Machine Learning
105602Module 1: Introduction
1(a). A model has 98% training accuracy and 72% test accuracy. Which of the following is the most likely explanation? (i) High bias (ii) Overfitting (iii) Underfitting (iv) Insufficient features20262m
Module 1: Introduction
View this question on its own page →A model has 98% training accuracy and 72% test accuracy. Which of the following is the most likely explanation? (i) High bias (ii) Overfitting (iii) Underfitting (iv) Insufficient features
Worked SolutionAnswer: (ii) Overfitting
Training accuracy is high (98%) but test accuracy is much lower (72%), indicating the model fits training-specific patterns and generalizes poorly. This is a typical high-variance/overfitting pattern.
1(b). Which statement about hypothesis space is correct? (i) A larger hypothesis space always gives better generalization (ii) Hypothesis space is unrelated to inductive bias (iii) A smaller hypothesis space always gives lower training error (iv) A larger hypothesis space can increase the risk of overfitting20262m
Module 1: Introduction
View this question on its own page →Which statement about hypothesis space is correct? (i) A larger hypothesis space always gives better generalization (ii) Hypothesis space is unrelated to inductive bias (iii) A smaller hypothesis space always gives lower training error (iv) A larger hypothesis space can increase the risk of overfitting
Worked SolutionAnswer: (iv) A larger hypothesis space can increase the risk of overfitting
A larger hypothesis space can represent more patterns, lowering training error, but may also fit noise. Generalization depends on data, regularization, model capacity and inductive bias; larger is not automatically better.
2(a). Explain the bias-variance trade-off and discuss how model complexity affects underfitting and overfitting.20267m
Module 1: Introduction
View this question on its own page →Explain the bias-variance trade-off and discuss how model complexity affects underfitting and overfitting.
Worked SolutionBias–variance trade-off
Bias is error from overly restrictive assumptions; variance is sensitivity to the particular training sample. Expected prediction error can be viewed as irreducible noise plus squared bias plus variance.
- Underfitting: model too simple; high bias; poor training and test performance.
- Overfitting: model too complex; low training error but high variance and poor test performance.
- Balanced model: captures real structure while generalizing to unseen data.
Increasing complexity often lowers training error but may eventually raise test error. Use cross-validation, regularization, more representative data, feature selection or early stopping to choose complexity.
MermaidDiagramRendering diagram…3(a). A test is applied to 1000 patients: 200 are sick, and 800 are healthy. It correctly detects 160 sick patients, while 40 healthy patients test positive. Construct the confusion matrix and calculate accuracy, precision, recall, and F1-score.20267m
Module 1: Introduction
View this question on its own page →A test is applied to 1000 patients: 200 are sick, and 800 are healthy. It correctly detects 160 sick patients, while 40 healthy patients test positive. Construct the confusion matrix and calculate accuracy, precision, recall, and F1-score.
Worked SolutionConfusion matrix and metrics
There are 200 sick patients and 800 healthy patients. True positives = 160; false negatives = 200 − 160 = 40. False positives = 40; true negatives = 800 − 40 = 760.
Actual / Predicted Positive Negative Total Sick TP = 160 FN = 40 200 Healthy FP = 40 TN = 760 800 Total 200 800 1000 - Accuracy = ((TP+TN)/N=(160+760)/1000=\mathbf{92%})
- Precision = (TP/(TP+FP)=160/200=\mathbf{80%})
- Recall = (TP/(TP+FN)=160/200=\mathbf{80%})
- F1 = (2PR/(P+R)=\mathbf{80%})
Accuracy measures overall correctness; precision measures reliability of positive predictions; recall measures how many actual sick patients were detected.
Q1(c). Define specific boundary.20232m
Q1(e). List the basic design issues to machine learning.20232m
Module 1: Introduction
View this question on its own page →List the basic design issues to machine learning.
Q1(g). What is the essential difference between analytical and inductive learning methods?20232m
Module 1: Introduction
View this question on its own page →What is the essential difference between analytical and inductive learning methods?
Q1(h). What do you understand by noise in the data? How it affect the result?20232m
Module 1: Introduction
View this question on its own page →What do you understand by noise in the data? How it affect the result?
Q2(a). “Machine learning cannot solve every problem”. Is this statement correct? Give justification to your answer with proper explanation.20237m
Module 1: Introduction
View this question on its own page →“Machine learning cannot solve every problem”. Is this statement correct? Give justification to your answer with proper explanation.
Q4(a). What is the necessity of feature transformation in learning?20237m
Module 1: Introduction
View this question on its own page →What is the necessity of feature transformation in learning?