Decision tree is a supervised machine learning algorithm suitable for solving classification and regression problems. Decision trees are recursively built by applying split conditions at each node that divides the training records into subsets with output variable of same class. The process starts from the root node of the decision tree and progresses by applying split conditions at each non-leaf node resulting into homogenous subsets. However, achieving pure homogenous subsets is not possible. Therefore, the goal at each node is to identify an attribute and a split condition on that attribute that minimizes the mixing of class labels, thus resulting into nearly pure subsets. Several splitting indices were proposed to evaluate the goodness of the split, common ones being GINI index and Information gain. The aim of this study is to conduct an empirical comparison of GINI index and information gain. Classification models are built using decision tree classifier algorithm by applying GINI index and Information gain individually. The classification accuracy of the models is estimated using different metrics such as Confusion matrix, Overall accuracy, Per-class accuracy, Recall and Precision. The results of the study show that, regardless of whether the dataset is balanced or imbalanced, the classification models built by applying the two different splitting indices GINI index and information gain give same accuracy. In other words, choice of splitting indices has no impact on performance of the decision tree classifier algorithm.

/pdf/evaluating-the-impact-of-gini-index-and-information-gain-on-38jdvis39e.pdf

Evaluating the Impact of GINI Index and Information Gain on Classification using Decision Tree Classifier Algorithm

Prediction performance of improved decision tree-based algorithms: a review

It is important to detect cerebral microbleed voxels from the brain image of cerebral autosomal-dominant arteriopathy with subcortical infarcts and Leukoencephalopathy (CADASIL) patients. Traditional manual method suffers from intra-observe and inter-observe variability. In this study, we used the susceptibility weighted imaging (SWI) to scan 10 CADASIL patients and 10 healthy controls. We used slicing neighborhood processing (SNP) to extract “input” and “target” dataset from the 20 brain volumetric images. Afterwards, the undersampling technique was employed to handle the class-imbalanced problem. The single-hidden layer feedforward neural-network with scaled conjugate gradient was used as the classifier. We compared three activation functions: logistic sigmoid (LOSI), rectified linear unit (ReLU), and leaky rectified linear unit (LReLU). Early stopping and K-fold cross validation (CV) was used to avoid overfitting and statistical analysis. In the experiment, we generated 68,847 CMB voxels, and 68,829 non-CMB voxels. We observed that LReLU achieved the best result with a sensitivity of 93.05%, a specificity of 93.06%, and an accuracy of 93.06%. We also observed the effect of early stopping and K-fold CV. We found the optimal number of hidden neuron was 10 by grid searching method. Besides, our method performs better than three state-of-the-art methods. The results show our method is promising. In addition, LReLU is a better activation function that may replace traditional logistic sigmoid function in other applications.

Voxelwise detection of cerebral microbleed in CADASIL patients by leaky rectified linear unit and early stopping

Rough pavements cause ride discomfort and energy inefficiency for road vehicles. Existing methods to address these problems are time-consuming and not adaptive to changing driving conditions on rough pavements. With the development of sensor and communication technologies, crowdsourced road and dynamic traffic information become available for enhancing driving performance, particularly addressing the discomfort and inefficiency issues by controlling driving speeds. This study proposes a speed control framework on rough pavements, envisioning the operation of autonomous vehicles based on the crowdsourced data. We suggest the concept of ‘maximum comfortable speed’ for representing the vertical ride comfort of oncoming roads. A deep reinforcement learning (DRL) algorithm is designed to learn comfortable and energy-efficient speed control strategies. The DRL-based speed control model is trained using real-world rough pavement data in Shanghai, China. The experimental results show that the vertical ride comfort, energy efficiency, and computation efficiency increase by 8.22%, 24.37%, and 94.38%, respectively, compared to an optimization-based speed control model. The results indicate that the proposed framework is effective for real-time speed controls of autonomous vehicles on rough pavements. 

Comfortable and energy-efficient speed control of autonomous vehicles on rough pavements using deep reinforcement learning

From its first appearance in early vertebrates, the spine evolved the function of protecting the spinal cord, avoiding excessive straining during body motion. Its stiffness and strength provided the basis for the development of the axial skeleton as the mechanical support of later animals, especially those which moved to the terrestrial environment where gravity loads are not alleviated by the buoyant force of water. In tetrapods, the functions of the spine can be summarized as follows: protecting the spinal cord; supporting the weight of the body, transmitting it to the ground through the limbs; allowing the motion of the trunk, through to its flexibility; providing robust origins and insertions to the muscles of trunk and limbs. This narrative review provides a brief perspective on the development of the spine in vertebrates, first from an evolutionary, and then from an embryological point of view. The paper describes functions and the shape of the spine throughout the whole evolution of vertebrates and vertebrate embryos, from primordial jawless fish to extant animals such as birds and humans, highlighting its fundamental features such as strength, stability, and flexibility, which gives it huge potential as a basis for bio-inspired technologies.

The Spine: A Strong, Stable, and Flexible Structure with Biomimetics Potential.

The performance of vehicle body vibration and ride comfort of active or semi-active suspension with proper control is better than that with passive suspension. The key to achieve good control effect is that the suspension control system should have strong real-time learning ability according to changes in the road surface and suspension parameters. In the control strategies adopted by previous researchers, the classical neural network controller has some learning ability, but it is mainly based on offline learning with a large number of samples. In this paper, the deep reinforcement learning strategy is used to solve the above problems.Aiming at the continuity of state space and execution action in vehicle active suspension system, the control of the semi-active suspension is realized by using improved DDPG (Deep Deterministic Policy Gradient) algorithm. To overcome the shortcoming of low efficiency of this algorithm in the initial stage of learning, the DDPG algorithm is improved and using empirical samples in the learning method is proposed. Based on Mujoco, the physical model of semi-active suspension is established, and its dynamic characteristics are analyzed under the condition of various road level and vehicle speed. The simulation results show that compared with the passive suspension, the semi-active suspension based on improved DDPG algorithm with learning method using experienced samples can better adapt to various road level, more effectively reduce the vertical acceleration of the vehicle body and the dynamic deflection of the suspension, and further improve the ride comfort.

/pdf/semi-active-suspension-control-based-on-deep-reinforcement-3elm5mcmcm.pdf

Semi-Active Suspension Control Based on Deep Reinforcement Learning

In data classification mining, the decision tree method is a key algorithm. ID3 (Iterative Dichotomiser 3) algorithm which was presented by Quinlan is a famous decision tree algorithms, but ID3 has some shortcomings such as high complex computation in computing the information entropy expression, multivalue bios problem in the process of selecting an optimal attribute, large scales, etc. In order to solve the above problems, the improved ID3 algorithm is proposed, which combines the simplified information entropy with coordination degree in rough set theory. The experiment result has proved the feasibility of the optimized way.

Improvement of ID3 algorithm based on simplified information entropy and coordination degree

At present, the path-planning algorithm based on the grid map is generally adopted in the field of quadruped robot and the obtained environmental information is represented by a standardized grid map. In this paper, the ADFA* algorithm introduces a dilation factor based on the DFA* to solve the path planning problem under the constraint of computing time and provide a path search result related to the time limit. Path-planning algorithms based on raster maps often equate robots with particles, causing problems, such as path blocking. The FA* algorithm adds raster tolerance to expand obstacles. DFA* uses a path-splitting approach that, such as the DA* algorithm, has better dynamic environment processing capabilities than the FA*. However, during the actual operation of the robot, the environmental information acquired is extremely frequent due to its instability. The robot itself is often in a relatively static state. Therefore, compared with obtaining the shortest path, it is more practical to improve the path search efficiency under dynamic map environment. ADFA* will gradually optimize the path and eventually get the optimal solution when time is sufficient. When the time is limited, ADFA* will search for the current optimal solution under the specified search time but may not be able to obtain the shortest path, which is called the second best solution.

/pdf/path-planning-based-on-adfa-algorithm-for-quadruped-robot-3vx08xr7rr.pdf

Path Planning Based on ADFA* Algorithm for Quadruped Robot

The single hidden layer feed-forward neural networks learning algorithms with good performance are divided into batch learning and sequential (on-line) learning algorithms. Compared with batch learning algorithms, the sequential learning algorithms of feed-forward neural networks are suitable for problems of real-time processing, and are more adaptable for general industrial applications. This paper summarizes different sequential learning algorithms of single hidden layer feed-forward neural networks and analyzes the advantages and disadvantages of various algorithms. The performance including the stability, learning speed, approximation and generalization ability of different sequential learning algorithms are compared in detail in terms of different chaotic time series prediction problems. The simulation results provide the theoretic guidance on real applications of sequential learning algorithms of feed-forward neural networks.

Review and performance analysis of single hidden layer sequential learning algorithms of feed-forward neural networks

Firstly, mathematic model of rock drill robot joint hydraulic drive system was analyzed, and the reduced-order approximative model was acquired. Then a two-order ADRC (Active Disturbance Rejection Controller) controller of the hydraulic system was designed, and the simulations were down within uncertainty environments. The simulation results show that the ADRC controller has ideal robustness to the system parameters' disturbances and the large load disturbance, and rapid and smooth control process and high steady precise performances can be implemented.

Rong Xuewen

Papers

Semi-Active Suspension Control Based on Deep Reinforcement Learning

Improvement of ID3 algorithm based on simplified information entropy and coordination degree

Path Planning Based on ADFA* Algorithm for Quadruped Robot

Review and performance analysis of single hidden layer sequential learning algorithms of feed-forward neural networks

Study on ADRC Controller Design and Simulation of Rock Drill Robot Joint Hydraulic Drive System