There is yet to be a precise definition as to what Artificial Neural Network is, though many researchers would agree that it concerns a network of austere processing elements – otherwise known as the neurons, which presents complex behaviour established by the relationship amongst processing and parametrical elements. The main inspiration that lead to the development of this technique was from the investigation of, no lesser than, our Central Nervous System and the neurons (including their axons, dendrites and synapses) which make up its most important information processing elements.
A neural network model would show us that simple nodes are connected forming a network of nodes — thus, its coining as “neural network. ” A Neural Network functions in 2 different manners – learning and testing. The former would literally mean, the system learns the ways it is supposed to behave while the latter is when rigorous repetition of training would eventually result to a stable system, defined by its giving of constant satisfactory outputs.
Most “abstract reasoning” of an Artificial Neural Networks are being implemented through three learning types – supervised, unsupervised and the reinforced learning, as has been introduced in the preceding paragraphs. Supervised learning entails a functional relationship between the input and the output. The system has to learn every possible IO pair that can be thought of. In case, there is a miss, all that has to be done is to input the said pair into the memory of the system hence when it resurfaces, the system knows how the handle it.
Hence, basically, the goal is to ‘teach’ the network to identify the given input with the desired output. (Sordo 2002) This is usually best achieved when function f has already been derived to represent the behaviour of the Neural Network system. For unsupervised learning, we feed an input and a function to the system and record what behaviour the system outputs with such input and function. To begin with the learning process, there are no IO-pairs as opposed to supervised learning.
Ultimately, the main goal of achieving the stable state will be attained through rigorous repetition of test with different sets of inputs. This type of systems – imploring unsupervised learning as its method of learning, are best displayed in statistical modelling, and the likes. Reinforcement learning stems its roots from the related psychological theory that has been conceived even before AI has been.
Dynamically, in this type of learning, the machine interacts with its environment by producing actions a1, a2, … These actions affect the state of the environment, which in turn results in the machine receiving some scalar rewards (or punishments) r1, r2, … The goal of the machine is to learn to act in a way that maximizes the future rewards it receives (or minimises the punishments) over its lifetime. Reinforcement learning is closely related to the fields of decision theory (in statistics and management science), and control theory (in engineering).
The fundamental problems studied in these fields are often formally equivalent, and the solutions are the same, although different aspects of problem and solution are usually emphasised. (Ghahramani 2004) Neural networks with always have the outstanding characteristic of deriving intelligence from the usually complicated and, oftentimes, fuzzy data stored in the neurons. These systems, oftentimes, offer to be easy utilities to deduce patterns and perceive trends that are difficult to be noticed by either human observation or by our current computer intelligence.
A trained neural network is regarded as an “expert” in the category of information it has been given to analyze. This expert can then be used to provide projections given new situations of interest and answer “what if” questions. (Chung et al 2007) It is used for adaptive learning on how to handle tasks based on the input provided for training or preliminary experience. It is a self-organizational tool that hones its own picture of the data it receives in as early as learning time. Neural networks another feature is that it is a real-time operation system where all calculation may be performed in parallel.
Fault Tolerance via Redundant Information Coding is another aspect of the neural system where partial destruction of a network leads to the corresponding degradation of performance. However, some network capabilities may be retained even with major network damage. The platform to a successful implementation. Several environments can be used in totally implementing a Cross-Language Translator through with the various and fast developments in computer technology since its introduction.
In the succeeding paragraphs we will be tackling some of those that has come the author’s A-list. Microsoft . NET Framework. This framework form part of Microsoft Windows operating systems, containing a vast number of pre-coded resolutions to general program requirements, and governing the performance of programs written particularly for the framework. This framework is a vital Microsoft contribution and is projected on being utilized by most applications created and to be created for Windows platform.