Improving of artifical neural networks performance by using gpu's a survey

Sundarapandian et al. (Eds) : ACITY, AIAA, CNSA, DPPR, NeCoM, WeST, DMS, P2PTM, VLSI - 2013
pp. 39–48, 2013. © CS & IT-CSCP 2013 DOI : 10.5121/csit.2013.3405
IMPROVING OF ARTIFICIAL NEURAL
NETWORKS PERFORMANCE BY USING GPU’S:
A SURVEY
Israel Tabarez-Paz1
, Neil Hernández-Gress2
and Miguel González
Mendoza2
.
1
Universidad Autónoma del Estado de México
Blvd. Universitario s/n, Predio San Javier Atizapán de Zaragoza,
México {itabarezp}@uaemex.mx
http://www.uaem.mx/cuyuaps/vallemexico.html
2
Tecnológico de Monterrey, Campus Estado de México,
Carretera Lago de Guadalupe km 3.5 Atizapán de Zaragoza Col. Margarita
Maza de
Juarez, Atizapán de Zaragora, México
{ngress, mgonza}@itesm.mxhttp://www.itesm.edu
ABSTRACT
In this paper we study the improvement in the performance of Artificial Neural Networks (ANN)
by using parallel programming in GPU or FPGA architectures. It is well known that ANN can
be parallelized according to particular characteristics of the training algorithm. We discuss
both approaches: the software (GPU) and the Hardware (FPGA). Different training strategies
are discussed: the Perceptron training unit, the Support Vector Machines (SVM) and Spiking
Neural Networks (SNN). The different approaches are evaluated by the training speed and
performance. On the other hand, algorithms were coded by authors in the hardware, like Nvidia
card, FPGA or sequential circuits that depends on methodology used, to compare learning time
with between GPU and CPU. Also, the main applications were made for recognition pattern,
like acoustic speech, odor and clustering According to literature, GPU has a great advantage
compared to CPU, this in the learning time except when it implies rendering of images, despite
several architectures of Nvidia cards and CPU’s. Also, in the survey we introduce a brief
description of the types of ANN and its techniques of execution to be related with results of
researching.
KEYWORDS
GPU, FPGA, Artificial Neural Networks, Spiking Neural Networks, Support Vector Machines.
1. INTRODUCTION
This paper presents a survey of the improvement of training time and performance of Artificial
Neural Networks (ANN). Some trends of topics are about ANN is Parallel Programming to solve
problems such as clustering (Herrero-Lopez [8]), pattern recognition (Olaf [28]), regression
(Carpenter [19]), building of ANN in a specific hardware, such as FPGAs (Papadonikolakis [7]).
Historically, ANN were developed in second half of century XX as a result of research made in
Second War World. First generations are MC Culloc (1943), Hebb (1949), Perceptron (1957),

40 Computer Science & Information Technology (CS & IT)
Adaline y Madaline (1959), Cognitron (1980) y Hopfield (1982). Then, in 1982 Hinton and
William developed the known algorithm called back propagation. This can be considered as
second generation of ANN. The third generation was studied since 1930, with British scientists
Alan Lloyd Hodgkin (1914-1998) y Andrew Fielding Huxley (he was born in 1917), nowadays
SNN are based on this model. These authors finished their research in 1950 and their
mathematical model is known as Hodgkin – Huxley [27]
The perceptron is the first training unit in which a training algorithm for linearly separable
problems, that consist of a single neuron. Mathematically, itis represented by anstraight line
equation [36]. However, non – linear problems can’t be solved with this methodology.
However, in decade of 1990 was started the development of SVM [35], even this methodology
had been invented since 1979 [32] by Vapnik, to solve more complex problems, linearlly
separable or non – separable. In this case, support vectors are a set of vectors placed in border of
clusters for classifying or densities detection. An advantage related to this methodology was that
scientist calculated its architecture that consists of three layers; however, SVM consume many
memory resources.
In 2003 was developed the methodology SNN also called Spike Response Model (SRM) (Bohte
[29], Olaf[28]). At the same time, Izhikevich [30] developed a reduced model of Hodgkin –
Huxley model that consist on two differential equations that explain behavior of mammal
neurons. The main diference between SRM and Izhikevich’s model are the differential equations.
In case of Hodgkin – Huxley model has four differential equations with partial non linear
derivatives, and depends on the space and time. This model describes propagation and generation
of potential of a big axon of squid in order to explain the main properties. SNN’s are based on the
model described in last paragraph, because is the model most similar to the neurons of mammals
[28]
Information in the mammalian neurons of the brain, is coded with spikes, around 55 mV. A spike
is an electrical pulse along its axon. The similar Artificial Neural Network with mammalian
neuron is Spiking Neural (SNN), what sends a response according to the data encoded in the time,
so it is more suited for applications where the timing of input signals carries important
information (e.g., speech recognition and other signal-processing applications). Also, SNN can be
applied to the same problems that depend on behavior of time of parameters because of its
singular characteristic of coding in the time [13]
Respect to parallel programming [38] calculations are carried out simultaneously, operating on
the principle that large problems can often be divided into smaller ones, that are solved
concurrently. There are several different forms of parallel computing: bit-level, instruction level,
data, and task parallelism. This manuscript is focused on instruction level and task. In the figure
1, we can see principle of parallel computing.
Problems of parallel programming can be solved in OpenGL, Cg, C, C++ and Fortran.
Finally, they are developed for CUDA language.
This paper is distributed as follows: in section 2 the state of the art of ANN is presented, in
section 3 the performance of ANN in parallel programming is included, finally in section 4 are
the conclusions.

Computer Science & Information Technology (CS & IT) 41
Figure 1. Parallel programming.
2. RELATING WORKS ON THE USE OF PARALLEL PROGRAMMING IN
ARTIFICIAL NEURAL NETWORKS
Recently works are shown in the following paragraph. According to Sridhar [1], proposes
application of a technology that consists of a chip thermal modelling based on Neural Network
and GPU’s. Learning time is compared between CPU and GPU for this application. As a
conclusion, GPU is faster then CPU.
Table 1. State of the Art Parallel.
AUTHOR CONTRIBUTION YEAR
Arash [2] Implements Izhikevich’s model in a GPU of Spiking Neural
Network. Comparisons between CPU and GPU are shown and
considerable speedup is achievable with the approach depending
on the system architecture and the SNN complexity.
2011
Lowe [3] He uses several GPU’s of Nvidia to compare an Artificial
Neural Network with eight neurons in the hide layer about its
training time. As a result, Nvidia GTX is faster than other cards
that he used. However, he doesn’t say which is the model of
Artificial Neural Network.
2011
Sergio Herrero [8] Compares learning time and precision between two GPU targets
and libraries of MatLab.
2010
Izhikevich [9] Designs a hybrid model for SNN in order to combine continuous
and discontinuous numeric methods.
2010
Bhuiyan [11] Compares the models of SNN such as Izhikevich and Hodgkin
Huxley, these models applied to recognition of characters.
2010
Yudanov [13] Implements a hybrid method with numeric integration of Parker
Sochacki (PS) with adaptative order. This is validated at the
moment in the comparision made between GPU and CPU in
their characteristics.
2010
Scanzio [12] He compares the speed of processing in CUDA of algorithms
feed fordward and back propagation.
2010
Xin Jin [14] He Applies a chip called SpiNNaker that contains several
processors ARM968 that has a speed of processing of 200MHz
each one. He compares the results usingMatLab. He evaluate the
state of actualization of the neuron, arrangement of entry,
processing of a new entry;. However he don’t experiment with
other databases.
2010
Papadonikolakis
[15]
He has focused on improving the speed of learning and
efficiency of SVMs using several methods. Also, he compares
these parameters between a GPU and a FPGA programming
Gilbert’ algorithm.
2009
Nageswaran [17] He presents a compilation of theIzhikevich’s models. 2009

Thomas andLuk
[18]
The author proposes a Simulation with maximum 1024 neurons
in a FPGA the Izhikevich’s model.
2009
Stewart andBair
[20]
He Applies Runge – Kutta’s method for Izhikevich and Hodgkin
Huxle’s models. As a result, second method is more efficient.
2009
Prabhu [21] He applies GPU for pattern classifier in images. He focuses on
the degree of parallelism of a problem. He uses maximum size
of image of 256 MB, and in a video memory GPU of 768 MB.
As a result, author compares Dual – Core AMD processor with a
Geforce 6150 GPU, and when the number of patterns increase,
the CPU is linearly slower than GPU, But when the network
size increase the curve isn’t linear.
2008
Catanzaro [22] The author proposes theSVM using algorithm of Sequential
Minimal Optimization (SMO), also he compares time of
learning and precision of classification between GPU and
libraries of SVM of MatLab.
2008
Martínez [23] He implementes Fuzzy ART Neural Network in a GPU in order
to compare results with CPU processor. Type of data applied are
images in format RGB. Also the author uses dual – core Pentium
4 at 3.2 GHz, although the Neural Network is only accelerated
in a GeForce 7800 GT card. As a conclusion, CPU is faster in
training stage that CPU, in spite of GPU is faster in testing stage.
2007
Philipp [24] He programs Spiking Neural Network in a FPGA. About results,
author says that this hardware can be simulated through
thousands of neurons, however author does not show
conclusions about it. On the other side, he focuses on
synchronization of nodes according to the frequency of the
signal.
2007
Pavlidis, et. al [26] He Applies evolution algorithms using SNN, however according
to author efficiency of the network was not calculated.
2005
Zhongwen [25] He calculates the learning time in a multilayer perceptron in a
CPU and a GPU, spending 11328 ms (CPU) and 46 ms (GPU)
respectively.
2005
Olaf [28] The author Shows theory about SNN and the problem of
codification of the entries.
2004
Bohte [29] He also made a comparison with different algorithms as
Spikeprop. The work does not show the learning time. On the
other hand, he presents differences between traditional ANN and
SNN.
2003
Izhikevich [30] He simulates in MatLab the Hodgkin – Huxley’s model. Also,
he gots to execute maximum 10000 neurons and 1000000 of
sinapsys.
2003
3. PARALLEL PROGRAMMING IN ANNS
3.1. Types of architectures for parallelization
Artificial Neural Networks can be parallelized in a specific hardware, for example GPU, FPGA
such a Thomas [18], sequential circuits, or specific card [14] However, according to the most
often used, we have focused on GPU and FPGA.
In the case of SVM’s, they are better designed in a GPUbecause the optimizing methodrequires of
solving repetitively operations of matrixes. However, for SNN methodology is better to use

FPGA or sequential circuits, although it can be simulated in GPU. The reason is that in a GPU
thelearning time is calculated with a non linear and exponential equation,which algebraic order
depends directly of number of neurons in the input. This equation is solved with mathematical
approximations, so if the number of neurons in a input layer is increasing, the computational
complexity too. However, to design SNN architecture in the circuit mentioned, the time does not
need a mathematical solution to find the threshold, this is only detected with a simplex circuit.
3.2. Configuration of a GPU
This section is focused in program CUDA in a GPU. Because of each author uses his own
resources, then there is not a standardhardware of interest for comparing the results with other
researchers, however results of other investigations can be used as a reference to improve the
algorithms.
One of the biggest problems to start isto configure the hardware selected. Some ANNs need to be
programmed with kernel in three dimensions, such as SNN. The simplest configuration in one or
two dimensions is perceptron method. Parameters to be considered in hardware are threads,
blocks and grids. In case of perceptron, each column of a block can be used as a layer;in contrast
for SNN each block can represent only one neuron. Sometimes computational resources are not
enough. Other problem to be considerated is thatthis method is recursive and weightsmust be
frequently adjusted because of local memory is not enough.
However, all details of programming depend on the ability to program each design it can be easy
or not. Also, efficiency can be better if the researcher knowswhen a specific methodology could
be used. But the most important depends on the researcher’s ability to decide when and how
much can be parallelized (the grade of parallelism).
The hardware configuration is selected in each kernel, the number of threads and the number of
blocks per grid are chosen [38]
3.3. Grade of parallelization
All ANNs can be parallelized in several levels. First, perceptron can be parallelized per layer. In
other words, all outputs of neuronsper layer can be calculated at the same time because of
thesimple form of its activation function. On the other hand, each layer must be calculated
sequentially because of input of hidden layer depends on output of previous layer. However,
circular buffer, this procedure can be designed, through FIFO principle (First in – First out).So,
when all the output values of the first iteration are known all the layers could be parallelized.
However, this can be implemented in other research.
Second, in case of SVM, is similar to perceptron but the difficulty increases when the quantity of
parameters increases to calculate the dimension in where the problem is separable. So, a handle
matrix of more than one columnis needed and requirements of computational resources are
bigger. In this case, we kernels in CUDA with more than one dimension could be needed.
Third, for SNN an implement algorithmof a mathematical or sequential method to calculate the
value of threshold in amplitude and time is needed. This impliesthatmany values of time are
needed, as well as, maybe hundreds or thousands per row in a block. So, a neuron is represented
as a grid in tree dimensions, where each row can represent a previous neuron to be added for one
output of the following neuron. This method requiresgood memory resources.
Therefore, sometimes it is not recommended to parallel more than hardware features allow you to
do.

3.4. Examples of parallelization in ANNs
Parallelization of ANN consists in two phases: first learning phase, and second execution phase.
Parallelization for learning phase in case of SVM is in parallelization of matrixes operations,
what is well defined in webpage of CUDA [37] although this depends on optimization method
applied what is focused on solving quadratic problem [38]. In case of SNN, the mathematical
calculus in phase of learning is more sequential because of its approximation method.But in both
SVM and SNN all weight of hidden layers neurons can be parallelized and calculated at the same
time.
The second phase refers about to evaluations of weights calculated according to inputs. The
sections A, B and C show a way of parallelizing of this phase.
Preceptron.Graphically, the following components of the model represent the actual activity of a
neuron. All inputs are summed altogether and modified by the weights. This activity is referred as
a linear combination. Finally, an activation function controls the amplitude of the output. This
process is described in the figure 2. Each thread represents a layer. In the GPU, parallelism for a
neuron of perceptron is focused on mathematical operations [38]
Figure 2. Simple perceptron.
Backpropagation.In this case has the samealgorithm in weight thatSNN. The weight needs to be
adjusted per iteration. Parallelization in two dimensionsis like the figure 3.
Support Vector Machines (SVM).In the figure 3, the parallelization of SVM could be observed,
in this case each thread represents a layer with n or k number of rows, however the mathematics
calculus are parallelized into of each neuron of hidden layer in other kernel, what is solved as a
optimization problem. The dimension of the Hessian matrix isequal to numberof input
parameters. So, the multiplication of matrix is another operation to parallelize that can be solved
in a separated kernel. In the figure 3, the y0 neuron gets at the same time all values multiplied per
its respective weight, the in other kernel in CUDA the mathematical operations are parallelized.
At the same time the other neurons of hidden layer computing its respective output. However, the
next layer cannot calculate its output without the previous layer has done.
Figure 3. Parallelization of Support Vector Machine (SVM), 2D Array.

Spiking Neural Networks (SNN). Thearrange of figure 4 represents a configuration of the GPU
device in three dimensions. This is a solution for parallelizing SNNalgorithm.The cube showedin
this figure is only a neuron of a hidden or output layer.There are as many cubes as neurons are
required. Each cube is divided in blocks, what depend on the length of time in the input [28]. All
neurons per layer can be calculated in parallel, but a disadvantage is thatthis procedure requires
many resources of memory.
Figure 4. Parallelization of Spiking Neural Network (3D Processor Array).
SNN has significant characteristics that must be considerated. The synapses of the biological
neuron are modeled as weights. Let’s remember that the synapse of the biological neuron is
which interconnects the neural network and gives the strength of the connection. For an artificial
neuron, the weight is a number, and represents the synapse. A negative weight reflects an
inhibitory connection, while positive values designate excitatory connections.Inherent parallelism
of commodity graphic hardware is used to accelerate the computationofANN.According to
Nikola [34], taxonomy of parallelization approaches for neurosimulations is represented in the
figure 5.
Figure 5. Taxonomy of parallelization approaches for neurosimulations.
Sridhar [1] says that the main advantage of GPU over CPU is high computational parallelism and
efficiency with a relatively low cost.However, it is difficult to design an algorithm. Also, the
author says that although exist Integrated Circuits (IC) for high parallelism, it is very difficult to
translate this parallelism in an efficient software.On the other hand, human brain can be trained to
solve complex problems, such as thermal modeling of specific IC layouts.
Prabhu [21] compares efficiency of the human brain with enormous computational powerand
parallel environs of GPU’s, so we understand that GPU has some limitations. According to him,
the role played by Graphical Processing Unit (GPU) is approaching to Artificial Neural Networks
to the nature of human brain. Also, GPU’s have been used for rendering high quality images in
real time, virtual reality simulations and games. Modern GPU’s can perform highly intensive
parallel tasks.
4. CONCLUSIONS
In this paper weconclude thatparallelismin ANNincrease speedof learning time. However, is very
difficult to design this sort of algorithms. On the other hand, we can parallelize by hardware
(FPGA) of software (GPU). Tendency is study to know which algorithm is the most efficient and

faster, because of their mathematical characteristics and their architecture. So, is better to solve a
problem with large database using SVM and SNN than traditional ANN.
The importance to compare the efficiency between these algorithms is to know the error in the
results and which is faster for learning according to quantity of instances and parameters per
instance. So, with this information is possible to know what applications are the most
appropriates for each application.
As a future work, there are some aspects, as parallelizing SVM or SNN in a GPU and SNN in a
FPGA, then compare learning time. However, also is necessary to propone an important
application to solve real problems.
REFERENCES
[1] Sridhar, A.; Vincenzi, A.; Ruggiero, M.; Atienza, D.; , "Neural Network-Based Thermal Simulation
of Integrated Circuits on GPUs," Computer-Aided Design of Integrated Circuits and Systems, IEEE
Transactions on , vol.31, no.1, pp.23-36, Jan. 2012
[2] Ahmadi, Arash; Soleimani, Hamid; , "A GPU based simulation of multilayer spiking neural
networks," Electrical Engineering (ICEE), 2011 19th Iranian Conference on , vol., no., pp.1-5, 17-19
May 2011
[3] Lowe, E.W.; Woetzel, N.; Meiler, J.; , "Poster: GPU-accelerated artificial neural network for QSAR
modeling," Computational Advances in Bio and Medical Sciences (ICCABS), 2011 IEEE 1st
International Conference on , vol., no., pp.254, 3-5 Feb. 2011
[4] B. Kirk, David; W. Hwu, Wen-mai “Programming Massively Parallel Processors” Ed. Morgan
Kaufmann, 2010.
[5] Qi Li; Salman, R.; Kecman, V.; , "An intelligent system for accelerating parallel SVM classification
problems on large datasets using GPU," Intelligent Systems Design and Applications (ISDA), 2010
10th International Conference on , vol., no., pp.1131-1135, Nov. 29 2010-Dec. 1 2010
[6] Tsung-Kai Lin; Shao-Yi Chien; , "Support Vector Machines on GPU with Sparse Matrix Format,"
Machine Learning and Applications (ICMLA), 2010 Ninth International Conference on , vol., no.,
pp.313-318, 12-14 Dec. 2010
[7] Papadonikolakis, M.; Bouganis, C.: "A novel FPGA-based SVM classifier," Field-Programmable
Technology (FPT), 2010 International Conference on , vol., no., pp.283-286, 8-10 Dec. 2010
[8] Sergio Herrero-Lopez, John R. Williams, Abel Sanchez,: Parallel Multiclass Classification using
SVMs on GPUs, November 2010.
[9] Izhikevich, E.M.: "Hybrid spiking models", vol. 368, issue 1930, pp. 5061-5070, Nov 2010
[10] Venkittaraman Vivek, Pallipuram Krishnamani: ACCELERATION OF SPIKING NEURAL
NETWORKS ON SINGLE- GPU AND MULTI-GPU SYSTEMS, ProQuest document ID:
204037751, Publication Number: AAT 147555, May 2010.
[11] Bhuiyan, M.A.; Pallipuram, V.K.; Smith, M.C.: "Acceleration of spiking neural networks in emerging
multi-core and GPU architectures," Parallel & Distributed Processing, Workshops and Phd Forum
(IPDPSW), 2010 IEEE International Symposium on , vol., no., pp.1-8, 19-23 April 2010
[12] Scanzio, S.; Cumani, S.; Gemello, R.; Mana, F.; Laface, P.; , "Parallel implementation of artificial
neural network training," Acoustics Speech and Signal Processing (ICASSP), 2010 IEEE
International Conference on , vol., no., pp.4902-4905, 14-19 March 2010.
[13] Yudanov, D.; Shaaban, M.; Melton, R.; Reznik, L.; GPU-Based Simulation of Spiking Neural
Networks with Real-Time Performance & High Accuracy, Feb 2010.
[14] XIN JIN: Parallel Simulation of Neural Networks on Spinnaker Universal Neuromorphic Hardware,
University of Manchester, 2010
[15] Papadonikolakis, M.; Bouganis, C.-S.; Constantinides, G.: "Performance comparison of GPU and
FPGA architectures for the SVM training problem," Field-Programmable Technology, 2009. FPT
2009. International Conferenceon , vol., no., pp.388-391, 9-11 Dec. 2009
[16] Fidjeland, A.K.; Roesch, E.B.; Shanahan, M.P.; Luk, W.; , "NeMo: A Platform for Neural Modelling
of Spiking Neurons Using GPUs," Application-specific Systems, Architectures and Processors, 2009.
ASAP 2009. 20th IEEE International Conference on , vol., no., pp.137-144, 7-9 July 2009

[17] Nageswaran, J.M., Dutt, N.: Krichmar, J.L.; Nicolau, A.; Veidenbaum, A.; , "Efficient simulation of
large-scale Spiking Neural Networks using CUDA graphics processors," Neural Networks, 2009.
IJCNN 2009. International JointConferenceon , vol., no., pp.2145-2152, 14-19 June 2009
[18] Thomas, D.B.;Luk, W.: "FPGA Accelerated Simulation of Biologically Plausible Spiking Neural
Networks," Field Programmable Custom Computing Machines, 2009. FCCM '09. 17th IEEE
Symposium on , vol., no., pp.45-52, 5-7 April 2009
[19] Carpenter, A.; "CUSVM: A CUDA IMPLEMENTATION OF SUPPORT VECTOR
CLASSIFICATION AND REGRESSION", Jan. 2009.
[20] Stewart R.D., Bair W.: "Spiking neural network simulation: numerical integration with the Parker-
Sochacki method", Jan. 2009
[21] Prabhu, R.D.; "SOMGPU: An unsupervised pattern classifier on Graphical Processing Unit,"
Evolutionary Computation, 2008. CEC 2008. (IEEE World Congress on Computational Intelligence).
IEEE Congress on , vol., no., pp.1011-1018, 1-6 June 2008
[22] Bryan Catanzaro, Narayanan Sundaram, KurtKeutzer: Fast Support Vector Machine Training and
Classification on Graphics Processors, International Conference on Machine Learning, Helsinki,
Finland, 2008.
[23] Martínez Z. M., Díaz P. F. J., Díez H. J. F, Antón R. M.: Fuzzy ART Neural Network Parallel
Computing on the GPU, Springer (2007).
[24] Philipp S., Grübl A., Meier K., Schemmel J.: Interconnecting VLSI Spiking Neural Networks Using
Isochronous Connections, Springer (2007).
[25] L. Zhongwen, L. Hongzhi and W. Xincai: Artificial Neural Network Computation on Graphic Process
Unit (2005).
[26] N.G. Pavlidis, D.K. Tasoulis, V.P. Plagianakos, G. Nikiforidis, M.N. Vrahatis: Spiking Neural
Network Training Using Evolutionary Algorithms, IEEE International Joint Conference on, pp: 2190
– 2194, vol. 4 ,december 2005.
[27] SeijasFossi C., Caralli D’ Ambrosio A.: Uso de las máquinas de soportepara la estimación del
potencial de acción cellular, Revista de Ingeniería UC. Vol. 11, Nº 1, 56 – 61 (2004).
[28] Olaf Booij,: Temporal Pattern Classification using Spiking Neural Networks, Intelligent Sensory
Information Systems Informatics Institute Faculty of Science Universiteit van Amsterdam (2004).
[29] Sander M. B.: Spiking Neural Networks, Universiteit Leiden (2003).
[30] Izhikevich, E.M.: "Simple model of spiking neurons," Neural Networks, IEEE Transactions on ,
vol.14, no.6, pp. 1569- 1572, Nov. 2003
[31] Sander M. Bohte, Joost N. Kok, Han La Poutre: Error-backpropagation in temporally encoded
networks of spiking neurons, Neurocomputing 48, pp: 17 – 37, (2002).
[32] Platt C. J.: Sequiential Minimal Optimization: A fast Algorithm for Tarining Support Vector
Machines (1998).
[33] Osuna E., Freund R., Girosi F.: An Improved Training Algorithm for Support Vector Machines, In
Proc. of IEEE NNSP'97.
[34] Nikola B. Serbediija: Simulating Artificial Neural Networks on Parallel Architectures (1996).
[35] Vapnik V., Cortes C., Support Vector Networks (1995).
[36] FausetLaurene, Fundamentals of Neural Networks, ARCHITECTURE, ALGORITHMS, AND
APPLICATIONS, Prentice Halls, 1994.
[37] A. L. Hodgkin and A. F. Huxley: A quantitative description of membrane current and its application
to conduction and excitation in nerve, J. Physiol. (1952).
[38] http://www.nvidia.com
[39] http://www.svms.org

Authors
M en C Israel Tabarez Paz: He works for Universidad Autónoma del Estado de Méxicoas a researcher. He
is focus on Artificial Intelligence, Artificial Neural Networks. Also he is a PhD. Student in Computing
Science in InstitutoTecnológico y de EstudiosSuperiores de Monterrey Campus Estado de México
(ITESM).
PhD. Neil Hernández Gress:He works for de InstitutoTecnológico y de EstudiosSuperiores de Monterrey
Campus Estado de México (ITESM)as a Director of Research Technological Development and Postgrade
in ITESM. He is focus on Artificial Intelligence, Artificial Neural Networks.
PhD. Miguel González Mendoza.Head of the P Program in Computer Sciences and Engineering in
ITESM.Secretary in Mexican Society on Artificial Intelligence.

Improving of artifical neural networks performance by using gpu's a survey

More Related Content

What's hot (17)

Viewers also liked (8)

Similar to Improving of artifical neural networks performance by using gpu's a survey (20)

Recently uploaded (20)

Improving of artifical neural networks performance by using gpu's a survey