The Multi-perturbation Shapley value Analysis (MSA) is a framework for deducing causal function localization from multiple perturbations data. It provides insights to the workings of a given system, the "players" taking part in carrying out the different functions and the functional interactions between those "players". Essentially, it makes use of a set of multiple perturbations that are afflicted upon the system, while measuring the system's performance score in each, to quantify the importance of each of the elements, as well as the interactions between them. See [1,2] below for more about the MSA.

The MSA has a wide range of potential applications. We are currently engaged in the following projects:

  • We have mainly utilized the MSA for analyzing the evolved neurocontrollers of autonomous agents, as a testbed for real biological networks but also in their own right. Such analyses provide insights into the workings of the neurocontrollers, leading to conclusions concerning the evolutionary algorithms and allowing to test hypotheses regarding the neural processing. See [2,3,4,6,7,8,9,10] below for applications of the MSA for understanding our agents' brains.
  • The MSA can be utilized for the analysis of neurophysiological models. Analyzing such models can extend their usefulness by allowing for a deeper understanding of their operation. In [1] we have demonstrated the workings of the MSA for the analysis of the building block of a model of lamprey swimming controller.
  • In neuroscience, the MSA may help in analyzing different types of lesioning experiments, such as reversible deactivation experiments, neuronal laser ablations and transcranial magnetic stimulation "virtual lesions". In [1,5] below the approach is applied to reversible cooling deactivation experiments in cats, establishing to contributions and interactions of cortical and collicular sites to the brain function of spatial attention to auditory and visual stimuli.
  • The MSA further offers a rigorous way of making sense out of gene knockout experiments [11], using technologies such as RNA interference, as well as analyzing models of genetic and metabolic networks.


MSA Publications
A. Keinan, B. Sandbank, C. C. Hilgetag, I. Meilijson, E. Ruppin, Fair attribution of functional contribution in artificial and biological networks, Neural Computation, 16(9), 1887-1915, 2004
A. Keinan, B. Sandbank, C. C. Hilgetag, I. Meilijson, E. Ruppin, Axiomatic scalable neurocontroller analysis via the Shapley value, Artificial Life, to appear
A. Keinan, C. C. Hilgetag, I. Meilijson, and E. Ruppin, Causal localization of neural function: The Shapley value method, Neurcomputing 58-60C, 215-222, 2004
K. Saggie, A. Keinan, and E. Ruppin, Spikes that count: Rethinking spikiness in neurally embedded systems, Neurocomputing, 58-60C, 303-311, 2004
A. Keinan, A. Kaufman, N. Sachs, C. C. Hilgetag, and E. Ruppin, Fair localization of function via multi-lesion analysis, Journal of Neuroinformatics (Special issue on Functional Connectivity), 2(2), 163-168, 2004
K. Saggie-Wexler, A. Keinan, and E. Ruppin, Neural processing of counting in evolved spiking and McCulloch-Pitts agents, Artificial Life, to appear
Z. Ganon, A. Keinan, and E. Ruppin, Evolutionary network minimization: Adaptive implicit pruning of successful agents, In W. Banzhaf, T. Christaller, P. Dittrich, J. T. Kim, & J. Ziegler (Eds.), Advances in artificial life - proceedings of the 7th european conference on artificial life (ECAL) (Vol. 2801, pp. 319??327). Springer Verlag Berlin, Heidelberg, 2003
K. Saggie, A. Keinan, and E. Ruppin, Solving a delayed response task with spiking and McCulloch-Pitts agents, In W. Banzhaf, T. Christaller, P. Dittrich, J. T. Kim, & J. Ziegler (Eds.), Advances in artificial life - proceedings of the 7th european conference on artificial life (ECAL) (Vol. 2801, pp. 199??208). Springer Verlag Berlin, Heidelberg, 2003
S. Boshy and E. Ruppin, Evolving Small Neurocontrollers With Self-Organized Compact Encoding, Artificial Life Vol. 9(2) pp 131-151, 2003
A. Keinan, Analyzing evolved fault-tolerant neurocontrollers, Proceedings of the ninth international conference on the simulation and synthesis of living systems (ALIFE9), MIT Press, 557-562, 2004
A. Kaufman, M. Kupiec and E. Ruppin, Multi-Knockout Genetic Network Analysis: The Rad6 Example, Computational Systems Bioinformatics, CSB2004
Software
We have created a Matlab{R} package, including routines for the computation of contributions of elements and interactions between them in a given system, using the various methods in the MSA framework. The package is freely available for academic use. To acquire it, please e-mail Nir Yosef.
Design and HTML Programming: Elhanan Borenstein, Dynamic Content Script: Shay Cohen
Background image adapted with permission from the E.Coli metabolic map
produced using SimPheny™ from Genomatica, Inc.
For any comment please contact the Webmaster
Last updated: 9:51, 10/10/2012