Soft Computing Applications

advertisement
Soft Computing Applications
Pavel Krömer
IT4 Knowledge Management
Outline
Soft computing
Research topics and applications
Soft computing
Roughly: search for approximate solutions to complex
problems
Goals: practical solutions to real world problems, analysis of
large data sets, classification, prediction
Means: bio-inspired algorithms, fuzzy systems, artificial neural
networks, algebraic methods (matrix factorization),
Emphasis of HPC aspects: parallel algorithms, many-core
implementations, large data
IT4KM portfolio in SC
Evolutionary computation: genetic algorithms, genetic
programming, artificial immune systems, differential evolution
Swarm intelligence: particle swarm optimization, ant colony
optimization
Artificial neural networks: self organizing maps, PCA
performing ANNs, maximum likelihood Hebbian learning,
flexible neural trees
Hybrid methods: ANFIS
Research topics and applications
Design and implementation
of EAs on the GPU
Manycore (many-threaded) SIMD versions of
• DE, GA, GP
Design and implementation
of EAs on the GPU
P. Gajdos, P. Moravec: Intruder Data Classification Using GM-SOM, in
CISIM (A. Cortesi, N. Chaki, K. Saeed, and S. T. Wierzchon, eds.),
vol. 7564 of LNCS, pp 92-100, Springer 2012
Combinatorial optimization
Data classification
P. Krömer, J. Platos, and V. Snásel, “Evolutionary clustering on CUDA,”
in ECAI 2012 (L. D. Raedt, C. Bessière, D. Dubois, P. Doherty,
P. Frasconi, F. Heintz, and P. J. F. Lucas, eds.), vol. 242 of Frontiers in
Artificial Intelligence and Applications, pp. 909–910, IOS Press, 2012.
GPU Accelerated Genetic Clustering, SEAL 2012 (LNCS), Accepted.
Evolutionary clustering
Many-threaded Differential Evolution on the GPU (book chapter),
Evolutionary Computation on Graphics Processing Units, Springer, the
Natural Computing Series. Accepted.
+ submitted/invited papers to journal special issue
(Concurrency and Computation: Practice and
Experience, Wiley)
Information security, compression,
(bio-)signal analysis
Intrusion detection, cryptography, data compression,
EEG signal analysis
E. Ochodková, J. Dvorský, P. Krömer, and P. Tuček, “On fitness function based upon quasigroups power sequences,” in International Joint
Conference CISIS12-ICEUTE12-SOCO12 Special Sessions, vol. 189 of Advances in Intelligent Systems and Computing, pp. 141–150, Springer
Berlin Heidelberg, 2013. 10.1007/978-3-642-33018-6_14.E. 2012.
J. Platos and P. Kromer, “Improving evolved alphabet using tabu set,” in Hybrid Artificial Intelligent Systems (E. Corchado, V. Snášel,
A. Abraham, M. Wozniak, M. Graňa, and S.-B. Cho, eds.), vol. 7208 of Lecture Notes in Computer Science, pp. 655–666, Springer Berlin /
Heidelberg, 2012. 10.1007/978-3-642-28942-2_59. 2012.
P. Dohnálek, P. Gajdoš, T. Peterek and M. Penhaker, “Pattern Recognition in EEG Cognitive Signals Accelerated by GPU,” in International Joint
Conference CISIS12-ICEUTE12-SOCO12 Special Sessions, vol. 189 of Advances in Intelligent Systems and Computing, pp. 477-485, Springer
Berlin Heidelberg, 2013. 10.1007/978-3-642-33018-6_14.E. 2012.
Bio-inspired methods for
combinatorial optimization
Focus on emerging / less used
algorithms
• Motivation the „No free
lunch“ theorem
(GAs), DE, AIS
• New methods,
encodings
Problems
• Linear Ordering Problem
• Independent Task Scheduling
P. Kromer, V. Snasel, J. Platos, A. Abraham, and
H. Ezakian, “Evolving schedules of independent
tasks by differential evolution,” in INCoSA, vol. 329
of Studies in Computational Intelligence,
pp. 79–94, Springer Berlin / Heidelberg, 2011.
Kromer, Platos, Snasel, „Independent Task
Scheduling by Artificial Immune Systems“,
Differential Evolution, and Genetic Algorithms,
INCoS 2012, IEEE, pp 28-32. 2012.
Practical Results of Artificial Immune Systems for
Combinatorial Optimization Problems, NaBIC 2012
(IEEE), Accepted.
+ submitted/invited papers to journal special issue
(Cluster Computing: The Journal of Networks,
Software Tools and Applications, Springer)
ANNs on the GPUs
Manycore (many-threaded) SIMD
versions of
• SOM, Neural PCA, MLHL
P. Gajdoš and J. Platoš: GPU Based Parallelism for SelfOrganizing Map, IHCI 2011, Advances in Intelligent Systems
and Computing, Springer 2013, Volume 179, Part 4, 231-242
P. Krömer, E. Corchado, V. Snášel, J. Platoš, and L. GarcíaHernández, “Neural PCA and maximum likelihood hebbian
learning on the GPU,” in Artificial Neural Networks and
Machine Learning – ICANN 2012 (A. E. Villa, W. Duch, P. Érdi,
F. Masulli, and G. Palm, eds.), vol. 7553 of Lecture Notes in
Computer Science, pp. 132–139, Springer, 2012.
+ submitted/invited papers to journal special issues
(Neurocomputing)
Evolution of fuzzy predictors and
classifiers for data mining
Application of fuzzy IR principles
and GP in data mining
• An evolution of query
optimization algorithms
• Used for classification,
function approximation,
time series analysis
Evolution of fuzzy predictors and
classifiers for data mining
Steel products quality
estimation
Photovoltaic power plant
output prediction
Reactor tension cycles
estimation
Traffic accident severity
estimation
Intrusion detection
P. Krömer, J. Platoš, V. Snášel, and A. Abraham, “Fuzzy classification by
evolutionary algorithms,” in IEEE SMC 2011, pp. 313 – 318, 2011.
T. Beshah, D. Ejigu, A. Abraham, V. Snášel, and P. Krömer, “Knowledge
discovery from road traffic accident data in ethiopia: Data quality,
ensembling and trend analysis for improving road safety,” Neural
Network World, vol. 22, no. 3, pp. 215–244, 2012.
P. Krömer, T. Novosad, V. Snásel, V. Vera, B. Hernando, L. GarcaHernandez, Hé. Quintian-Pardo, E. Corchado, R. Redondo, J. Sedano and
A. E. Garcia, " Prediction of Dental Milling Time-Error by Flexible Neural
Trees and Fuzzy Rules " in H. Yin, J. A. F. Costa and G. D. A. Barreto (Eds.),
IDEAL 2012, LNCS, vol. 7435, pp. 842-849, Springer, 2012
P. Kromer, L. Prokop, V. Snasel, S. Misak, J. Platos, and A. Abraham,
“Evolutionary prediction of photovoltaic power plant energy production,”
in GreenGEC@GECCO 2012, GECCO Companion ’12, (New York, NY, USA),
pp. 35–42, ACM, 2012.
+ submitted/invited papers to journal special issues (Logic Journal of the
IGPL)
(Social) network analysis
Network analysis
Forcoa.net
DBLP analysis
Community detection
Traffic routing
V. Snásel, P. Krömer, J. Platos, M. Kudelka, and Z. Horak, “On spectral partitioning of co-authorship networks,” in CISIM (A. Cortesi, N. Chaki,
K. Saeed, and S. T. Wierzchon, eds.), vol. 7564 of LNCS, pp. 302–313, Springer, 2012.
P. Krömer, V. Snásel, J. Platos, M. Kudelka, and Z. Horak, “An aco inspired weighting approach for the spectral partitioning of co-authorship
networks,” in IEEE Congress on Evolutionary Computation, pp. 1–7, IEEE, 2012.
M. Kudelka, Z. Horák, V. Snášel, P. Krömer, J. Platoš, and A. Abraham, “Social and swarm aspects of co-authorship network,” Logic Journal of the
IGPL, vol. Special Issue: HAIS 2010, 2012.
V. Snásel, P. Krömer, J. Platos, M. Kudelka, Z. Horak, and K. Wegrzyn-Wolska, “Two new methods for network analysis: Ant colony optimization and
reduction by forgetting,” in Advances in Intelligent Web Mastering - 3, vol. 86 of Advances in Soft Computing, pp. 225–234, Springer, 2011.
P. Krömer, J. Martinovic, M. Radecký, R. Tomis, and V. Snášel, “Ant colony inspired algorithm for adaptive traffic routing,” in Nature & Biologically
Inspired Computing, Third World Congress on, NABIC 2011, pp. 336 – 341, IEEE, 2011.
Future work
GPU -> multi GPU -> (hybrid) GPU clusters
Superparallel meta-heuristics
Hybrid bio-inspired algorithms
Applications
Download