作者：
Yao, XLiu, YLaboratory of Chromatography
DEPg.Fac.Quimica Universidad Nacional Autonoma de Mexico Circuito interior Cd Universitaria/CP 04510 Mexico D.F.Mexico
This paper presents a new evolutionary system, i.e., EPNet, for evolving artificial neural networks (ANN's), The evolutionary algorithm used in EPNet is based on Fogel's evolutionary programming (EP), Unlike m...
详细信息
This paper presents a new evolutionary system, i.e., EPNet, for evolving artificial neural networks (ANN's), The evolutionary algorithm used in EPNet is based on Fogel's evolutionary programming (EP), Unlike most previous studies on evolving ANN's, this paper puts its emphasis on evolving ANN's behaviors, This is one of the primary reasons why EP is adopted, Five mutation operators proposed in EPNet reflect such an emphasis on evolving behaviors, Close behavioral links between parents and their offspring are maintained by various mutations, such as partial training and node splitting, EPNet evolves ANN's architectures and connection weights (including biases) simultaneously in order to reduce the noise in fitness evaluation, The parsimony of evolved ANN's is encouraged by preferring node/connection deletion to addition, EPNet has been tested on a number of benchmark problems in machine learning and ANN's, such as the parity problem, the medical diagnosis problems (breast cancer, diabetes, heart disease, and thyroid), the Australian credit card assessment problem, and the Mackey-Glass time series prediction problem, The experimental results show that EPNet can produce very compact ANN's with good generalization ability in comparison with other algorithms.
A hybrid learning algorithm designed for feedforward neural networks is proposed. Presented procedure combines the advantages of both global evolutionary programming search and local gradient tuning through cooperatio...
详细信息
A hybrid learning algorithm designed for feedforward neural networks is proposed. Presented procedure combines the advantages of both global evolutionary programming search and local gradient tuning through cooperation of hidden neurons. Such an approach allows to implement search in a single network that differs from traditionally employed evolutionary simulations.
A simulated battlefield, containing airborne lasers that shoot ballistic missiles down, provides an excellent test-bed for developing adaptive controllers. An airborne laser fire controller, which can adapt the strate...
详细信息
A simulated battlefield, containing airborne lasers that shoot ballistic missiles down, provides an excellent test-bed for developing adaptive controllers. An airborne laser fire controller, which can adapt the strategy it uses for target selection, is developed. The approach is to transform a knowledge-based controller into an adaptable connectionist representation, use supervised training to initialize the weights so that the adaptable controller mimics the knowledge-based controller, and then use directed search with simulation-based performance evaluation to continuously adapt the controller behavior to the dynamic environmental conditions. New knowledge can be directly extracted from the automatically discovered controllers. Three directed search methods are characterized for production training, and compared with the better characterized gradient descent methods commonly used for supervised training. Automated discovery of improved controllers is demonstrated, as is automated adaptation of controller behavior to changes in environmental conditions.
With the help of the family of non-linear projections and fitness functions introduced here, and using a standard evolutionary programming procedure, a broad class of parametric geometric primitives may be discovered ...
详细信息
With the help of the family of non-linear projections and fitness functions introduced here, and using a standard evolutionary programming procedure, a broad class of parametric geometric primitives may be discovered in an image. The generalization of the notion of Mahalanobis distance proposed here permits this approach to be extended to non-parametric primitives. The suggested approach is highly robust. (C) 1997 Elsevier Science B.V.
This paper compares the accuracy of parametric and nonparametric classifiers on the problem of predicting Bankruptcy. Among the single classifiers, an artificial neural network is found to provide the best results. Tw...
详细信息
This paper compares the accuracy of parametric and nonparametric classifiers on the problem of predicting Bankruptcy. Among the single classifiers, an artificial neural network is found to provide the best results. Two ways of combining classifiers are considered and an additive aggregation method is proposed. We show that both ways of combining produce classifiers whose forecasts are more accurate than the ones obtained with any single model. We suggest that an optimal system for risk rating should combine two or more different techniques.
It has been previously shown that evolving modular programs can improve the efficiency of induction for a specific class of modular programs called automatically defined functions (ADFs). ADFs, a variant of genetic pr...
详细信息
ISBN:
(纸本)0819425877
It has been previously shown that evolving modular programs can improve the efficiency of induction for a specific class of modular programs called automatically defined functions (ADFs). ADFs, a variant of genetic programming, induce hierarchically decomposed programs similar to those a human programmer might construct. This paper demonstrates that multiple interacting programs (MIPs), an evolutionary program that induces systems of equations, also provides an efficiency increase when multiple equations are evolved rather than a single equation. This result is demonstrated on two different boolean problems using both a numerical and boolean representation.
evolutionary programming was originally proposed in 1962 as an alternative method for generating machine intelligence. This paper reviews some of the early development of the method and focuses on three current avenue...
详细信息
evolutionary programming was originally proposed in 1962 as an alternative method for generating machine intelligence. This paper reviews some of the early development of the method and focuses on three current avenues of research: pattern discovery, system identification and automatic control. Recent efforts along these lines are described. In addition, the application of evolutionary algorithms to autonomous system design on parallel processing computers is briefly discussed.
This paper explores the applicability of clustering methods for obtaining an optimal partition of a network. In order to make the network management fault-tolerant, more than one management center is assigned to each ...
详细信息
This paper explores the applicability of clustering methods for obtaining an optimal partition of a network. In order to make the network management fault-tolerant, more than one management center is assigned to each cluster of nodes in the partition. Gradient descent partition methods converge to locally optimal partitions. In contrast, a stochastic search method called evolutionary programming is employed to search for a globally optimal partition that minimizes the communication cost.
In this paper, we describe TOGAPS, a Testability-Oriented Genetic Algorithm for Pipeline Synthesis. The input to TOGAPS is an unscheduled data flow graph along with a specification of the desired pipeline latency. TOG...
详细信息
In this paper, we describe TOGAPS, a Testability-Oriented Genetic Algorithm for Pipeline Synthesis. The input to TOGAPS is an unscheduled data flow graph along with a specification of the desired pipeline latency. TOGAPS generates a register-level description of a datapath which is near-optimal. in terms of area, meets the latency requirement, and is highly testable, Genetic search is employed to explore a 3-D search space, the three dimensions being the chip area, average latency, and the testability of the datapath. Testability of a design is evaluated by counting the number of self-loops in the structure graph of the data path. Each design is characterized by a four-tuple consisting of (i) the latency and schedule information, (ii) the module allocation, (iii) operation-to-module binding, and (iv) value-to-register binding. Accordingly, we maintain the population of designs in a hierarchical manner. The topmost level of this hierarchy consists of the latency and schedule information, which together characterize the timing performance of the design. The middle level of the hierarchy consists of a number of allocations for a given latency/schedule duplet. The lowest level of the hierarchy consists of a number of bindings for a specific latency/schedule/allocation. An initial population of designs is constructed from the given data flow graph using different latency cycles whose average latency is in the specified range. Multiple scheduling heuristics are used to generate schedules for the DFG. For each of the resulting scheduled data flow graphs, we decide on an allocation of modules and registers based on a lower bound estimated using the schedule and latency information. The operation-to-module binding and the value-to-register binding are then carried out. A fitness measure is evaluated for each of the resulting data paths;this fitness measure includes one component for each of the three search dimensions. Crossover and mutation operators are used to generate ne
In training a back-propagation neural network, the learning speed of the network is greatly affected by its learning rate. None, however, has offered a deterministic method for selecting the optimal learning rate. Som...
详细信息
In training a back-propagation neural network, the learning speed of the network is greatly affected by its learning rate. None, however, has offered a deterministic method for selecting the optimal learning rate. Some researchers have tried to find the sub-optimal learning rates using various techniques at each training step. This paper proposes a new method for selecting the sub-optimal learning rates by an evolutionary adaptation of learning rates for each layer at every training step. Simulation results show that the learning speed achieved by our method is superior to that of other adaptive selection methods.
暂无评论