Filtyp
We have developed a prognostic index model for survival data based on an ensemble of artificial neural networks that optimizes directly on the concordance index. Approximations of the c-index are avoided with the use of a genetic algorithm, which does not require gradient information. The model is compared with Cox proportional hazards (COX) and three support vector machine (SVM) models by Van Bel
The mean field theory approach to knapsack problems is extended to multiple knapsacks and generalized assignment problems with Potts mean field equations governing the dynamics. Numerical tests against 'state of the art' conventional algorithms shows good performance for the mean field approach. The inherently parallelism of the mean field equations makes them suitable for direct implementations i
[abstract missing]
The majority of microarray studies evaluate gene ex- pression differences between various specimens or con- ditions. However, the causes of this variability often re- main unknown. Our aim is to identify underlying causes of these patterns, a process that would eventually enable a mechanistic understanding of the deregulation of gene expression in cancer. The procedure consists of three phases: pr
Recent evidence has revealed the role of mechanical cues in the development of shapes in organisms. This thesis is an effort to test some of the fundamental hypotheses about the relation between mechanics and patterning in plants. To do this, we develop mechanical models designed to include specific features of plant cell walls. These are heterogeneous stiffness and material anisotropy as well as
A mean field feedback artificial neural network (ANN) algorithm is developed and explored for the set covering problem. A convenient encoding of the inequality constraints is achieved by means of a multilinear penalty function. An approximate energy minimum is obtained by iterating a set of mean field equations, in combination with annealing. The approach is numerically tested against a set of pub
A novel artificial neural network approach to constraint satisfaction problems is presented. Based on information-theoretical considerations, it differs from a conventional mean-field approach in the form of the resulting free energy. The method, implemented as an annealing algorithm, is numerically explored on a testbed of K-SAT problems. The performance shows a dramatic improvement over that of
A novel method is presented and explored within the framework of Potts neural networks for solving optimization problems with a non-trivial topology, with the airline crew scheduling problem as a target application. The key ingredient to handle the topological complications is a propagator defined in terms of Potts neurons. The approach is tested on artificial problems generated with two real-worl
A variational method for computing conformational properties of molecules with Lennard-Jones potentials for the monomer-monomer interactions is presented. The approach is tailored to deal with angular degrees of freedom, rotors, and consists of the iterative solution of a set of deterministic equations with an annealing in temperature. The singular short-distance behavior of the Lennard-Jones pote
We consider the thermodynamics of a uniformly charged polyelectrolyte with harmonic bonds. For such a system there is at high temperatures an approximate scaling of global properties like the end-to-end distance and the interaction energy with the chain length divided by the temperature. This scaling is broken at low temperatures by the ultraviolet divergence of the Coulomb potential. By introduci
A new Monte Carlo method for computing thermodynamical properties of very large polyelectrolytes is presented. It is based on a renormalization group relating the original polymer to a smaller system, where, in addition to the naively rescaled forces, a corrective nearest-neighbor interaction originating from the short distance Coulomb cutoff is introduced. The method is derived for low T but is i
The Apollonian tiling of the plane into circles is analyzed with respect to its group properties. The relevant group, which is noncompact and discrete, is found to be identical to the symmetry group of a particular geometric tree graph in hyperbolic three-space. A linear recursive method to compute the radii is obtained. Certain modifications of the problem are investigated, and relations to other
We discuss the Langevin equation for a complex Boltzmann distribution, allowing for modifications of the process. The relation between the two different time development operators involved is analyzed, with emphasis on their spectra.
We describe and test an implicit solvent all-atom potential for simulations of protein folding and aggregation. The potential is developed through studies of structural and thermodynamic properties of 17 peptides with diverse secondary structure. Results obtained using the final form of the potential are presented for all these peptides. The same model, with unchanged parameters, is furthermore ap
Designing amino acid sequences that are stable in a given target structure amounts to maximizing a conditional probability. A straightforward approach to accomplish this is a nested Monte Carlo where the conformation space is explored over and over again for different fixed sequences. In this paper we discuss an alternative Monte Carlo approach, multisequence design, where conformation and sequenc
We study the phase behavior of single homopolymers in a simple hydrophobic/hydrophilic off-lattice model with sequence independent local interactions. The specific heat is, not unexpectedly, found to exhibit a pronounced peak well below the collapse temperature, signalling a possible low-temperature phase transition. The system size dependence at this maximum is investigated both with and without
We study a simple heteropolymer model containing sequence-independent local interactions on both square and triangular lattices. Sticking to a two-letter code, we investigate the model for varying strength κ of the local interactions; κ=0 corresponds to the well-known HP model [K. F. Lau and K. A. Dill, Macromolecules 22, 3986 (1989)]. By exhaustive enumerations for short chains, we obtain all str
A method for sequence optimization in protein models is presented. The approach, which has inherited its basic philosophy from recent work by Deutsch and Kurosky [Phys. Rev. Lett. 76, 323 (1996)] by maximizing conditional probabilities rather than minimizing energy functions, is based upon a different and very efficient multisequence Monte Carlo scheme. By construction, the method ensures that the
Folding properties of a two-dimensional toy protein model containing only two amino acid types, hydrophobic and hydrophilic, respectively, are analyzed. An efficient Monte Carlo procedure is employed to ensure that the ground states are found. The thermodynamic properties are found to be strongly sequence dependent in contrast to the kinetic ones. Hence, criteria for good folders are defined entir
