• No se han encontrado resultados

CAPÍTULO II – DEL CONCURSO DE ACREEDORES A LA MEDIACIÓN

2.5. Derecho comparado

The preceding example illustrates how adaptive devices use the set of rules as their only element for represent- ing and handling knowledge.

A rule (here, a transition) may handle parametric information in its components (here, the transition’s origin and destination states, the token labeling the transition, the adaptive function it calls, etc.).

Rules may be combined together in order to represent some non-elementary information (here, the sequences

of transitions consuming tokens “b” and “c” keep track

of the value of n in each particular sentence). This way, rules and their components may work and may be in- terpreted as low-level elements of knowledge.

Although being impossible to impose rules on how to represent and handle knowledge in systems repre-

1 2 3 a /A() ε ε 1 2 3 a /A() ε ε

Figure 1. Initial configuration of the illustrative adap- tive automaton

Figure 2. Adaptive function

A

( )

x ε 2 ? [ ] 2 ε y ? [ ] x ε 2 ε y – [ ] 2 ε ε + [ x b b c c c y ] { A() = } x ε 2 ? [ x ε 2 ] ? [ ] 2 ε y ? [ 2 ε y ] ? [ ] x ε 2 ε y – [ x ε 2 ε y ] – [ ] 2 ε ε + [ x b b ε 2 ε c c c y ] + [ x b b c c c y ] { A() = }

Adaptive Technology and Its Applications

A

sented with adaptive devices, the details of the learning process may be chosen according to the particular needs of each system being modeled.

In practice, the learning behavior of an adaptive

device may be identified and measured by tracking

the progress of the set of rules during its operation and interpreting the dynamics of its changes.

In the above example, when transitions are added to the automaton by executing adaptive action

A

( ), one may interpret the length of the sequence of transitions

consuming “b” (or “c”) as a manifestation of the knowl- edge that is being gathered by the adaptive automaton on the value of n (its exact value becomes available

after the sub-string of tokens “a” is consumed).

FUTURE TRENDS

Adaptive abstractions represent a significant theoreti- cal advance in Computer Science, by introducing and exploring powerful non-classical concepts such as: time-varying behavior, autonomously dynamic rule sets, multi-level hierarchy, static and dynamic adap- tive actions.

Those concepts allow establishing a modeling style, proper for describing complex learning systems, for

efficiently solving traditionally hard problems, for

dealing with self-modifying learning methods, and for providing computer languages and environments for comfortable elaboration of quality programs with dynamically-variant behavior.

All those features are vital for conceiving, modeling,

designing and implementing applications in Artificial Intelligence, which benefits from adaptivity while expressing traditionally difficult-to-describe Artificial

Intelligence facts.

Listed below are features Adaptive Technology

offers to several fields of Computation, especially to Artificial Intelligence-related ones, indicating their

main impacts and applications.

• Adaptive Technology provides a true computation model, constructed around formal foundations.

Most Artificial Intelligence techniques in use

are very hard to express and follow since the connection between elements of the models and information they represent is often implicit, so

their operation reasoning is difficult for a human

to track and plan. Adaptive rule-driven devices concentrate all stored knowledge in their rules, and the whole logic that handles such information, in their adaptive actions. Such properties open for

Artificial Intelligence the possibility to observe,

understand and control adaptive-device-modeled phenomena. By following and interpreting how and why changes occur in the device set of rules, and by tracking semantics of adaptive actions, one can infer the reasoning of the model reactions to its input.

• Adaptive devices have enough processing power to model complex computations. In (Neto, 2000) some well-succeeded use cases are shown with Figure 3. Configurations of the adaptive automaton after executing A ( )once and twice

1 1 2 3 a /A() ε ε b b c c c a /A() ε ε 2 b b b b 3 c c c c c c

Adaptive Technology and Its Applications

simple and efficient adaptive devices used instead

of complex traditional formulations.

• Adaptive Devices are Turing Machine-equiva- lent computation models that may be used in the

construction of single-notation full specifications

of programming languages, including lexical, syntactical, context-dependent static-semantic is- sues, language built-in features such as arithmetic operations, libraries, semantics, code generation and optimization, run-time code interpreting, etc.

• Adaptive devices are well suited for representing complex languages, including idioms. Natural language particularly require several features to

be expressed and handled, as word inflexions, or- thography, multiple syntax forms, phrase ordering, ellipsis, permutation, ambiguities, anaphora and others. A few simple techniques allow adaptive devices to deal with such elements, strongly sim- plifying the effort of representing and processing them. Applications are wide, including machine translation, data mining, text-voice and voice-text conversion, etc.

• Computer art is another fascinating potential application of adaptive devices. Music and other artistic expressions are forms of human language. Given some language descriptions, computers can capture human skills and automatically generate interesting outputs. Well-succeeded experiments

were carried out in the field of music, with excel- lent results (Basseto, 1999).

• Decision-taking systems may use Adaptive Deci- sion Tables and Trees for constructing intelligent systems that accept training patterns, learn how to classify them, and therefore, classify unknown patterns. Well-succeeded experiments include: classifying geometric patterns, decoding sign languages, locating patterns in images, generat- ing diagnoses from symptoms and medical data, etc.

• Language inference uses Adaptive Devices to generate formal descriptions of languages from samples, by identifying and collecting structural information and generalizing on the evidence of repetitive or recursive constructs (Matsuno, 2006).

• Adaptive Devices can be used for learning pur- poses by storing as rules the gathered information on some monitored phenomenon. In educational

systems, the behavior of both students and train- ers can be inferred and used to decide how to proceed.

• One can construct Adaptive Devices whose underlying abstraction is a computer language. Statements in such languages may be considered

as rules defining behavior of a program. By at- taching adaptive rules to statements, the program

becomes self-modifiable. Adaptive languages are

needed for adaptive applications to be expressed naturally. For adaptivity to become a true pro- gramming style, techniques and methods must be developed to construct good adaptive software, since adaptive applications developed so far were usually produced in strict ad-hoc way.

CONCLUSION

Adaptive Technology concerns techniques, methods and subjects referring to actual application of adaptivity.

Adaptive automata (Neto, 1994) were first proposed

for practical representation of context-sensitive lan- guages (Rubinstein, 1995). Adaptive grammars (Iwai, 2000) were employed as its generative counterpart (Burshteyn, 1990), (Christiansen, 1990), (Cabasino, 1992), (Shutt, 1993), (Jackson, 2006).

For specification and analysis of real time reactive

systems, works were developed based on adaptive versions of statecharts (Almeida Jr., 1995), (Santos,

1997). An interesting confirmation of power and

usability of adaptive devices for modeling complex systems (Neto, 2000) was the successful use of Adap- tive Markov Chains in a computer music-generating device (Basseto, 1999).

Adaptive Decision Tables (Neto, 2001) and Adap- tive Decision Trees (Pistori, 2006) are nowadays being experimented in decision-taking applications.

Experiments have been reported that explore the potential of adaptive devices for constructing language inference systems (Neto, 1998), (Matsuno, 2006).

An important area in which adaptive devices shows

its strength is the specification and processing of natural

languages (Neto, 2003). Many other results are being achieved while representing syntactical context-depen- dencies of natural language.

Simulation and modeling of intelligent systems are other concrete applications of adaptive formalisms, as illustrated in the description of the control mechanism

Adaptive Technology and Its Applications

A