In this paper, we considerpartitioned edge learning(PARTEL), which implements parameter-server training, a well known distributed learning method, in a wireless network. Thereby, PARTEL leverages distributed computation resources at edge devices to train a large-scaleartificial intelligence(AI) model by dynamically partitioning the model into parametric blocks for separated updating at devices. Targeting broadband channels, we consider the joint control of parameter allocation, sub-channel allocation, and transmission power to improve the performance of PARTEL. Specifically, the policies for joint SUbcarrier, Parameter, and POweR allocaTion (SUPPORT) are optimized under the criterion of minimum learning latency. Two cases are considered. First, for the case of decomposable models (e.g., logistic regression), the latency-minimization problem is a mixed-integer program and non-convex. Due to its intractability, we develop a practical solution by integer relaxation and transforming it into an equivalent convex problem of model size maximization under a latency constraint. Thereby, a low-complexity algorithm is designed to compute the SUPPORT policy. Second, consider the case ofdeep neural network(DNN) models which can be trained using PARTEL by introducing some auxiliary variables. This, however, introduces constraints on model partitioning reducing the granularity of parameter allocation. The preceding policy is extended to DNN models by applying the proposed techniques of load rounding and proportional adjustment to rein in latency expansion caused by the load granularity constraints.

In this paper, a novel accumulate-repeat-accumulate-check-accumulate (ARACA) code is proposed as a subclass of protograph-based low-density parity-check (LDPC) codes. The key feature of the proposed ARACA code is represented by the outer connection doping in the protograph. This feature can provide the linear minimum distance growth (LMDG) property at a good iterative decoding threshold while maintaining an efficient encoder structure. The effect of the outer connection doping on the typical minimum distance, the iterative decoding threshold, and the LMDG property is discussed and analyzed by comparing case examples and using the asymptotic protograph ensemble weight enumerator. Some good ARACA code protographs are provided for a wide range of code rates. In addition, an efficient and universal encoding procedure and the corresponding encoder structure are provided for them. The performance of the proposed ARACA code is evaluated and compared with well-known good LDPC codes. The simulation results confirm the superiority of the proposed ARACA codes in terms of encoding complexity and frame error rate performance, especially at low-rates in an ultra-reliable regime.

In this paper, we consider a delay and energy constrained wireless ad hoc network with node density of λn, where a packet should be delivered to the destination within D(λn) seconds using at most E(λn) energy in joules while satisfying the target outage probability. The performance metric for analyzing the network is the delay and energy constrained random access transport capacity (DE-RATC), i.e., $Cϵ(D(λn), E(λn), which quantifies the maximum end-to-end distance weighted rate per unit area of a delay and energy constrained network using a random access protocol. It is shown that a slotted ALOHA protocol is order-optimal under any delay and energy constraints if equipped with additional features such as power control, multi-hop control, interference control, and rate control, and the delay and energy constraints can be divided into three regions according to the relation between the physical quantities due to the constraints and those due to the node density and network size. The three regions are the non-constrained (NC) region, where the DE-RATC is given by Θ(√λn/logλn); the delay-constrained (DC) region, where the DE-RATC depends only on the delay constraint as Θ(D(λn); and the non-achievable (NA) region where a packet delivery under the given constraints is impossible. Also, it is shown that an arbitrary tradeoff between the rate of each source node and the number of source nodes can be achieved while keeping the optimal capacity scaling as long as λs=Ω√λn/logλn, Dλn))).

We study joint feedback design for multiple access channel (MAC) over fading Gaussian Channel. The transmitters use partial channel-state information (CSI), which is obtained via an optimized feedback link. We develop a simple iterative algorithm for the design of CSI quantization to maximize the expected rate. Also, we show the performance by simulation result.