I have in mind, that Logic is mainly about thinking, abstract thinking, particularly reasoning. Reasoning is a process, structured by steps, when one conclusion usually is based on a previous one, and at the same time it can be the base, the foundation of further conclusions. Despite the mostly intuitive character of the algorithm as a concept (even not taking into account Turing and Markov theories/machines), it has a step by step structure, and they are connected, even one would say that logically connected (when they are correct algorithms). The different is, of course, the formal character of the logical proof.
Teachable neural networks are currently the main direction in the development of artificial intelligence (AI). These systems belong to the category of inductive anthropomorphic, which is associated with their well-known advantages, disadvantages and areas of application. It is generally accepted that anthropomorphism is the exclusive factor in the advantage of AI, however, there are actual areas of AI application, in which the use of deductive systems is more effective, which use less data and computational resources, but are capable of learning. Examples of such systems: knowledge generators, expert systems, active marketing systems, and many others.
The basis of AI deductive imitation systems are extremely general models and algorithms that imitate the objects (objects) under consideration. The generality limit is determined both by the amount of knowledge about the object and by the calculated capabilities. General models (metamodels) are usually formed in the form of systems of equations that link together the variables that describe the object under consideration, as well as parameters that determine the form of partial equations formed when transforming the equations of the general model. The solution of such systems of equations is carried out by methods of global optimization: stochastic (Monte Carlo methods) or heuristic (evolutionary algorithms and others). Simulation AI presupposes the solution of two basic problems: inverse (implementing the system training options) - the formation of partial equations (models) according to the values of certain variable characteristics of objects; direct - definitions of values of variables for particular configurations of models. The use of simulation metamodels in combination with the randomization of variables and parameters provides high flexibility and adaptability of intelligent systems.
In other words, the question is: does proving consistency of deductive structures help to develop skills in the activity of writing algorithms? I have in mind the common, the similar mental actions, taking place in both process. Having skills, abilities, in formal deduction, could support the development of algorithmic skills (in fact, formal proving could be considered a "kind" of algorithm).
Dear Mirzakhmet, could you specify what concretely you have in mind when saying "to start from examples practicing deductive style…"?
When referring to studying deductive structures as a way of support the development of algorithmic skills, I have not considered, necessarily, any particular approach to study them. Probably, a good introduction could be presenting problems, where deductive approach is a recommended way for solving problems.
- Badges
- Science topic
- Similar topics
- Psychology
- Mental Processes
- Learning