Untitled Quiz

Unit 1

• Artificial Intelligence (AI) is a multidisciplinary field of computer science that seeks to create machines capable of performing tasks that normally require human intelligence.
• Thinking Humanly: AI here mimics how humans think, focusing on learning, reasoning, and problem-solving. The aim is to make machines act intelligently by simulating human thought processes.
• Thinking Rationally: This approach centers on logical and rational thinking. AI systems follow formal rules and laws of thought to make deductions and inferences, arriving at conclusions through structured reasoning.
• Acting Humanly: AI, in this approach, aims to act like a human. Think of it like a machine taking the Turing Test, interacting with people in a way that's natural and human-like, involving language processing, knowledge representation, and learning.
• Acting Rationally: AI is concerned with optimal behavior, not necessarily imitating humans. It's about making machines act in the best way possible based on a set of principles, maximizing success or utility in a given situation.
• Natural Language Processing (NLP): AI used in NLP to enable machines to understand, interpret, and generate human-like text; examples include chatbots, language translation services, and sentiment analysis.
• Machine Learning (ML): A subset of AI, involves the development of algorithms that allow systems to learn and improve from experience. ML is used in image recognition, recommendation systems, and fraud detection.
• Computer Vision: AI is employed in computer vision to enable machines to interpret and make decisions based on visual data. Examples include facial recognition, object detection, and autonomous vehicles.
• Robotics: AI is crucial in robotics, where it enables machines to perceive their environment, make decisions, and execute tasks. This is seen in industrial robots, drones, and even in healthcare, where robots assist in surgeries.
• Expert Systems: AI is used to develop expert systems that emulate the decision-making ability of a human expert in a specific domain; examples include diagnosis in medicine and troubleshooting in technical support

Unit 2

• Knowledge Representation is a crucial aspect of artificial intelligence. It involves structuring information in a way that allows a computer program to understand, interpret, and manipulate knowledge about the world.
• Declarative Knowledge: Represents factual information; "knowing that."
• Procedural Knowledge: Represents knowledge about how to do something; involves a sequence of steps.
• Semantic Knowledge: Understanding the meaning of words, concepts, and relationships between them.
• Episodic Knowledge: Knowledge of particular instances or occurrences.
• Conceptual Knowledge: Understanding abstract ideas, principles, or categories.
• Tactical Knowledge: Strategies and plans for achieving goals.
• Heuristic Knowledge: Rules of thumb or guidelines used to solve problems based on experience and judgment.
• Meta-Knowledge: Knowledge about knowledge; how information is organized, evaluated, and used.
• Domain-Specific Knowledge: Knowledge specific to a particular field.

Unit 3

• Statistical learning in AI involves constructing models based on data to make predictions or decisions without explicit programming.
• Bayesian Learning Methods: These methods involve using observations to update a prior distribution over hypotheses, making optimal predictions in the presence of uncertainty and noise.
• Learning Probability Models: Statistical learning often involves using probability models like Bayesian networks to represent and reason about uncertain knowledge effectively.
• Expectation Maximization (EM) Algorithm: A powerful statistical technique in AI; useful for estimating parameters in probabilistic models with incomplete data or missing values; alternating between Expectation (E) and Maximization (M) steps.
• Temporal Difference (TD) Learning: A central concept in reinforcement learning to learn from incomplete sequences. It bootstraps estimates based on subsequent estimates; TD Error is the difference between the estimated value of the current state and the combined reward and estimated value of the next state.

Additional Details

• Simple Reflex Agent: Selects actions based on the current percept, ignoring past or future.
• Model-Based Agent: Maintains an internal model of the environment and makes decisions based on this model.
• Utility-Based Agent: Makes decisions by assessing the value (or utility) of actions in order to achieve goals.
• Goal-Based Agent: Operates by pursuing predefined goals in a given environment.
• Learning Agent: A learning agent learns and makes better decisions over time. This involves components capable of perception, learning, and decision making.
• Depth-First Search (DFS): Explores a path as deeply as possible before backtracking. Suitable for scenarios where the solution is deep.
• Breadth-First Search (BFS): Explores all nodes at the current depth before moving to the next. Suitable for scenarios where the solution is close to the root.
• Uniform Cost Search (UCS): Expands nodes in the order of their cumulative cost from the initial state to the current state. Optimal but may require more memory (especially in large spaces).
• Iterative Deepening Depth-First Search (IDDFS): A search algorithm that combines the advantages of both depth-first and breadth-first searches; it retains the memory efficiency of depth-first while ensuring completeness by gradually increasing the depth limit.
• Bidirectional Search: Simultaneously explores from the start and goal states, meeting in the middle. Suitable for large search spaces where the goal is well-defined.
• Thinking Rationally: AI systems aiming to emulate human thought processes and reasoning, using rules, logic, and symbolic representations.
• Thinking Humanly: AI systems that strive to mimic human cognitive functions (perception, learning, problem-solving).
• Acting Humanly: AI systems that produce behaviour indistinguishable from humans.
• Acting Rationally: Creating AI systems to make optimal decisions based on available information to achieve specific goals.
• Parametric Models: Make explicit assumptions about the underlying relationship between variables; the number of parameters is fixed. Simple and easier to understand in certain circumstances.
• Non-parametric Models: Flexible and can adapt to complex relationships without assuming a specific functional form in the underlying data; often used when the relationship is unknown or complex and doesn't allow for formal modelling.
• Naive Bayes: A simple probabilistic classifier based on Bayes' Theorem; assumes features are independent given the class. Relatively efficient and suited to high-dimensional data.
• Hidden Markov Models (HMMs): Models that represent systems with hidden states. The system's internal processes (states) aren't directly observed but can be inferred from observable output sequences (e.g., speech recognition, biological process).
• Reinforcement Learning(RL): Learning through interactions with environments, receiving rewards/penalties for actions.
• Q-Learning: A model-free RL algorithm; learns an action-value function and estimates the optimal action-value in a given state; useful when the environment model is unknown.
• Association Rule Mining: Discovers relationships or patterns between items in a dataset; uses metrics like support, confidence, and lift; commonly uses in business (market basket analysis)
• Supervised Learning: Learns from examples with known inputs and outputs.
• Unsupervised Learning: Learns from unlabeled data, finding patterns and relationships.
• Linear Regression: Predicts a continuous variable based on one or more independent variables (assumes a linear relationship between variables).
• Logistic Regression: Predicts the probability of a binary outcome based on one or more independent variables (used for classification problems.
• K Nearest Neighbours (KNN): A simple instance-based learning machine learning algorithm which assumes there's a relationship between the new data and the existing data; frequently used in classification.
• Decision Trees: Learns a hierarchical representation of decisions based on attribute values; generally suited for classification problems.
• Random Forests: A composite of decision trees; increases the robustness and accuracy of the classification algorithms.
• Gradient Descent: An iterative optimization algorithm used to minimize the cost function in machine learning models.