Computational intelligence, its basics and applications

about computational intelligence

1. Computational Intelligence

2. Can Computers be Intelligent ? Who’s smarter – you or computer ? The answer is increasingly complex , and depends on definitions in flux. Today, computers can earn faster than humans, e.g., IBM’s Watson can read and remember a the research on cancer , no human can do that How its possible ?

3. Artificial Intelligence (AI) Artificial Intelligence refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions’ Most AI examples that we hear about today – from playing chess by computers to self-driving cars. “ The ability to learn/understand/deal with new situations “ “ The study of how to make computers do things at which people are doing better. [IEEE 1996] “ “[...] area of computer science that study techniques to create Intelligent systems . [Nilsson, 1998] “ “ Intelligent behavior involves perception, reasoning, learning, communicating and action in complex environments. [Nilsson, 1998 ] ”

4. History of AI

5. Computational Intelligence (CI) CI is a subset of Artificial Intelligence (AI). Ability of a computer to learn a specific task from data or experiemental observation. CI comprises of practical adaptation and self-organization concepts, paradigms, algorithms and impementations that enable or faciliate appropriate actions (intelligent behavior) in complex and changing environments -By Eberhart

6. Computational Intelligence (CI) Computational Intelligence (CI) is a dynamic branch of AI focused on developing systems that learn, adapt, and solve complex problems even in uncertain environments. Unlike traditional AI, CI mimics natural intelligence through methods like neural networks and fuzzy logic. Creating adaptive systems that learn from data, rather than being explicitly programmed for every scenario. Utilises paradigms inspired by biological and natural phenomena, such as brain function and evolution. To build machines that can "think" and make decisions with human-like intuition, but with superior speed and accuracy.

7. Human vs Machine Intelligence 1. Learning and Adaptability Human Intelligence: Humans learn from life experiences, emotions, social interactions, and abstract thought. This learning is not limited to structured data; we learn by watching others, reading between the lines, and even making mistakes. Our adaptability is highly dynamic. A child can learn a language simply by being immersed in it, without formal training. Machine Intelligence: Machines learn through data, massive volumes of structured or semi-structured information fed into models. For example, Google Translate has learned to convert languages by processing millions of translated sentence pairs. While machines can learn quickly from large datasets, their ability to generalize outside trained data is limited. Example : Google Translate doesn’t truly “understand” language the way humans do, it predicts statistically likely translations based on patterns in data.

8. Human vs Machine Intelligence 2. Reasoning and Decision-Making Human Intelligence: Humans reason using both logic and intuition. We can make decisions even with incomplete or ambiguous data, using experience, emotional context, and gut feelings. Our decision-making can also be influenced by ethical, moral, or cultural values. Machine Intelligence: Machine reasoning is strictly logical, based on the rules or data it’s been trained on. It doesn’t “understand” consequences or context unless explicitly programmed. Even complex systems like Tesla’s Autopilot follow programmed safety protocols, path planning algorithms, and sensor inputs to make decisions, but they don't truly understand what “danger” feels like. Example : Tesla’s Autopilot can change lanes and maintain distance based on input from radar, lidar, and cameras. However, it can’t yet “intuit” a reckless driver in the next lane or read subtle human cues, like a pedestrian’s body language.

9. Human vs Machine Intelligence 3. Memory and Knowledge Storage Human Intelligence: Humans have limited but flexible memory. We may forget details but remember concepts, emotions, and context. Our memories can evolve, reinterpret, or become biased over time. Machine Intelligence: Machines have virtually limitless memory , they can store and recall data precisely, repeatedly, and without fatigue. But they don’t “remember” the way we do. There’s no emotional context, no subjective filter. Siri, for instance, can recall your recent appointments instantly but doesn't remember the conversation you had last week unless explicitly stored. Example : Siri can schedule a meeting or set a reminder based on your voice command — but she doesn't understand why you’re scheduling that meeting or how you feel about it.

10. Human vs Machine Intelligence 4. Communication Human Intelligence: Humans communicate using language, gestures, emotions, and subtle non-verbal cues. A raised eyebrow or a shift in tone can completely change meaning. Machine Intelligence: Machines like Siri communicate via speech synthesis and natural language processing (NLP). They can parse commands, recognize voice inputs, and generate appropriate responses. However, they lack emotional nuance or the ability to infer intent beyond what’s been programmed or trained. Example : Siri may respond to “How’s the weather today?” with a factual weather report, but it won’t respond empathetically to “I’m feeling down today” unless it’s part of a pre-defined response set.

11. Human vs Machine Intelligence 5. Creativity and Emotions Human Intelligence: Creativity, imagination, and emotion are deeply human traits. We create art, poetry, and innovations not just from logic but from emotion and intuition. Emotions influence our decisions, sometimes irrationally but often insightfully. Machine Intelligence: Machines can generate content (like text, music, or images) using patterns in data, as seen in AI art generators or chatbots. However, this is not true creativity; it’s recombination of learned patterns. Machines do not experience emotions or consciousness. Example : ChatGPT can write a poem that mimics human style, but it doesn’t feel anything. It doesn’t create because it’s inspired — it creates because it’s prompted.

12. Why Computational Intelligence? Computational Intelligence (CI) is essential in today’s world because it enables machines to solve complex, real-world problems that are difficult or even impossible to handle using traditional rule-based programming. CI mimics aspects of human-like learning, reasoning, and adaptation, allowing systems to work in environments with uncertainty, imprecision, and incomplete data. Handles Uncertainty and Incomplete Data Traditional algorithms require precise, complete, and noise-free data. However, real-world data such as human language, sensor readings, or medical records is often noisy, vague, or missing. Example : A fuzzy logic system can diagnose a patient with “mild fever” even if there’s no exact threshold something classical binary logic can't easily do.

13. Why Computational Intelligence? Mimics Human Intelligence CI techniques like neural networks, fuzzy systems, and evolutionary algorithms are inspired by the human brain, biological evolution , and human reasoning under uncertainty . Neural Networks learn patterns like a brain. Fuzzy Logic reasons with uncertainty like humans do. Evolutionary Computation adapts and optimizes like nature does.

14. Why Computational Intelligence? Self-Learning and Adaptation CI models can learn from experience (data) and adapt to new environments over time without explicit programming. Example : Spam filters using neural networks learn new patterns as spammers evolve their tactics. Solves Complex, Non-Linear Problems Many real-world problems like medical diagnosis, image recognition, or autonomous driving are non-linear and have no simple mathematical solution. CI handles such ill-defined and multidimensional problems effectively.

15. Why Computational Intelligence? Drives Intelligent Systems CI forms the backbone of many modern AI systems and intelligent automation from personal assistants like Siri and Alexa, to self-driving cars, recommender systems, and industrial robots.

16. Myths about CI AI works like the human brain AI is a computer engineering discipline. In its current state, it consists of software tools aimed at solving problems. Some forms of AI might give the impression of being clever, but it woud be unrealistic to think that current AI is similar or equivalent to human intelligence. Although some forms of machine learning (ML) – a category of AI – have been inspired by the human brain, they are not equivalent. Image recognition technology, for example, is more accurate than most humans, but of no use when it comes to soving a math problem . The rule with AI today is that it solves one task exceedingly well, but if the conditions of the task change only a bit, it fails.

17. Myths about CI Intelligent machines learn on their own A finished ML product gives the impression that it is able to learn on its own. However, experienced human data scientists frame the problem, prepare the data, determine appropriate datasets, remove potential bias in the training data and most importantly, continually update the software to enable the integration of new knowledge and data into the next learning cycle.

18. Myths about CI AI can be 100% objective Every AI technology is based on data, rules and other kind of input from human experts. Because all humans are intrinsically biased in one way or another, so is the AI. Systems that are frequently retrained – for example , using new data from social media – are even more vulnerable to unwanted bias or international malevolent influences.

19. Myths about CI AI will only replace mundane jobs AI enables businesses to make more accurate decisions via predictions, classifications and clustering. These abilities have enabled aI-based solutions to reach deep into work environments, not only replacing mundane tasks, but also augmenting those that are more complex. Take, for example, the use of imaging AI in healthcare. A chest X-ray application based on AIcan detect diseases faster than radiologists. In the financial and insurance industry, robo advisors are being used for wealth management and fraud detection. These capabilities don’t eliminate human involvement in those tasks but will eventually limit it to observing and dealing with unusual cases. Adjust job profiles and capacity planning and offer retraining options for existing staff.

20. Myths about CI My business does not need an AI strategy Every organisation should consider the potential impact of AI on its strategy and investigate how this technology can be applied to the organisation’s business problems. In many ways, eschewing AI exploitation is the same as forgoing the next phase of automation, and could place enterprises at a competitive disadvantage. “Even if your current AI strategy is ‘no AI’ , this should be a conscious decision based on research and consideration . And –as with every other strategy – it should be periodically revisited and changed according to the organisation’s needs,” ined says.

21. Paradigms of CI CI is an umbrella term for techniques inspired by natural intelligence. It doesn’t rely on exact models but adapts through learning and experience. The three main paradigms are: Neural Networks – mimic the brain’s neuron structure Evolutionary Computation – use natural selection principles Fuzzy Systems – handle imprecise data and uncertainty

22. Neural Networks Inspired in biological neural systems. Ability to learn, memorize and generalize. H uman brain - billions of interconnected neurons; Neural network consists of layers of interconnected nodes or “artificial neurons”. A neural network contains three types of layers: an input layer, one or more hidden layers, and an output layer. Neural networks excel at learning from examples, making them extremely powerful for tasks such as image recognition, speech processing, language translation, and medical diagnosis. One of the most revolutionary advancements in this area is deep learning, which involves neural networks with many hidden layers—called deep neural networks.

23. Neural Networks Applications: Time - series approximation. Control process and optimization. Pattern recognition Pattern classification Clustering Associative memories Example: Google Translate uses deep neural networks to convert text from one language to another. Instead of translating word-for-word, the network understands entire phrases, grammar rules, and even context to generate fluent, human-like translations.

24. Evolutionary Computation Mimics processes from natural evolution Have family of algorithms inspired by the process of biological evolution. Genetic algorithms, Genetic programming, Evolutionary programming, Evolution strategies etc. The key strength of evolutionary computation lies in its ability to search through large and complex problem spaces. This paradigm has found widespread use in various domains such as robotics, automated design, game strategies, and machine learning hyperparameter tuning.

25. Evolutionary Computation Applications: Data mining Combinational optimization Fault diagnosis Classification and clustering Time series approximation Example: NASA has used evolutionary algorithms to design antennas for spacecraft. Instead of manually engineering the design, the algorithm evolves optimal shapes over generations, often producing novel and efficient structures that human engineers might not have considered.

26. Fuzzy Systems Inspired from human reasoning. Fuzzy logic allows variables to take on degrees of truth. Foundation of fuzzy systems lies in fuzzy sets. For example, in a fuzzy set representing “warm temperatures,” a temperature of 28°C might have a membership value of 0.6—meaning it’s somewhat warm—but not completely. Fuzzy inference system (FIS) processes fuzzy inputs using a set of fuzzy rules. (Eg. IF temperature is high AND humidity is high THEN fan speed is high.) These rules are evaluated using fuzzy logic operators like AND, OR, and NOT.

27. Fuzzy Systems The system then combines (aggregates) the outputs and applies a method called defuzzification to produce a crisp numerical output that can be used for control or decision-making. Fuzzy systems are especially useful in scenarios where expert knowledge can be encoded as rules, and the data is uncertain, noisy, or incomplete. Applications: Control systems Controlling lifts Classification and clustering Gear transmission and braking systems Example: Many household appliances, like washing machines and air conditioners, use fuzzy logic to optimize their operations. A fuzzy washing machine can adjust the water level, cycle duration, and spin speed based on fuzzy sensor inputs such as “lightly soiled” or “medium load,” ensuring efficient performance without requiring exact measurements.

28. Why Paradigms? Each of these three paradigms—Neural Networks, Evolutionary Computation, and Fuzzy Systems—solves a different type of problem. Neural networks learn from data and recognize patterns. Evolutionary algorithms optimize solutions in complex or unknown spaces. Fuzzy systems reason with uncertainty and linguistic information. When they are used together in hybrid systems, we can build intelligent machines that can learn, adapt, reason, and operate in real-world conditions.

29. Spam Detection – Neural Networks Neural networks are particularly effective in identifying spam emails by learning patterns in vast datasets of emails. Each email is converted into a vector of features — such as frequency of certain keywords ("free", "offer", "click here"), presence of suspicious links, sender reputation, and formatting style. The network is trained on labeled datasets using supervised learning. After training, the model can predict whether a new email is spam with high accuracy. Neural networks work well their ability to capture complex, non-linear relationships between words, context, and structure makes them ideal for dynamic spam detection — especially when spammers constantly change their strategies.

30. Medical Diagnosis – Evolutionary Computation Evolutionary algorithms start with a population of potential solutions (e.g., diagnostic rules or classification parameters). These candidates evolve over time using genetic operators like selection, crossover, and mutation Suppose the goal is to classify whether a patient has diabetes based on age, BMI, glucose level, and blood pressure. An evolutionary algorithm will generate different combinations of diagnostic thresholds and test them against patient data to see which rule-set performs best. Optimization focus: The objective could be maximizing diagnostic accuracy while minimizing false positives. Evolutionary strategies have been used in cancer detection, radiology image analysis, drug discovery, and optimizing treatment plans for chronic conditions like diabetes and hypertension.

31. Fuzzy Logic in Medical Diagnosis Instead of sharp classifications like “fever = yes or no,” fuzzy logic allows reasoning in terms like “high fever,” “moderate fever,” or “low fever,” based on continuous temperature values. A fuzzy expert system might combine multiple imprecise symptoms (mild headache, moderate fever, slight cough) and infer the possibility of conditions like viral flu or COVID-19 using a rule-based fuzzy inference engine.

32. Expert Systems R eplicates the decision-making abilities of a human expert. They are designed to solve complex problems by reasoning through bodies of knowledge, primarily represented as IF-THEN rules. An expert system typically consists of three main components: Knowledge Base Inference Engine User Interface (UI)

33. Knowledge Base This is the brain of the expert system. it contains domain-specific facts and heuristics. These are often expressed as IF-THEN rules that represent the knowledge of a human expert. Example Rule: IF patient has high fever AND cough THEN possible diagnosis is influenza. Characteristics: Static in nature (doesn't change unless updated by developers or learning modules) Must be comprehensive and accurate for the system to work well

34. Inference Engine This is the reasoning mechanism that applies logical rules to the knowledge base to deduce new information or reach conclusions. Works in two ways: Forward Chaining (Data → Conclusion): Starts from known facts and applies rules to infer conclusions. Example: From symptoms, infer disease. Backward Chaining (Goal → Supporting Data): Starts from a goal and works backward to verify supporting facts. Example: Want to confirm if it’s cancer → check symptoms → run rules.

35. User Interface (UI) This is the point of interaction between the system and the user. It allows users to input queries and receive responses, recommendations, or diagnoses. Could be graphical or textual Ensures the system is usable even for non-technical users Example: The user enters symptoms via the UI → The inference engine queries the knowledge base → It deduces a diagnosis → Presents results back through the UI.

36. How Rule-Based Systems Work ? User inputs symptoms: Fever = Yes Rash = Yes System checks the rule base: Finds a matching rule: IF fever AND rash THEN diagnosis = measles System outputs: Diagnosis: Measles

37. Medical Diagnosis System Rule 1: IF cough AND fever THEN illness = flu Rule 2: IF sore throat AND no fever THEN illness = cold Rule 3: IF high BP AND chest pain THEN illness = heart attack Scenario: Patient input: cough = yes, fever = yes Inference engine matches Rule 1 Output: illness = flu

38. Uncertainty Management An expert system consist of knowledge base that contains domain - specific facts and rules, an inference engine which applies logical rules to the knowledge base to derive conclusions and user interface to interact with users. Early systems like MYCIN (MYCIN was an early backward chaining expert system that used artificial intelligence to identify bacteria causing severe infections.) designed for diagnosing blood infections were surprisingly accurate. But real world decisions are rarely black-and-white. Experts often expresses their reasoning using phrases like ‘likely’, ‘possibly’ or ‘maybe’. Thats where the need to handle uncertainty comes in.

39. Why Uncertainty? Let’s say you’re designing a medical expert system. One rule might be: “If the patient has sore throat and a fever, they may have a strep throat “ This rule is probable but not certain. A sore throat and fever could indicate many conditions. Uncertainty arises due to: Incomplete information: The system may not have access to all patient data. Imprecise rules: Not all rules can be stated with certainty. Conflicting data: One symptom might point to multiple diagnoses. Noisy sensors or vague input: Imagine a robot using a camera to identify cracks — poor lighting might cause uncertainty in object detection.

40. Techniques to Handle Uncertainty Certainty Factors (CFs) Bayesian Reasoning Fuzzy Logic

41. Techniques to Handle Uncertainty Certainty Factors (CFs) (Used in MYCIN) “Certainty Factors were developed as a practical approach for medical expert systems. They express how confident we are in a rule’s conclusion, on a scale from -1 to +1: +1 means complete certainty 0 means no information -1 means certain the conclusion is false Its a numerical value representing the degree of belief in a conclusion or statement, ranging from -1 to +1. It's used in expert systems and other AI applications to handle uncertainty where probabilities are unknown , difficult to obtain, or too complex to compute . Positive values indicate belief, negative values indicate disbelief, and 0 indicates no information.

42. Let’s take an example: Rule: If the patient has chills and fever, they might have malaria — with CF = 0.7 If chills are observed with CF 0.8, and fever with CF 0.6, the system combines these values — often using the minimum, average, or other combining functions — to determine how confident it is in the diagnosis.” Example 2: In a rule-based expert system, a rule might state: "If the patient has a fever (evidence) and the aerobicity of the organism is anaerobic (evidence), then there is suggestive evidence (0.5) that the identity of the organism is Bacteroides (conclusion)”.The CF of 0.5 represents the degree of belief in the conclusion based on the given evidence.

43. Calculation: CFs are often calculated based on measures of belief and disbelief, derived from the evidence. Applications: Certainty factors are used in expert systems, particularly in fields like medical diagnosis (e.g., the MYCIN system). They help in reasoning with uncertain medical knowledge and making recommendations based on symptoms and test results. Limitations: While CFs are useful, they have limitations, such as the difficulty in assigning accurate certainty values and the limited numeric range.

44. 2. Bayesian Reasoning “Bayesian reasoning uses probability theory to quantify uncertainty. It answers: Given the symptoms, what is the probability the patient has a specific disease? It uses Bayes’ Theorem: P(H|E) is the probability of hypothesis H given evidence E P(E|H) is the likelihood of seeing E if H is true P(H) is the prior probability of H P(E) is the probability of observing E

45. Bayesian reasoning is a powerful technique in artificial intelligence that allows systems to reason under uncertainty by updating probabilities based on new evidence. It's particularly useful in tasks involving incomplete or noisy data, such as medical diagnosis, risk assessment, and machine learning. This approach uses Bayes' theorem to calculate the probability of a hypothesis given prior knowledge and observed evidence, enabling AI to make more informed and robust decisions. Bayes' theorem provides a mathematical framework for calculating conditional probabilities, allowing AI systems to refine their beliefs based on observed data.

46. How it Works: Prior Probability: The initial belief about a hypothesis, before considering any new evidence. Evidence: The observed data or information that becomes available. Likelihood: The probability of observing the evidence given the hypothesis is true. Posterior Probability: The updated probability of the hypothesis after considering the evidence. Bayes' theorem combines these elements to calculate the posterior probability, which represents the AI's updated belief about the hypothesis.

47. Key Concepts: Bayes' Theorem : The mathematical formula that forms the foundation of Bayesian reasoning. Prior Probability : The initial belief about a hypothesis before any evidence is considered. Posterior Probability : The updated belief about a hypothesis after considering new evidence. Bayesian Networks : A graphical model that represents probabilistic relationships between variables, allowing for efficient reasoning under uncertainty.

48. 3. Fuzzy Logic Fuzzy logic in Artificial Intelligence (AI) is a method of reasoning that mimics human decision-making processes, especially in situations involving uncertainty, imprecision, or vagueness. Unlike traditional binary logic, which operates on strict true/false (0 or 1) values, fuzzy logic allows for degrees of truth, represented by values between 0 and 1. Deals with reasoning that is approximate rather than fixed and exact.

49. Key aspects of fuzzy logic in AI include: Handling Uncertainty: Fuzzy logic is designed to manage imprecise or incomplete data, making it suitable for real-world scenarios where information may not be exact. Degrees of Truth: Instead of absolute true or false, fuzzy logic assigns a "degree of membership" to a set, indicating how strongly an element belongs to that set. For example, a temperature might be "moderately high" rather than simply "high" or "normal." Linguistic Variables: It uses linguistic terms (e.g., "hot," "cold," "fast," "slow") to represent concepts, which are then mapped to fuzzy sets with associated membership functions. This allows AI systems to process and reason with human-like qualitative descriptions. Fuzzy Rule-Based Systems: Fuzzy logic systems often employ a set of "if-then" rules that incorporate these linguistic variables and fuzzy sets to make decisions or control outputs.

50. Applications: Control Systems: For example, controlling temperature in HVAC systems, optimizing washing machine cycles, or managing anti-lock braking systems in cars. Decision Making: Assisting in medical diagnosis, financial analysis, and expert systems. Image Processing: Enhancing images and pattern recognition. Natural Language Processing: Understanding and interpreting nuanced human language.

51. Assignment Collect research papers on any of Neural networks, Fuzzy logic and Evolutionary computation. Understand how they have implemented these in the methodology. Each of you prepare a 5 minute presentation with the description of how this technology is used in that methodology mentioned in the paper. Presentation on 13/08/2025