(Aktu Btech) Machine Learning Techniques Important Unit-1 Introduction

B.Tech AKTU Quantum Book will take you on a journey through the world of Machine Learning Techniques. Access critical notes, frequently asked questions, and important insights to grasp this innovative profession. Unit-1 Introduction

Dudes 🤔.. You want more useful details regarding this subject. Please keep in mind this as well.

Important Questions For Machine Learning Techniques:
*Quantum               *B.tech-Syllabus
*Circulars                *B.tech AKTU RESULT
* Btech 3rd Year     * Aktu Solved Question Paper

Q1. Define the term learning What are the components of a learning system ?

Ans.

• 1. Learning is the behaviour change a subject makes in response to a situation after experiencing it repeatedly, provided that the behaviour change cannot be explained by the individual’s innate reaction inclinations, biases, or transitory moods.
• 2. The idea behind a learning agent is that it consists of a performance element that chooses what actions to take and a learning element that improves the performance element’s decision-making.
• 3. The design of a learning element is affected by three major issues:
  • a. Components of the performance element. 
  • b. Feedback of components. 
  • c. Representation of the components.   

The important components of learning are:  

• 1. Acquisition of new knowledge: 
  • a. Acquiring new knowledge is one component of learning. 
  • b. Basic data collecting is simple for computers but challenging for humans. 
• 2. Problem solving: 
  • The third component of learning is problem solving, which is essential for both integrating new knowledge into the system and deducing new information when required facts are not supplied.

Q2. Describe briefly reinforcement learning ? 

Ans.

• 1. Reinforcement learning is the study of how a machine may learn to optimise its behaviour in the face of rewards and punishments.
• 2. Reinforcement learning algorithms that are closely related to methods of dynamic programming, which is a generic approach to optimum control, have been created.
• 3. Reinforcement learning phenomena have been observed in animal behaviour research as well as neurobiological studies of neuromodulation and addiction.

• 4. The objective of reinforcement learning is to use observed incentives to learn the best environmental policy.
• 5. An optimal strategy is one that maximises the overall expected payoff.
• 6. Without any input on what is excellent and bad, the agent will be unable to decide which action to make.
• 7. The agents must understand that when it wins, something good has occurred, and when it loses, something awful has occurred.
• This type of feedback is known as a reward or reinforcement.
• 9. Reinforcement learning is extremely useful in the realm of robotics, where the tasks to be done are frequently complicated enough to resist encoding as programmes and there is no available training data.
• 10. The robot’s duty is to determine which actions are appropriate in a given situation through trial and error (or success).
• 11. Humans learn in many ways that are extremely similar.
• 12. For example, when a child learns to walk, he or she normally does so without instruction, but rather through reinforcement.
• 13. Successful attempts at work are rewarded with forward movement, whereas poor attempts are punished with painful falls.
• 14. Positive and negative reinforcement are also key components in good school and sports learning.
• 15. Reinforcement learning is the only practical approach to train a programme to perform at high levels in many complicated domains.

Q3. Describe well defined learning problems role’s in machine learning. 

Ans. Well defined learning problems role’s in machine learning : 

• 1. Learning to recognize spoken words: 
  • a. Effective voice recognition systems use some sort of machine learning.
  • b. The SPHINX system, for example, develops speaker-specific techniques for detecting primitive sounds (phonemes) and words from the observed speech signal.
  • c. Neural network learning methods and hidden learning methods Markov models are useful for automatically tailoring to certain speakers, vocabularies, microphone characteristics, background noise, and so on.
• 2. Learning to drive an autonomous vehicle: 
  • a. Machine learning techniques have been used to train computer-controlled vehicles to steer accurately when driving on various types of roads.
  • b. For example, the ALYINN system, among other cars, has used its learnt tactics to drive independently at 70 miles per hour for 90 miles on public roadways.
• 3. Learning to classify new astronomical structures: 
  • a. Machine learning approaches have been used to learn general regularities implicit in a number of huge databases.
  • b. For example, NASA employed decision tree learning algorithms to learn how to identify celestial objects from the second Palomar Observatory Sky Survey.
  • c. This method is used to classify all objects in the Sky Survey, which contains three terabytes of visual data.
• 4. Learning to play world class backgammon: 
  • a. Machine learning algorithms are at the heart of the most successful computer systems for playing games like backgammon.
  • b. For example, TD-GAMMON, the world’s top computer backgammon software, learnt its approach by playing over one million practise games against itself.

Q4. Write short note on Artificial Neural Network (ANN).

Ans.

• 1. Artificial Neural Networks (ANN) or neural networks are computational algorithms designed to mimic the behaviour of biological systems made up of neurons.
• 2. ANNs are computer models inspired by the central nervous systems of animals.
• 3. It is capable of both machine learning and pattern recognition.
• 4. A neural network is a graph that is directed. It is made up of nodes connected by arcs that resemble neurons in the biological comparison.
• It is related to dendrites and synapses. Each arc is assigned a weight at each node.
• 6. A neural network is a computer learning technique that is based on a human neuron model. Millions of neurons make up the human brain.
• 7. It communicates and processes electrical and chemical impulses.
• 8. These neurons are linked together by a structure known as synapses. Synapses allow neurons to communicate with one another.
• 9. An Artificial Neural Network is a method of processing information. It functions similarly to how the human brain processes information.
• 10. ANN is made up of a large number of interconnected processing units that work together to process data. They also get meaningful results out of it.

Q5. What are the various clustering techniques ?

Ans.

• 1. Clustering techniques are used for combining observed examples into clusters or groups which satisfy two following main criteria : 
  • a. Each group or cluster is homogenous, which means that examples from the same group are similar to one another.
  • b. Each group or cluster should be distinct from the others, i.e., instances from one cluster should be distinct from examples from other clusters.
• 2. Depending on the clustering techniques, clusters can be expressed in different ways: 
  • a. The detected clusters may be exclusive, so that each example belongs to only one.
  • b. They may overlap, i.e., an example may belong to more than one cluster.
  • c. They may be probabilistic, in which case one example is assigned to each cluster with a specific probability.
  • d. Clusters may be hierarchical in nature.

Major classifications of clustering techniques are: 

• a. Once a criterion function is chosen, clustering becomes a well-defined discrete optimisation issue. We look for partitions of the set of samples that deviate from the criterion function.
• b. The sample set is finite; there are only a limited number of partitions conceivable.
• c. Exhaustive enumeration can always solve the clustering problem.

1. Hierarchical clustering: 

• a. This approach works by clustering data objects into a tree.
• b. This method is further characterized based on whether the hierarchical breakdown is created from the bottom up (merging) or the top down (splitting).

Following are the two types of hierarchical clustering:

• a. Agglomerative hierarchical clustering: Each object is initially placed in its own cluster as part of this bottom-up approach, which then combines these atomic clusters into ever larger clusters until every object is contained within a single cluster. 
• b. Divisive hierarchical clustering:
  • i. By beginning with all items in one cluster, this top-down technique does the opposite of the agglomerative strategy.
  • ii. It breaks the cluster up into smaller and smaller fragments until each item can stand alone and create a cluster.

2. Partitional clustering: 

• a. A set of initial partitions, each of which represents a cluster, are first created by this method.
• b. The clusters are generated to optimise an objective partition criterion such as a dissimilarity function based on distance such that the objects within a cluster are similar whereas the objects of separate clusters are dissimilar. 

Following are the types of partitioning methods: 

• a. Centroid based clustering:
  • i. In this, a group of objects is divided into a number of clusters based on an input parameter, producing intra cluster similarity that is high but inter cluster similarity that is low.
  • ii. The mean value of the objects in the cluster, which can be thought of as the cluster’s centroid or centre of gravity, is used to measure cluster similarity.
• b. Model-based clustering: This method hypothesizes a model for each of the cluster and finds the best fit of the data to that model.

Q6. What are the advantages and disadvantages of decision tree method ? 

Ans. Advantages of decision tree method are: 

• 1. Decision trees can produce rules that are easy to understand.
• 2. Classification is carried out using decision trees without the need for calculation.
• 3. Both continuous and categorical variables can be handled by decision trees.
• 4. Decision trees clearly identify the fields that are crucial for categorization or prediction.

Disadvantages of decision tree method are: 

• 1. Estimation assignments where the objective is to anticipate the value of a continuous attribute are less suitable for decision trees.
• 2. In classification problems with multiple classes and a limited number of training samples, decision trees are vulnerable to errors.
• 3. It takes a lot of computing power to train a decision tree. Each candidate splitting field at each node must first be sorted in order to determine which split is optimal.
• 4. Combinations of fields are employed in decision tree methods, and the best combining weights must be found. Due to the need to create and evaluate numerous candidate sub-trees, pruning algorithms can also be costly.

Machine Learning Techniques Btech Quantum PDF, Syllabus, Important Questions

Machine Learning Techniques Quantum PDF | AKTU Quantum PDF:

AKTU Important Links | Btech Syllabus

Important Links-Btech (AKTU)