( ),,, f C high low low low =
read more
2. What is the advantage of the reinforce learning?
One advantage of the reinforce learning is the result-oriented strategy, in which the authors do not needto worry about the accuracy of simulations or other factors that could make the designed parametersdeviate from the measurement results.
read more
3. What is the effect of the calibration process of the space attenuation coefficients?
The coupling effect cannot beeliminated by the calibration process of the space attenuation coefficients, because it is a nonlinearinteraction.
read more
4. What is the advantage of the reinforce learning process?
The reinforce-learning process doesnot need pre-prepared training data, while updates the configuration parameters according to thefeedback by interacting with environment.
read more
![Fig. 2 | Experimental results of image classifications using PAIM. a-d, Two kinds of patterns (letter ‘I’ and bracket ‘[]’) are represented by the distributions of bias voltages for 8×8 meta-atoms in first PAIM layer. The input image consists of 8×8 blocky squares of colors and each square represents the bias voltage of the meta-atom. The meta-atoms in the remaining four layers are assigned the bias voltages designed from the training process and could recognize the two patterns by ranking the receiving energies from the two receiving antennas. The patterns of letter ‘I’ in (a)](/figures/fig-2-experimental-results-of-image-classifications-using-3qq6mrc5.webp)
