Data Analysis and Machine Learning

Image Credit: Michael Taylor

If we believe that the natural laws of physics are encoded in all observational data, the question then is, how do we extract them? This has occupied me for the best part of a decade now and I have developed generalisations of dimensional analysis using similarity transforms as well as evolutionary optimisation neural networks aiming to extract equations from data. At the moment I am working on dimensional reduction and developing Volterra neural networks based on nonlinear autoregressive integrated moving-average exogenous inputs (NARIMAX) input-output models.

Peer-Reviewed Articles:

1. Taylor M, Daglis IA, Anastasiadis A, Vassiliadis D (2011) Volterra network modeling of the nonlinear finite-impulse response of the radiation belt flux. American Institute of Physics [Modern Challenges in Nonlinear Plasma Physics] 1320:221-226. doi:10.1063/1.3544328 [Poster]  [arXiv] [post-print] [journal article] [BibTeX]
2. Taylor M, Daglis IA, Anastasiadis A, Vassiliadis D (2010) Identification of nonlinear space weather models of the Van Allen radiation belts using Volterra networks.Publications of the Astronomical Society of the Pacific [9^th International Conference of the Hellenic Astronomical Society] 424:92-98. [Poster] [arXiv] [post-print] [journal article] [BibTeX]
3. Taylor M, Diaz AI, Jodar-Sanchez LA, Villanueva-Mico RF (2008) A matrix generalisation of dimensional analysis: new similarity transforms to address the problem of uniqueness. Advanced Studies in Theoretical Physics, 2(20):979-995. [post-print] [journal article] [BibTeX]
4. Taylor M, Diaz AI, Jodar-Sanchez LA, Villanueva-Mico RF (2007) 100 years of dimensional analysis:  new steps toward empirical law deduction. [arXiv] [BibTex] (3 citations)
5. Taylor M and Diaz AI (2007) On the deduction of galaxy abundances with evolutionary neural networks. Publications of the Astronomical Society of the Pacific [From Stars to Galaxies: Building the Pieces to Build Up the Universe] 374:104-110. [arXiv] [post-print] [journal article] [BibTeX] (2 citations)
6. Perakakis P, Taylor M, Buela-Casal G (2005) A neuro-fuzzy system to calculate a journal internationality index. Congreso Español de Informática [Symposium on Fuzzy Logic and Soft Computing] 1:157-163.  [post-print] [BibTeX] (4 citations)
7. Taylor M (2005) An AI approach to quantitative modelling in Astrophysics. Congreso Español de Informática [Symposium on Fuzzy Logic and Soft Computing] 1:69-81. [post-print] [BibTeX]

Conference Talks:

1. Taylor M (2009) How well can machines be trained to predict space weather storms? ISARS-NOA Lecture Series National Observatory of Athens (Penteli), Athens, Greece [PPT Talk]
2. Taylor M (2007) Unravelling the nonlinear physics of high metallicity galaxies and HII regions with A.I. neural networks. ESTALLIDOS Star Formation Conference, IAA, Granada, Spain. [PPT Talk]
3. Taylor M (2006) Using artificial intelligence to deduce galactic abundances. STARGAL 2006, Venice, Italy. [PPT Talk]
4. Perakakis P, Taylor M, Buela-Casal G (2005) A neuro-fuzzy system to calculate a journal internationality index. Congreso Español de Informática [Symposium on Fuzzy Logic and Soft Computing] 1:157-163.  [PPT Talk]