Genetic programming

Genetic programming (GP) should be the Holy Grail of scientific computing, and indeed of many other sorts of problem solving. The goal of GP is not to evolve the solution to a problem, but to evolve a computer program that when executed will yield that solution. 

The potential of such a method is very considerable since, if it can be made to work, there would no longer be a need to write an explicit program to solve a problem. Instead a description of the problem would be fed into the GP program, the user would retire from the scene, and in due course a complete program to solve the problem would emerge. 

The GP is initialized with strings that code for random sequences of instructions and the fitness of the strings is assessed. Those strings that 

In constructing a GP program, it is necessary not only to define GA-like parameters, such as the population size, the type of crossover, and the number of generations, but also parameters that relate specifically to the components of the candidate programs that the GP will build. These include  

This program would be allocated maximum fitness as it exactly matches the required output. However, the same GP run might create a second program that used the function  

we start the calcnlation with some initial first member From this guess, the generation is filled by creating new members throngh a combined process of single-member mutation and dual-member crossing. For each A = 2, 3. p, we add a new member 

To decide which process to use to addx t, we generate a random number Uk, uniformly distributed on [0,1]. If Uk am, we generate by selecting at random some previous member of the population with j k, and mutating it. A simple mutation is to displace 

If Uk dm and k 2, we generate by a random crossing of two previous members of the population. We select at random some J k and I kto serve as parents. One simple way to perform the crossing is to select at random for each component the value of one parent or the other  

When considering a member x l for admittance into the set of survivors, we check to see whether for all previously identified survivors d j d. Only if x is sufficiently different from all of the previous survivors will it be accepted. If we cannot find s sufficiently diverse survivors, we move on to the next generation with less than n, as propagating two nearly identical members does little good. 

The supplemental material at the accompanying website contains an additional example of a genetic algorithm. This example converts the process of solving a Sndoku puzzle into a discrete optimization that is then solved by a genetic algorithm. 

---

The events which occur prior to the death of short-term cultures have been referred to as the process of cellular aging, or apoptosis. Whether in vitro cellular aging is due to an inherited genetic program, or whether in vitro aging is simply a consequence of an imperfect cell culture environment is a matter of controversy. The possibility that the cellular aging which occurs in vitro resembles cellular aging in vivo is unresolved. 

Almost invariably, a neuron is genetically programmed to synthesize and release only a single type of neurotransmitter. Therefore, a given synapse is either always excitatory or always inhibitory. Once a neurotransmitter has bound to its receptor on the postsynaptic neuron and has caused its effect, it is important to inactivate or remove it from the synapse in order to prevent its continuing activity indefinitely. Several mechanisms to carry this out have been identified ... 

Eberhart2 is one of the few recent introductions to that field, but also covers genetic algorithms and genetic programming by way of setting the scene. 

Koza, J.R., Keane, M.A., and Streeter, M.J., What s AI done for me recently Genetic Programming s human-competitive results, IEEE Intell. Systs. 18, 25, 2003. 

The unmodulated functions of the developmental program of a developmental Cartesian genetic programming (DCGP) node. 

Genetic programming, a specific form of evolutionary computing, has recently been used for predicting oral bioavailability [23], The results show a slight improvement compared with the ORMUCS Yoshida-Topliss approach. This supervised learning method and other described methods demonstrate that at least qualitative (binned) predictions of oral bioavailability seem tractable directly from the structure. 

Molecular Genetics Program, Victor Chang Cardiac Research Institute (VCCRI), Sydney, Australia and... 

---