2017 IEEE Symposium Series on Computational Intelligence (SSCI)

Conference Information
Name: 2017 IEEE Symposium Series on Computational Intelligence (SSCI)
Location: Honolulu, United States
Date: 2017-11-27 - 2017-12-1

Latest articles from this conference

Hongwei Mo, He Qu, Lifang Xu, Chaomin Luo, Lu Ding, Qirong Tang
Abstract:
In order to overcome the limitation of the traditional bridge detection method, under the background of the unmanned aerial vehicle detection bridge, taking into account the signal without GPS, the information acquisition process and the grid method are combined to establish an environmental model, extremely covered the bottom of the bridge. In the path planning, most of the algorithms can only plan the shortest path, and the algorithm is complex, requires a lot of storage space, and it is difficult to carry out the whole area coverage path planning. Now the unit decomposition method and the two-point search strategy are proposed and applied in this context, so the algorithm has the advantages of simple computation and covering intelligence, It has great potential in the coverage path planning. Simulation experiments show that it can achieve the whole area coverage under the conditions of low power consumption.
Adrian Horzyk, Janusz A. Starzyk
Abstract:
This paper presents a fast self-organization of neural network structure using a new simplified pulsing model of neurons. These neurons incorporate the concept of time while simplifying many functional aspects of spiking models. This model is attractive because it is computationally very efficient. It allows for fast association of experimental data using conditional plasticity rules built-in neurons. It can be used for the representation of sequential and non-sequential data in neural network architectures. It also allows for the creation of synaptic connections that represent similarity, sequence, proximity, or defining dependencies between data and objects. Thus, this model can be used to develop complex neural graph structures for knowledge representation and retrieval. Such neural structures can be further used for fast search of related data or objects, clustering, classification, recognition, data mining, knowledge exploration, data retrieval, as well as for various cognitive tasks.
Jean Meunier, Dominique Knittel
Abstract:
Photovoltaic panels are a way of providing electricity without altering natural resources. Due to its intermittent nature and to meet consumers demand, this energy has to be stored, for example by using batteries. In this work, an energy model for buildings with battery is developed based on electrical consumption and production data. This model takes into account the depth of discharge, state of charge and efficiency over a cycle of a lithium type battery. Three rule-based strategies are then described. This leads to our optimization problem. The optimization is applied on two time slots: one day (for different algorithm strategies) and one week. Robust multi-objective optimization is performed in order to reduce the impact of consumption prediction errors.
Cristian D. Rodriguez Rodriguez, Diego Mayorga Gomez, Miguel A. Melgarejo Rey
Abstract:
This paper presents a method to forecast spatiotemporal patterns of criminal activity, through a novel time series approach from fuzzy clustering, in the city of San Francisco, USA. The developed analysis comprises from the cluster quantity selection to the forecast. A memetic algorithm is proposed in order to execute the series forecast, as well as, a problem-oriented fitness function. Results show that series approach of fuzzy clustering for criminal patterns is a feasible method to produce a forecast of criminal patterns.
Wen-Chi Yang
Abstract:
The selfish herd hypothesis highlights the importance of individual short-term fitness to the collective behavior. Previous agent-based models have demonstrated how selfish prey agents evolve into cohesive groups where individuals attempt to enter the central positions. However, these simulations either treated an agent as a point or allowed overlaps between agent bodies. Hence, the condition when a herd is too crowded to enter has long been neglected. In this paper, an agent-based model is built to simulate the behavioral evolution of a prey population in two-dimensional open space. These prey agents are specifically assigned rigid bodies so that overlapping is forbidden in the model. By introducing a genetic algorithm that evolves neural networks with incremental complexity, adaptive strategies can be developed automatically in evolution. The simulation output stresses the significant impact of the overlap-free condition on the behavioral evolution of gregarious prey. It is shown that given agents able to squeeze into a group by pushing away others, evolution will drive selfish prey agents to leave smaller heaps and assemble larger ones. In contrast, given that agents cannot squeeze into the crowd, selfish prey agents will evolve to exhibit various appearances of coordinated movement. This collective motion is due to malignant competition, which decreases the group benefit compared with the transitional states. These findings reveal a novel perspective on the collective behavior of group-living animals in nature.
, Benkai Li,
Abstract:
This paper concerns with a novel generalized policy iteration (GPI) algorithm with approximation errors. Approximation errors are explicitly considered in the GPI algorithm. The properties of the stable GPI algorithm with approximation errors are analyzed. The convergence of the developed algorithm is established to show that the iterative value function is convergent to a finite neighborhood of the optimal performance index function. Finally, numerical examples and comparisons are presented.
, Christoph Doell, Matthias Hewelt, Rudolf Kruse
Abstract:
This work builds up on previous research by Sievers and Helmert, who developed an Monte Carlo Tree Search based doppelkopf agent. This four player card game features a larger state space than skat due to the unknown cards of the contestants. Additionally, players face the unique problem of not knowing their teammates at the start of the game. Figuring out the player parties is a key feature of this card game and demands differing play styles depending on the current knowledge of the game state. In this work we enhance the Monte Carlo Tree Search agent created by Sievers and Helmert with a decision heuristic. Our goal is to improve the quality of playouts, by suggesting high quality moves and predicting enemy moves based on a neural network classifier. This classifier is trained on an extensive history of expert player moves recorded during official doppelkopf tournaments. Different network architectures are discussed and evaluated based on their prediction accuracy. The best performing network was tested in a direct comparison with the previous Monte Carlo Tree Search agent by Sievers and Helmert. We show that high quality predictions increase the quality of playouts. Overall, our simulations show that adding the decision heuristic increased the strength of play under comparable computational effort.
Weitai Hu, Jingling Li, Hong Huo, Tao Fang
Abstract:
Color selectivity and color constancy are important properties of human visual system, enabling human not only to distinguish different colors but also to perceive objects' real color invariant of the colorful illumination on them. In order to get a robust and biomimetic color image encoding method for color-biased images, we propose a spiking neural network (SNN) to model how the color selectivity and color constancy appear in human visual cortex. The hierarchical structure of the our SNN is consistent with human visual pathway from retina to secondary visual cortex(V2). The feed-forward connections are structured simulating the single opponent and double opponent receptive fields in cortex, and are simulated using excitatory and inhibitory synaptic connections. Lateral connections in cortex is also employed. Unsupervised learning rule: Spike-Timing-Dependent-Plasticity (STDP) is applied during the network training process under stimuli of natural images. After training, neurons response discriminatively to different color stimuli and the hue map is drawn to show preferred color of every neuron. And the hue map of our network highly ensembles biologically experiment result. Finally color-preferring neurons are used to encode color images in several methods. And classification tests are done using the commonly used SFU Lab dataset, showing the encoding methods are robust to color-biased situations.
Xiaoya Liao, Rui Zhang, Raymond Chiong
Abstract:
A bi-objective single machine scheduling problem with energy consumption constraints is studied, in which the objective functions are the total weighted completion time and the total weighted tardiness. Given the NP-hard nature of the problem, a multi-objective particle swarm optimization (MOPSO) algorithm is adopted to solve the problem. Since the original version of the MOPSO was designed for continuous optimization problems, it is crucial to decode its results in order to obtain feasible schedules. After the algorithm framework is determined, key parameters of the MOPSO are analyzed. A design of experiments (DOE) approach based on the Taguchi method is used to optimize parameters of the MOPSO algorithm for both small-scale and large-scale problem instances. To assess the algorithm's performance, we compare it to a well-known multi-objective evolutionary algorithm, the NSGA-II. DOE analysis is also carried out for tuning the parameters of the NSGA-II. Comprehensive computational experiments with different performance measures confirm that the modified MOPSO performs well on both small-scale and large-scale instances tested, and its performance is often superior compared to the NSGA-II.
Simon M. Lucas, , Diego Perez-Liebana
Abstract:
The compact genetic algorithm is an Estimation of Distribution Algorithm for binary optimisation problems. Unlike the standard Genetic Algorithm, no cross-over or mutation is involved. Instead, the compact Genetic Algorithm uses a virtual population represented as a probability distribution over the set of binary strings. At each optimisation iteration, exactly two individuals are generated by sampling from the distribution, and compared exactly once to determine a winner and a loser. The probability distribution is then adjusted to increase the likelihood of generating individuals similar to the winner. This paper introduces two straightforward variations of the compact Genetic Algorithm, each of which leads to a significant improvement in performance. The main idea is to make better use of each fitness evaluation, by ensuring that each evaluated individual is used in multiple win/loss comparisons. The first variation is to sample n > 2 individuals at each iteration to make n(n - l)/2 comparisons. The second variation only samples one individual at each iteration but keeps a sliding history window of previous individuals to compare with. We evaluate the methods on two noisy test problems and show that in each case they significantly outperform the compact Genetic Algorithm, while maintaining the simplicity of the algorithm.
Back to Top Top