Analysis of Soft Decision Trees for Passive-Expert Reinforcement Learning

This paper explores the use of soft decision trees [1] in basic reinforcement applications to examine the efficacy of using passive-expert like networks for optimal Q-Value learning on Artificial Neural Networks (ANN). The soft decision tree networks were built using the PyTorch machine learning and the OpenAi’s Gym environment frameworks. The conducted research study aimed at assessing the performance of soft decision tree networks on Cartpole as provided in the OpenAi Gym software package. The baseline performance metric that the soft decision tree networks were compared against was a simple Deep Neural Network using several linear layers with ReLU and Softmax activation functions for the input and output layers, respectively. All networks were trained using the Backpropagation algorithm provided generically by PyTorch’sAutograd module.

Soft Decision Tree for Regression

Decision tree(DT)plays an important role in pattern recognition and machine learning,which is widely used for regression tasks because of its natural interpretability.Nevertheless,the traditional decision tree is constructed by recursive Boolean division.The discrete decision-making process in DT makes it non-differentiable,and causes the problem of hard decision boundary.To solve this problem,a probability distribution model—Staired-Sigmoid is proposed in this paper.The Staired-Sigmoid model is used to differentiate the decision-making process,by which the samples can be assigned to two sub-trees more finely.Based on Staired-Sigmoid,we further propose the soft decision tree(SDT)for regression tasks,where the samples are assigned to different sub-nodes according to a continuous probability distribution.This process is differentiable,and all parameters in SDT can be optimized by gradient descent algorithms.Owing to its constructing rules,SDT is more stable than decision tree,and it is easier to overcome the problem of overfitting.We validate SDT on several datasets obtained from UCI.Experiments demonstrate that SDT achieves better performance than decision tree,and it significantly alleviates the overfitting.

LDPC Code’s Decoding Algorithms for Wireless Sensor Network:a Brief Review

As an effective error correction technology,the Low Density Parity Check Code(LDPC)has been researched and applied by many scholars.Meanwhile,LDPC codes have some prominent performances,which involves close to the Shannon limit,achieving a higher bit rate and a fast decoding.However,whether these excellent characteristics are suitable for the resource-constrained Wireless Sensor Network(WSN),it seems to be seldom concerned.In this article,we review the LDPC code’s structure brief.ly,and them classify and summarize the LDPC codes’construction and decoding algorithms,finally,analyze the applications of LDPC code for WSN.We believe that our contributions will be able to facilitate the application of LDPC code in WSN.

Sensor node-assisted asynchronous cooperative spectrum sensing for cognitive radio network

In order to take advantage of the asynchronous sensing information, alleviate the sensing overhead of secondary users （SUs） and improve the detection performance, a sensor node-assisted asynchronous cooperative spectrum sensing （SN-ACSS） scheme for cognitive radio （CR） network （CRN） was proposed. In SN-ACSS, each SU is surrounded by sensor nodes （SNs）, which asynchronously make hard decisions and soft decisions based on the Bayesian fusion rule instead of the SU. The SU combines these soft decisions and makes the local soft decision. Finally, the fusion center （FC） fuses the local soft decisions transmitted from SUs with different weight coefficients to attain the final soft decision. Besides, the impact of the statistics of licensed band occupancy on detection performance and the fact that different SNs have different sensing contributions are also considered in SN-ACSS scheme. Numerical results show that compared with the conventional synchronous cooperative spectrum sensing （SCSS） and the existing ACSS schemes, SN-ACSS algorithm achieves a better detection performance and lower cost with the same number of SNs.

Iterative multi-user detection and decoding for space-time block coding systems

To restrain the interference of co-channel users using space-time block coding （STBC）, the proposed Gaussian-forcing soft decision multi-user detection （GFSDMUD） algorithm is applied in fiat-fading channels by using the relation among the users＇ signals, which can enhance the capacity by introducing co-channel users. During iterations, extrinsic information is calculated and exchanged between a soft multi-user detector and a bank of turbo decoders to achieve refined estimates of the users＇ signals. The simulations show that the proposed iterative receiver techniques provide significant performance improvement around 2 dB over conventional noniterative methods. Furthermore, iterative multi-user space-time processing techniques offer substantial performance gains around 8 dB by adding the number of receiver antennas from 4 to 6, and the system performance can be enhanced by using this strategy in multi-user STBC systems, which is very important for enlarging the system capacity.