Evolution of Neuro-controllers for Multi-link Robots

This paper presents the development of several efficient learning algorithms for neural networks (ELEANNE). The ELEANNE 1 and ELEANNE 2 are two recursive least-squares learning algorithms, proposed for training single-layered neural networks with analog output. This paper also proposes a new optimization strategy for training single-layered neural networks, which provides the basis for the development of a variety of efficient learning algorithms. This optimization strategy is the source of the ELEANNE 3, a second-order learning algorithm for training single-layered neural networks with binary output. A simplified version of this algorithm, called ELEANNE 4, is also derived on the basis of some simplifying but reasonable assumptions. The two algorithms developed for single-layered neural networks provide the basis for the derivation of ELEANNE 5 and ELEANNE 6, which are proposed for training multilayered neural networks with binary output. The ELEANNE 7 is an efficient algorithm developed for training multilayered neural networks with either binary or analog output. >

The use of numerical prediction models are essential to modern society. Data assimilation is a technique that aims to increase the prediction accuracy by combining a model output with observational data, resulting in a state that is closer to the true state of the problem. Depending on the size of the model output and the number of observations to assimilate, the combination of these two sources of information may require intensive computing and become a challenge, even for supercomputers used in this type of application. Thus neural networks have been proposed as an alternative to perform high quality data assimilation at lower computational cost. This paper investigates the use of NeuroEvolution of Augmenting Topologies (NEAT) in data assimilation. NEAT is capable of adapting the connections weights and the neural network topology using principles of evolutionary computation in a search for a minimum topology and best performance. In this work, two different models were used for testing: the Lorenz Attractor and Shallow Water model. The experiments compared the results obtained with NEAT and backpropagation neural networks, using as benchmark the Best Linear Unbiased Estimator (BLUE). In the experiment with the Lorenz Attractor, NEAT was able to emulate the data assimilation task with smaller error at lower computational cost. For the Shallow Water model, tested using different grid sizes, it was observed that the errors obtained with both neural networks were small, but NEAT showed high error values. On the other hand, NEAT always gets a topology with significantly fewer operations, and the computational cost difference increases with the grid size.

The Sixth International Symposium on Neural Networks (ISNN 2009)

Cars are becoming more and more intelligent, embedded with a range of sensors to give them local perception of their environment (LIDARs, cameras, etc.). Trendy companies like Google and Tesla are actively testing cars on American roads that can drive without any human interaction [1]. Neural networks are the modern approach for autonomous cars. However, an inefficient neural network algorithm will make the learning process slower and will result in a less reliable autonomous vehicle. In this paper, we will introduce a platform built in JAVA named LAOP (Learning Algorithm Optimization Platform) [2] while explaining the solutions we found to make it easy for researchers to test and compare their own algorithms. Then, we will show how we have integrated a natural selection algorithm with a neural network in order to improve them. Moreover, we will demonstrate how the Fully Connected Neural Network and the NeuroEvolution of Augmenting Topologies (NEAT) [3] algorithms are implemented in the context of vehicular learning on LAOP. Finally, we will display the different results extracted from LAOP by tuning several various parameters such as the weight mutation chance and the car density in the simulation.

Decision-making of online rescheduling procedures using neuroevolution of augmenting topologies

The paper proposes a fast online learning method for neural network structures by using genetic algorithm (GA) and dynamic back propagation algorithm (BP) jointly. GA is used in the coarse tuning process which adjusts interconnection weights of the neural network. The dynamic back propagation algorithm is subsequently applied to achieve fine adjusting of the network weights. The fitness function, based on the squared error between the teaching signal and the network output value, is redefined at every time step and the proposed GA based algorithm solves a nonstationary function optimization task. At every time step the solution with the best fitness function is used for current representation of the neural network weights and biases. It is shown through the simulations and real time temperature control of drying oven that this learning algorithm has faster convergence ability and better performance on reducing mapping error in the online learning neural network structures. This leads to an improvement of the transient response of neuro adaptive systems. The proposed method has the potential to be applied to many practical areas such as system modeling and control, signal processing and pattern recognition.

Nowadays artificial intelligence (AI) has been applied to many areas, including digital game with various genres. One of them is the fighting game. Implementation of AI in the fighting game aims to make players who are not constrained by the limiting factors which humans possess, such as speed of reflexes and hand gestures. One platform that can be used to make fighting game AI is FightingICE. This platform is official platform used in Computational Intelligence and Games (CIG) conference. One approach that can be used in the process of building the AI in the fighting game is opponent modeling, a method that utilize machine learning to make model based on the behavior of the opponent. There is one AI that utilize k-Nearest Neighbor (kNN) algorithm called MizunoAI. This AI has outstanding performance, but it suffers from inability to adjust its parameter automatically. To solve that problem, in this paper the algorithm used for modelling is changed to Neuroevolution of Augmenting Topologies (NEAT). NEAT is an algorithm with Artificial Neural Network (ANN) as basic architecture and Genetic Algorithm (GA) as the architecture optimizer. In the implementation, NEAT will receive six inputs of distance in the x-axis and y-axis, both players' hitpoints, and both players' stamina. Output of NEAT is the probability of all movements that can be done by the opponent. The entire output will be processed by a simulator to determine the best countermeasure based on the predictions. For testing, the AI made will be faced against four AI winners of CIG 2013, namely MizunoAI, Kaiju, T, and SejongAI. The results showed that the AI is capable to do well in modeling process. Although the performance is still behind the MizunoAI, the learning process can be seen along the match.

Neuroevolution is a powerful and general technique for evolving the structure and weights of artificial neural networks. Though neuroevolutionary approaches such as NeuroEvolution of Augmenting Topologies (NEAT) have been successfully applied to various problems including classification, regression, and reinforcement learning problems, little work has explored application of these techniques to larger-scale multiclass classification problems. In this paper, NEAT is evaluated in several multiclass classification problems, and then extended via two ensemble approaches: One-vs-All and One-vs-One. These approaches decompose multiclass classification problems into a set of binary classification problems, in which each binary problem is solved by an instance of NEAT. These ensemble models exhibit reduced variance and increasingly superior accuracy as the number of classes increases. Additionally, higher accuracy is achieved early in training, even when artificially constrained for the sake of fair comparison with standard NEAT. However, because the approach can be trivially distributed, it can be applied quickly at large scale to solve real problems. In fact, these approaches are incorporated into DarwinTM, an enterprise automatic machine learning solution that also incorporates various other algorithmic enhancements to NEAT. The resulting complete system has proven robust to a wide variety of client datasets.

Accurate prediction of energy products future price is required for effective reduction of future price uncertainty as well as risk management. Neural Networks (NNs) are alternative to statistical and mathematical methods of predicting energy product prices. The daily prices of Propane (PPN), Kerosene Type Jet fuel (KTJF), Heating oil (HTO), New York Gasoline (NYGSL), and US Coast Gasoline (USCGSL) interrelated energy products are predicted. The energy products prices are found to be significantly correlated at 0.01 level (2-tailed). In this study, NNs learning algorithms are used to build a model for the accurate prediction of the five (5) energy product price. The aptitudes of the five (5) NNs learning algorithms in the prediction of PPN, KTJF, HTO, NYGSL, and USCGSL are examined and their performances are compared. The five (5) NNs learning algorithms are Gradient Decent with Adaptive learning rate backpropagation (GDANN), Bayesian Regularization (BRNN), Scale Conjugate Gradient backpropagation (SCGNN), Batch training with weight and bias learning rules (BNN), and Levenberg-Marquardt (LMNN). Results suggest that the LMNN and BRNN can be viewed as the best NNs learning algorithms in terms of R2 and MSE whereas GDANN was found to be the fastest. Results of the research can be use as a guide to reduce the high level of uncertainty about energy products prices, thereby provide a platform for developmental planning that can result in the improvement economic standard.

We estimate the spectrum envelope of speech by using neural networks. The neural networks system proposed in our study consists of two parts. The first part is the network for detecting the spectrum peaks, and the other is for interpolating between the extracted peaks. The structure of the latter is derived from the Neville interpolation. The former is based on the peak picking method. We present an operator for detecting the peaks using neural networks. Unfortunately, the result is that the neural networks do not only detect the general peaks but also the local peaks and the extremely closed peaks. We train the neural networks to detect only the general peaks. The learning algorithm uses a steepest descent method such as the backpropagation (BP) algorithm. After the learning, we show the detected peaks and the spectrum envelope when we use the learned pattern as the input data. In addition, we perform an experiment to examine whether the proposed operator corresponds to the XOR Boolean function or not.

The goal of this paper is to create a hybrid system based on a Multi-Agent Architecture that will investigate the evolution of some neural network methods along with technical and fundamental analysis methods on stock market indexes and how this information influences the stock market behavior in order to improve the profitability of a short or medium time period investment. The proposed system compares the results of Standard Feed Forward Neural Network, Elman and Jordan Recurrent Neural Networks and a Neural Network evolved with Neuro Evolution of Augmenting Topologies (NEAT) in order to investigate which network gives the most accurate result and time performance by taking in consideration the close price of a stock. The system also finds correlations between the pattern recognition methods and technical and fundamental methods results in order to find the direction of the market trend, to predict the next day price of a stock and to trigger a useful buy/sell signal. We are also interested in finding a correlation between the evolution of price, volume, number of transactions in order to have a better view on which is the effect of stock liquidity on a stock price. In order to validate our model a prototype was developed and applied to the Bucharest Stock Exchange Market indexes.

A way in which a neural network can implement a genetic algorithm as its learning algorithm is shown. This model is called GLANN (genetic learning algorithm for neural networks). The components of GLANN can be shown to be biologically plausible. The algorithm itself can be classified as a reinforcement learning algorithm. The neural network has a fixed architecture and processes binary strings using genetic operators. Learning is stored in the form of newly created patterns, which can then be stored in some kind of associative memory. The benefits of GLANN reside in the proven optimizing capabilities of genetic algorithms, and in its parallel implementation. The shallow two-level architecture translates into system scalability, an issue that has not been successfully resolved in the case of other neural network algorithms. >

Fractional calculus has been adopted in the modelling of many scientific processes and systems. Due to the inherent feature of long term memory of fractional derivatives, it has been used in the learning process of neural networks. A fractional order derivative based back propagation learning algorithm in neural networks is proposed in this paper. Specifically, Riemann-Liouville (R-L), Caputo (C) and Caputo Fabrizio (CF) fractional Derivative based on the back propagation algorithms in a three layer feed-forward neural network employed. To get a faster learning rate without oscillation, momentum factor is incorporated. The effect of fractional order and momentum factor is investigated and compared. The performance of these fractional derivatives based algorithms with integer derivatives based algorithm in terms of mean square error (MSE), particularly the salary based on years of experience is predicted. Results demonstrate that fractional derivative based learning algorithms outperform the integer derivatives.

The construction of a Boolean competitive neural network using ideas from immunology

We used multiobjective genetic algorithms with neuroevolution of augmenting topologies (NEAT) to evolve effective micro behaviors for opposing groups with heterogeneous compositions in StarCraft II, an RTS game. We used the Fast Nondominated Sorting Genetic Algorithm to maximize damage done and minimize damage received in a skirmish, and used this two objective fitness to guide NEAT to evolve the structure and weights of a neural network based controller. The evolved NEAT network controls the movement and attack commands for each unit. We show that non-dominated selection and NEAT can be used together to generate effective micro for groups with two types or three types of units on each side. The evolved micro also generalized well to random configurations, doing well along both objectives. We also manually co-evolved against the best performing individuals produced during a run for multiple cycles and show that this improves micro resulting in better performance against the default Starcraft II AI.