Artificial Intelligence

 Radial Basis Function Network (RBFN)

A Radial Basis Function Network (RBFN) is a type of neural network. As seen in the diagram, the input has been given to the first layer that is the input layer which is passed to the second layer which is the output layer. Finally, the output has been obtained. The XOR function has been used in the input and output layer both.

RBF and Backpropagation Learning Comparison

The BPNN is the abbreviation of a backpropagation neural network. The BPNN system structures a bit of MLP structural building with a supervised studying method. Thus, it is a kind of subsistence forward neural framework.

Based on the function approximation theory, the RBFNN is the abbreviation of a radial basis function neural network. RBFNN has a distinctive training algorithm consisting of unsupervised and supervised both.

The RBFNN-based model is a more accurate predictor than the BPNN model when performance indices are compared.

Hopfield, RBF, and Kohonen networks all integrates unsupervised approach

Hopfield, RBF, and Kohonen networks all integrates unsupervised approach which can be validated as follows:

In Radial Basis Function Network, the first layer is built in an unsupervised way. These networks obtain their potential from the greater number of nodes in the unsupervised layer. The hidden layer and its parameters are trained in an unsupervised way. If unsupervised methods are used, they are efficient in training which is an appealing characteristic.

Studying masses of the Hopfield network tacitly makes an unsupervised model of the learning data set. Hopfield networks produce unsupervised models from training data.

Despite the lower usage, Kohonen networks have a noteworthy capacity in the unsupervised setting the latest generation of deep learning.

Prolog

Prolog is a logic programming language associated with computational linguistics and artificial intelligence. Prolog is intended mainly as a declarative programming language. Unlike many other programming languages, Prolog has its roots in first-order logic. The program logic is represented as rules and facts, expressed in terms of relations. By running a query, a computation is started over these relations. Prolog is one of the most popular languages currently, with various commercial and free implementations available. All prolog compilers do not support modules and due to the issues of compatibility between module systems of the aim prolog compilers, programming in large is considered complicated. For having a high-performance penalty, developed software in Prolog has been criticized as compared to non-declarative programming languages.

Heuristic

The process of solving problem or self-examination that make use of a practical method which is enough for attaining a short-term, instant objective but no guaranteed to be ideal, logical or best can be defined as a heuristic. Heuristic techniques can be used for speeding up the task of discovering a satisfying result, where searching the best result is impractical or impossible. Heuristics can be time-saving and in making the cognitive burden easy for taking a decision. Thumb rule, an educated guess, or trial and error are the examples that make use of heuristics.

Admissible Heuristic

The admissible heuristic can be defined as a heuristic function that never overrates the price of achieving aim which means price it rates for reaching the aim is not greater than the least feasible price from the present position in the approach regards to computer science, especially in pathfinding algorithms.

Greedy Search

The search which pursues the heuristic of solving a problem for selecting the best solution locally at each phase to discover comprehensive optimum can be called a greedy search. A greedy search does not generally outcomes an ideal result in many problems, but a greedy search may provide optimal results locally which is near to an optimal solution globally in a rational portion of the time.

Expectimax Search

Expectimax search is a decision-making/search algorithm that increases the expected(average) benefit to its maximum capacity. The expectimax search is usually applicable to trees with stochastic nodes, where the output of an action is not certain.

Chance node s_i: expected value of outcomes s_ij

V* (s_i)= E[V* (s_i)]= ∑_j p_ijV* (s_ij)

Odds

The probability that an incident will happen is the fragment of times expected to see the respective events in numerous tests. Probabilities every time scales in between 0 and 1. Conditional probability is an estimate of the probability of an event happening provided that another event has happened by evidence, presumption, assertion, or assumption. In probability theory, the conditional probability concept is one of the most basic and vital, but it can be quite unpredictable and need attentive understanding. For instance, there has not to be a causal relationship between A and B, and they should not occur at the same time. The division of the probability that the event will occur to the probability that the event will not occur can be defined as the odds.