Abstract Deep nets have done well with early adopters, but the future will soon depend on crossing the chasm. The goal of this paper is to make deep nets more accessible to a broader audience including people with little or no programming skills, and people with little interest in training new models. A github is provided with simple implementations of image classification, optical character recognition, sentiment analysis, named entity recognition, question answering (QA/SQuAD), machine translation, speech to text (SST), and speech recognition (STT). The emphasis is on instant gratification. Non-programmers should be able to install these programs and use them in 15 minutes or less (per program). Programs are short (10–100 lines each) and readable by users with modest programming skills. Much of the complexity is hidden behind abstractions such as pipelines and auto classes, and pretrained models and datasets provided by hubs: PaddleHub, PaddleNLP, HuggingFaceHub, and Fairseq. Hubs have different priorities than research. Research is training models from corpora and fine-tuning them for tasks. Users are already overwhelmed with an embarrassment of riches (13k models and 1k datasets). Do they want more? We believe the broader market is more interested in inference (how to run pretrained models on novel inputs) and less interested in training (how to create even more models).

Most previous researches on the rate analysis of the power allocation (PA) schemes in two-way decode-and- forward (DF) relay systems typically focus on maximizing the instantaneous objective rates. However, the ergodic rate analysis of the optimized rates is not well considered. In this paper, we investigate the ergodic rates of the PA schemes in a two-way DF relay system, where the physical-layer network coding (PNC) protocol is adopted. Specifically, under a total power constraint, we consider three heuristic PA schemes, in which the optimization objectives are: 1) maximizing the sum rate; 2) maximizing the sum rate with an additional rate fairness constraint; and 3) maximizing the minor rate of the two directions of the system. The cumulative distribution functions (CDFs) of the equivalent system SNR are approximately obtained and the analytical expressions for the calculations of the ergodic rates are theoretically derived. Numerical results show that the derived analytical rates are converged to the simulated results in high SNR regions.

In this paper, we investigate the power allocation (PA) schemes in a two-way relay system, where the relay adopts physical layer network coding (PNC) based on decode-and-forward (DF) protocol. The objective is to determine the optimal power allocation (OPA) scheme at both the source nodes and at the relay that minimizes the end-to-end symbol error probability (SEP) performance under a total power constraint. The perfect channel state information (CSI) is assumed to be available at both the source nodes and at the relay. Since numerical methods (e.g., exhaustive search) are required to find the solution of the OPA scheme, which might be infeasible for practical systems, based on upper and lower bounds on the optimal solution, we propose a near optimal power allocation (NOPA) scheme. The asymptotic SEP expression of the proposed NOPA scheme is also derived. By both the analytical and simulated results, it is shown that the proposed NOPA scheme provides the asymptotic SNR gains of about 1.76dB and 2.22dB over two well-known schemes in high SNR. Furthermore, simulation results show that the performance of proposed NOPA scheme is quite close to that of the OPA scheme.

In this study, the authors examine the power allocation (PA) in a two‐way relay system, where the relay adopts physical layer network coding based on decode‐and‐forward protocol. In the PA, the optimisation objective is to minimise the end‐to‐end symbol error probability (SEP) under a total power constraint. Since the optimal power allocation (OPA) scheme requires numerical methods (for exhaustive search) to find the solution, which might be infeasible for practical systems, the authors propose a near optimal power allocation (NOPA) scheme that has a closed‐form solution by studying the properties of the OPA scheme. The asymptotic SEP expression of the proposed NOPA scheme is also derived. Simulation results show that the performance of proposed NOPA scheme is quite close to that of the OPA scheme. Furthermore, it is shown that the NOPA scheme can provide asymptotic SNR gains of 1.76 dB and 2.22 dB over two well‐known schemes.

In this paper, we study the impact of outdated channel state information (CSI) on the error rate performance in multiple-antenna two-way decode-and-forward (DF) relay systems, where physical-layer network coding (PNC) is adopted. We consider two variations of relay selection, namely, single best relay selection (S-RS) and multiple successfully-participated relay selection (MSP-RS). Under perfect and outdated CSI assumptions, closed-form expressions for the end-to-end system symbol error probability (SEP) are derived for the S-RS scheme, along with the asymptotic SEP expressions in high SNR regions for the MSP-RS scheme. By both analytical and simulation results, it is clearly shown that a full cooperative diversity gain can be achieved with perfect CSI. The results also manifest that the S-RS scheme can achieve only the transmit diversity gain with outdated CSI. On the other hand, the MSP-RS scheme can achieve a full cooperative diversity gain even if the CSI is outdated at the expense of high complexity.