AN ARTIFICIAL INTELLIGENCE BASED NEURO-FUZZY SYSTEM WITH EMOTIONAL INTELLIGENCE

1. C.SARANYA, 2.A.RAJAM, U.G STUDENTS DEPT OF CSE, P.S.R RENGASAMY COLLEGE OF ENGINEERING FOR WOMEN ,SIVAKASI. E-MAIL ID:saranyanemo@gmail.com , PH-NO:7708343887.

ABSTRACT- In a very real sense we have two minds, one that thinks (rational mind) and one that feels (emotional mind). These two fundamentally different ways of knowing interact to construct our mental life. This results in two different kinds of intelligence i.e, rational and emotional. Intellect cannot work at its best without emotional intelligence. If our aim in AI is to build systems that behave like human beings then it is necessary that we incorporate elements of both rational and emotional intelligence into the system. The main concept in this paper is to develop an Intelligent Agent that exhibits the notion of emotional intelligence. The architecture that will be considered for implementation is a Neuro-Fuzzy system with concepts of artificial intelligence incorporated into the functionality of the system. The approach chosen for the implementation is Soft Computing which is basically a synergistic integration of three computing paradigms: neural networks, fuzzy logic and probabilistic reasoning, to provide a flexible framework to construct computationally intelligent systems. KEYWORDS- Fuzzification, Defuzzification, Hopfield Network, Patter Matching, Linguistic variables, Membership Function, Activation, Emotional State. I. INTRODUCTION Agents are entities capable of autonomous goal-oriented behavior in some environment. An intelligent agent is a device that interacts with its environment in flexible, goaldirected ways, recognizing important states of the environment and acting to achieve desired results. They can have one or more attributes such as Autonomous, Adaptive, Communicative, Collaborative, Personal and Mobile. The attribute that will be considered for the current system is Personal which enables the agent to exhibit believable character behavior (emotion). Research in the domain of emotion has found that intelligence and emotion are linked. Emotional Agent has an emotional state

which gets altered by stimuli from environment or internal elements. Emotional state steers the decision process. An emotional agent can be implemented by developing a computational approach. Emotional intelligence represents an ability to validly reason with emotions and to use emotions to enhance thought. Physiological research was revealing the existence of neural pathways for unconscious emotional processing. Neurologists have made progress in demonstrating that emotion is as, or more, important than reason in the process of making decisions and deciding actions. The significance of these findings should not be overlooked in a world that is increasingly reliant on computers to accommodate to user needs. Emotional Intelligence involves the following concepts: o o o o Managing Emotions Understanding Emotions Facilitating Thought Perceiving Emotions

Neural networks are designed to operate in numeric frameworks. They deal with problem space variables that are numerically quantifiable, or that are measurable. However, much of the information available in the real world cannot be measured. In order to exploit such vast repositories of information while developing models of real world processes, one needs a mechanism to embed linguistic information into neural network operational environments. Fuzzy set theory provides precisely this modeling platform. Fuzzy logic and neural networks are complementary technologies. They work at different levels of abstraction and individually provide rich functionality, which when brought together in a cohesive fashion provide us with intelligent systems. A combination of these two technologies endows systems with a two fold advantage. Fuzzy logic provides a high level framework for approximate reasoning that can appropriately handle the uncertainty and imprecision in linguistic semantics.

Neural networks provide self-organizing framework for low level representation of information with on-line adaptation capabilities. It therefore seems both plausible and justified to attempt combining both these approaches in the design of intelligent systems. II. NEURO-FUZZY SYSTEM: Neuro-Fuzzy system is a fuzzy system that uses a learning algorithm derived from or inspired by neural network theory to determine its parameters (fuzzy sets and fuzzy rules) by processing data samples. Neuro-fuzzy modeling is a pivotal technique in soft computing wherein neural network learning concepts are incorporated in fuzzy inference systems. The system comprises of three different layers :

emotional multi-agent project. The model only describes basic emotions and innate reactions; however, it presents a good starting point for building computer simulations of emotion. The basic emotions that were simulated in the model are anger, fear, distress/sadness, enjoyment/ happiness, disgust, and surprise. III. MODEL OF THE SYSTEM

Architecture Unit

1. Fuzzification Layer: Here each neuron represents an

input membership function of the antecedent of a fuzzy rule.

Data converter

2. Fuzzy Inference Layer : In this layer, fuzzy rules are fired and the value at the end of each rule represents the initial weight of the rule, and will be adjusted to its appropriate level at the end of the training. 3. Defuzzification Layer : Here each neuron represents a consequent proposition and its membership function can be implemented by combining sigmoid functions and linear functions.

Event processor (Neuro fuzzy system)

Emotional state calculator (Hop field network)

Fuzzification layer

Fuzzy Inferences Layer

Defuzzifi -cation Layer

Output unit

Fig 1 : Neuro Fuzzy System

Fig 2:Schematic representation of the architecture

ADAPTIVE MODEL OF EMOTIONS FLAME-Fuzzy Logic Adaptive Model of Emotions was modeled to produce emotions and to simulate the emotional intelligence process. FLAME was built using fuzzy rules to explore the capability of fuzzy logic in modeling the emotional process. Psychology, Neurology, Philosophy and Cognitive Science have been concerned with modeling the mind and its behavior for many years. By the 1990s, the Japanese researchers were interested in a system that can communicate with humans. Emotions were regarded as one of the most important factors in communication The MIT lab is also producing an

The system comprises of the following units : 1) Input Unit ( accepts incoming events ). 2) Data converter ( for suitable conversions ). 3) Event Processor ( Neuro-Fuzzy system ) with the following five layers : The input layer The input Fuzzy layer The conjunction layer The output fuzzy layer The output layer 4) Emotional State Calculator ( Hopfield neural

network )

5) Output Unit ( generates O/P in desired form) IV. Methodology

Step 1 : The input to the system is an external event which is a string representing the real world event. Step 2 : The string is converted into corresponding linguistic variables and fed to the event processor which is a neuro-fuzzy system. Step 3 : In the event processor the activation level of an input unit is the value of a certain input variable in the given instance. The input value is passed onto fuzzy set units, which then translate the value into a degree of membership as the activation level of a fuzzy set unit. The conjunction unit will take the minimum of the inputs ( degrees of membership ) it receives from the input fuzzy set units below. Now, the output fuzzy set unit will collect information from one or more conjunction units (each corresponding to a fuzzy rule). An output fuzzy unit may take the maximum or sum of its inputs or employ a combination function. Then the output unit generates the final result by integrating the information from the output fuzzy set units. How the output unit calculates its activation level depends on the defuzzification scheme adopted. The activation function used in each layer is as follows : 1. 2. 3. 4. 5. The input layer : no activation function. The input fuzzy layer : the membership function. The conjunction layer : the min function. The output fuzzy layer : the max function. The output layer : no activation function. Consider a single node from each layer .Let us denote f as the transfer function (which is a weighted sum in a normal neural network), a as the activation function, ukp as the inputs to the node, wkp as the input weights and oki as the outputs of the node, where k=15, p and i are the total number of the input and output variables respectively. The process of each node can be described as follows:

in terms of a bell-shaped fuzzy membership function.

o Layer 3: this layer is a fuzzy rule layer where

each node represents one fuzzy rule. The links in this layer are used to perform antecedent matching fuzzy logic rules and also the fuzzy AND operation. The link weight, w3i, at the third layer is also equal to unity. consequences of the rule nodes. The links in the third and fourth layer, in association, function as a fuzzy inference engine. The nodes in this layer have two operation modes: the down-up transmissions and the up-down transmission mode. In the down-up mode, the links in the layer should perform the fuzzy OR operation to integrate the fired rules which have the same consequence. a = min(1,f) is also equal to unity. In the up-down mode, the nodes in the layer 4 and the layer 5 function exactly the same way of the nodes in the layer 2 except that only a single node is used to perform a membership function for the output l Layer 5: is an output linguistic layer, which has two nodes for each input variable. The first type of node is used to feed the networks for training it again.

o Layer 4: the links to this layer define

o Layer 5 : The link weight, w4i, at the fourth layer

Step 4 : The output layer of the Event Processor generates a numeric pattern corresponding to the input event. Step 5 : This pattern is taken up by the Emotional State Calculator i.e., a Hopfield Neural Network which is a simple artificial network which is able to store certain memories or patterns in a manner rather similar to the brain - the full pattern can be recovered if the network is presented with partial information. The Hopfield network is created by supplying input data vectors, or pattern vectors, corresponding to the different classes of emotions. These patterns are called class patterns. In an n-dimensional data space the class patterns should have n binary components {1,-1}; that is, each class pattern corresponds to a corner of a cube in an n-dimensional space. The network stores a set of predefined patterns and their mapping onto the emotional states, in a coded form.

o Layer 1: consists of the input nodes and directly

transmits the input linguistic variables to the next layer. The link weight, w1i, at the layer one is unity.

o Layer 2: is the input term layer where each node

acts as a membership function and represents the terms of the respective linguistic variables. Instead of calculating a weighted sum as in a normal neural network model, it calculates, for example,

number of calculations associated with each. As well, most neural fuzzy systems can be trained by going through the data once, whereas a neural network may need to be exposed to the same training data many times before it converges. Less computational resources: Neural fuzzy system is smaller in size and contains fewer internal connections than a comparable ANN, hence it is faster and use significantly less resources.
Fig 4 : Hopfield Neural Network

Step 6 : Using a pattern matching algorithm, the emotional state corresponding to the pattern is extracted, decoded and generated as final output. The result of the system is an emotional state in response to the input external event. The above procedure can be programmed in an appropriate programming language or framework and implemented on a computer. Whereupon the implementation can be embedded into a computer system with due enhancements and modifications , which could possibly give it the potential of emotional intelligence.

Offer the possibility to extract the rules: This is a major advantage over ANNs in that the rules governing a system can be communicated to the human users in an easily understandable form. Limitation of ANF systems: The greatest limitation in creating adaptive systems is known as the "Curse of Dimensionality", which is named after the exponential growth in the number of features that the model has to keep track of as the number of input attributes increases. Each attribute in the model is a variable in the system, which corresponds to an axis in a multidimensional graph that the function is mapped into. The connections between different attributes correspond to the number of potential rules in the system as given by the formula:

FUZZY LOGIC EXPLAINED Fuzzy logic enables a computer program to arrive at a decision based on multiple factors with different levels of importance. Rather than "yes" or "no" answers, this type of decision-making software deals with "maybe" and "sometimes". During the decision-making process, this type of artificial intelligence (AI) program evaluates weighted factors to reach a conclusion. Each factor that applies to the decision made by the fuzzy logic application will have a "weight", or level of importance, assigned to it. The application will compare not only the positive or negative results, but the relative importance of each result. The calculations required to process each decision can be extremely complex. AI applications performing fuzzy logic calculations often use neural networks for processing power. FUTURE DEVELOPMENTS Advantages of ANF systems: Although there are many ways to implement a Neuro-fuzzy system, the advantages described for these systems are remarkably uniform across the literature. The advantages attributed to Neuro-fuzzy systems as compared to ANNs are usually related to the following aspects: Faster to train: This is due to the massive number of connections present in the ANN, and the non-trivial

This formula becomes more complicated if there are different numbers of linguistic variables (fuzzy sets) covering each attribute dimension. Fortunately there are ways around this problem. As the neural fuzzy system is only approximating the function being modeled, the system may not need all the attributes to achieve the desired results. Another area of criticism in the Neuro-fuzzy field is related to aspects that can't be learned or approximated. One of the most known aspects here is the caveat attached to the universal approximation. In fact, the function being approximated has to be continuous; a continuous function is a function that does not have a singularity, a point where it goes to infinity. Other functions that Adaptive Neurofuzzy systems may have problems learning are things like encryption algorithms, which are purposely designed to be resistant to this type of analysis. Future developments: Predicting the future has always been hard; however for ANF technology the future expansion has been made easy because of the widespread

use of its basis technology (neural networks and fuzzy logic). Mixing of these technologies creates synergies as they remediate to each other weaknesses. ANF technology allows complex system to be grown instead of someone having to build them. One of the most promising areas for ANF systems is System Mining. There exist many cases where we wish to automate a system that cannot be systematically described in a mathematical manner. This means there is no way of creating a system using classical development methodologies (i.e. Programming a simulation.). If we have an adequately large set of examples of inputs and their corresponding outputs, ANF can be used to get a model of the system. The rules and their associated fuzzy sets can then be extracted from this system and examined for details about how the system works. This knowledge can be used to build the system directly. One interesting application of this technology is to audit existing complex systems. The extracted rules could be used to determine if the rules match the exceptions of what the system is supposed to do, and even detect fraud actions. Alternatively, the extracted model may show an alternative, and or more efficient manner of implementing the system USING AI IN BANKING AND LOANS Banks and other lending institutions are increasingly making use of AI applications in the loan pre-screening process. Instead of devoting valuable man-hours to the credit application process, lenders can employ AI programs to pre-screen applicants. By providing different levels of importance to each lending criteria, the lender ensures that the application evaluation software is able to correctly process loan applications. Each lender has different policies, and will apply unique levels of importance to each criteria. A credit score, for example, may carry a different weight depending on the actual number. A very high credit score, or a very low credit score, will affect the credit application more than an average score, in most cases. Even in cases where the credit score is very low, there are mitigating factors which may limit the impact of the score, all of which are part of the fuzzy logic algorithm. V. CONCULSION A man is a natural intelligent human being associated with physical and mental capabilities. A machine may be defined as an apparatus for applying mechanical/electrical power etc, having inter-related parts which may give useful service to human. The AI major thrust is to develop computer function normally associated with human intelligence. Intelligent devices cannot be designed unless a

fairly solid conceptual bridge has been constructed for man to put the available knowledge of brain operation to the most efficient use. The concept in this paper is an attempt at building one such conceptual bridge. Emotion is fundamental to human experience, influencing cognition, perception, and everyday tasks such as learning, communication, and even rational decision-making. However, technologists have largely ignored emotion . This research aims to restore a proper balance between emotion and cognition in the design of technologies for addressing human needs. If we can simulate a system that could reflect the emotional state in response to an event and if such a system can be embedded into a computer then we could have a computer system that has elements of the rational mind and also the emotional mind and which will eventually take the processing prowess of computers to new heights.

[5]Daniel Goleman, Emotional Intelligence, Bantam Books: New York, 1995. [6]K. Inoue, K. Kawabata and H. Kobayashi, On a decision Making System With Emotion, IEEE Int. Workshop on Robot and Human Communication, pp. 461465, 1996. [7]Joseph LeDoux, The Emotional Brain, Simon & chuster:USA,1996. REFERENCES [1] J. Bates, A. Bryan Loyall and W. Scott Reilly, An Architecture for Action, Emotion, and Social Behavior, School of Computer Science, Carnegie Mellon University, Pittsburg, PA, Technical Report CMU-CS-92-144, May 1992. [2]J. Bates, A. Bryan Loyall and W. Scott Reilly, Integrating Reactivity, Goals and Emotion in a Broad Agent, School of Computer Science, Carnegie Mellon University, Pittsburgh, PA, Technical Report CMU-CS92-142, May 1992 [3]Aard-Jan van Kesteren , Rieks op den Akker, Mannes Poel, Anton Nijholt. Simulation of emotions of agents in virtual environments using neural networks, Department of Computer Science, University of Twente, Enschede, Netherlands. [4]Magy Seif El-Nasr , John Yen, Agents, Emotional Intelligence and Fuzzy Logic , Texas A&M University.

[8]E. Masuyama, A Number of Fundamental Emotions and Their Definitions, IEEE International Workshop on Robot and Human Communication, pp. 156-161, 1994. [9]James.A.Freeman, David M.Skapura, Neural Networks Algorithms, applications and programming techniques, Pearson Education, 2007. [10]Tim Jones, Artificial Intelligence Application Programming, Second Edition, Dreamtech Press, 2006