Deep Reservoir Computing Meets 5G MIMO-OFDM Systems in Symbol Detection
Authors: Zhou Zhou, Lingjia Liu, Vikram Chandrasekhar, Jianzhong Zhang, Yang Yi1266-1273
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical experiments show that the deep RC based receiver can offer a faster learning convergence and effectively mitigate unknown non-linear radio frequency (RF) distortion yielding twenty percent gain in terms of bit error rate (BER) over the shallow RC structure. |
| Researcher Affiliation | Collaboration | 1Department of Electrical and Computer Engineering, Virginia Tech, USA {zhouzhou, ljliu, yangyi8}@vt.edu 2Samsung Research America, USA {v.chandrasek, jianzhong.z}@samsung.com |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any explicit statements about releasing open-source code or links to a code repository. |
| Open Datasets | Yes | The generation of the MIMO-OFDM dataset strictly follows the 5G NR specification in (3GPP TS 38.211 2019). For example, the OFDM signal generation follows OFDM baseband signal generation defined in section 5.3.1; the functional block of modulation follows the modulation mapper defined in section 7.3.1.2; MIMO operation follows the layer mapping and the antenna port mapping defined in section 7.3.1.3 and section 7.3.1.4. The wireless transmission channel follows the WINNER II channel (Meinil a et al. 2009) which is an established channel model based on practical measurement instead of analytical models. |
| Dataset Splits | No | The paper mentions "calculate the validation loss" but does not provide specific details on the dataset split used for validation (e.g., percentages or sample counts). It only describes a training-testing procedure. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the experiments (e.g., GPU/CPU models, memory, or cloud instances). |
| Software Dependencies | No | The paper does not provide specific version numbers for any key software components or libraries used in the experiments. |
| Experiment Setup | Yes | The number of neurons is set as 128 for both shallow RC and each RC in the deep RC structure. Furthermore, a time window is added to the input layer of each RC unit, where the length is set to be 128. The state transition matrix W s is generated randomly to satisfy the echo state property (Jaeger 2001), where we choose ρ(W s) < 1. The input weight W in is generated randomly from a uniform distribution. Meanwhile, we ignore the usage of the teacher-forcing in the state update, therefore, W fb is set as zero. ... To be specific, we set Eb/N0 = 15 d B and train different RCs under the same channel realization. ... For the training set of each RC in the deep RC structure, we uniformly choose 5 delay parameters from 0 to Ncp. While for the shallow timedomain RC, we uniformly choose 50 delay parameters from 0 to Ncp. ... Here we fix the number of iterations for the ALS in solving each RC layer to be 5. |