AKTU Machine Learning Techniques · Important Questions 2026Source: ryzenstudy.com PYQs
BCS055 Machine Learning Techniques Important Questions 2026 (AKTU)
Repeat analysis + most-asked questions, unit-wise
What we did: built from all 5 AKTU Machine Learning Techniques papers (2021-22–2025-26): 124 questions grouped into 74 distinct questions, 23 of which were asked in more than one paper.
- of marks were repeat questions
- 53%
- of marks were repeat questions
- questions asked in 2+ papers
- 23
- questions asked in 2+ papers
- papers analysed (2021-22–2025-26)
- 5
- papers analysed (2021-22–2025-26)
- units, mapped to current syllabus
- 5
- units, mapped to current syllabus
How much of the paper repeats?
- Repeats are real: when we checked each newer BCS055 paper against the papers before it, on average 53% of its marks were questions already asked in an earlier paper.
- 2023-24 paper: 69% of marks were repeats (10 of 16 repeated questions were in our top list).
- 2024-25 paper: 43% of marks were repeats (5 of 9 repeated questions were in our top list).
- 2025-26 paper: 48% of marks were repeats (8 of 10 repeated questions were in our top list).
- Topic weightage is a guide, not a guarantee: our Tier 1 + Tier 2 topics covered 58% of the next paper's marks — picking the same number of topics at random would have covered about 48%.
Across 10 subjects and 23 papers we tested, 43% of marks repeated earlier questions on average.
1.Unit-wise most-asked questions
Ranked by how many papers asked them, then by how recently. Click a paper to open it.
Unit 1:
21% of past marks| Question | Asked |
|---|---|
| Explain types of learning in Machine Learning with suitable examples. | Asked 4× |
| What is a well-defined learning problem? | Asked 3× |
| Describe the steps involved in designing a machine learning system. Illustrate each step with a real-world example. | Asked 2× |
| Compare regression, classification and clustering in machine learning along with suitable real-life applications. | Asked 2× |
Unit 2: SUPPORT VECTOR MACHINE
18% of past marks| Question | Asked |
|---|---|
| Explain the Maximum Likelihood and Least Squared Error hypotheses with an example. | Asked 2× |
| Explain hyperplane (decision boundary) in SVM. Categorize various popular kernels associated with SVM. | Asked 2× |
| Explain the "Concept Learning" task, giving an example. | Asked 2× |
| Discuss support vectors in SVM. | Asked 2× |
Unit 3:
24% of past marks| Question | Asked |
|---|---|
| What is overfitting in decision tree learning, and how can it be avoided? | Asked 4× |
| Explain the k-Nearest Neighbour (k-NN) algorithm and its limitations. | Asked 3× |
| Explain the ID3 algorithm for Decision Tree learning. | Asked 2× |
| Compare and contrast Information Gain, Gain Ratio, and Gini Index in detail. | Asked 2× |
Unit 4:
18% of past marks| Question | Asked |
|---|---|
| Explain the architecture of Convolutional Neural Networks (CNN). | Asked 4× |
| Derive the Backpropagation algorithm for training neural networks. | Asked 4× |
| What is gradient descent delta rule? | Asked 2× |
| Discuss applications of Deep Learning with suitable case studies. 7 marks2025-26 Q5(b) | Asked 1× |
Unit 5: GENETIC ALGORITHMS
19% of past marks| Question | Asked |
|---|---|
| Describe Q-Learning algorithm and its applications. | Asked 4× |
| Explain Genetic Algorithm cycle with suitable example. | Asked 4× |
| Explain Reinforcement Learning with Markov Decision Process. | Asked 2× |
| Explain the concept of Reinforcement Learning (RL). Discuss how it differs from supervised and unsupervised learning. | Asked 2× |
2.Topic weightage: which topics carry the marks
Blue = Tier 1 + Tier 2 (topics that carried ~80% of past marks). Grey = the rest. It shows where marks have gone, not a promise of what comes next.
Types of learning · U1
7.8% (cum 8%)
Issues in decision trees · U3
7.3% (cum 15%)
Bayes theorem and Naive Bayes · U2
6.7% (cum 22%)
GA components and cycle · U5
5.1% (cum 27%)
Multilayer networks and backpropagation · U4
5.1% (cum 32%)
Q-learning · U5
5.1% (cum 37%)
CNN layers and architecture · U4
4.7% (cum 42%)
Linear regression · U1
4.4% (cum 46%)
History, applications, ML vs AI/DS · U1
4.1% (cum 50%)
Instance-based and lazy learning · U3
3.9% (cum 54%)
Reinforcement learning basics · U5
3.7% (cum 58%)
Perceptron and activation · U4
3.6% (cum 62%)
k-Nearest Neighbour · U3
3.4% (cum 65%)
Well-defined problems and design · U1
3.4% (cum 68%)
Concept learning · U2
3.3% (cum 72%)
Entropy and information gain · U3
3.3% (cum 75%)
Decision tree and ID3 · U3
3.1% (cum 78%)
SVM introduction and properties · U2
2.9% (cum 81%)
Hyperplane and decision surface · U2
2.8% (cum 84%)
Bayesian belief networks · U2
1.6% (cum 85%)
Markov decision process · U5
1.6% (cum 87%)
Inductive bias and inference · U3
1.5% (cum 88%)
Gradient descent and delta rule · U4
1.5% (cum 90%)
Logistic regression · U1
1.4% (cum 91%)
Locally weighted regression · U3
1.4% (cum 93%)
RL applications · U5
1.3% (cum 94%)
Deep learning case studies · U4
1.1% (cum 95%)
Genetic algorithm basics · U5
1.1% (cum 96%)
Crossover and mutation · U5
1.1% (cum 97%)
EM algorithm · U2
1% (cum 98%)
Generalization · U4
0.8% (cum 99%)
SOM algorithm · U4
0.5% (cum 100%)
Deep learning introduction · U4
0.4% (cum 100%)
| Topic | Papers | Share |
|---|---|---|
| Types of learning · Unit 1 · Tier 1 | 4/5 | 7.8% |
| Issues in decision trees · Unit 3 · Tier 1 | 4/5 | 7.3% |
| Bayes theorem and Naive Bayes · Unit 2 · Tier 1 | 5/5 | 6.7% |
| GA components and cycle · Unit 5 · Tier 1 | 5/5 | 5.1% |
| Multilayer networks and backpropagation · Unit 4 · Tier 1 | 4/5 | 5.1% |
| Q-learning · Unit 5 · Tier 1 | 4/5 | 5.1% |
| CNN layers and architecture · Unit 4 · Tier 1 | 4/5 | 4.7% |
| Linear regression · Unit 1 · Tier 1 | 3/5 | 4.4% |
| History, applications, ML vs AI/DS · Unit 1 · Tier 1 | 3/5 | 4.1% |
| Instance-based and lazy learning · Unit 3 · Tier 2 | 3/5 | 3.9% |
| Reinforcement learning basics · Unit 5 · Tier 2 | 2/5 | 3.7% |
| Perceptron and activation · Unit 4 · Tier 2 | 3/5 | 3.6% |
| k-Nearest Neighbour · Unit 3 · Tier 2 | 3/5 | 3.4% |
| Well-defined problems and design · Unit 1 · Tier 2 | 3/5 | 3.4% |
| Concept learning · Unit 2 · Tier 2 | 3/5 | 3.3% |
| Entropy and information gain · Unit 3 · Tier 2 | 3/5 | 3.3% |
| Decision tree and ID3 · Unit 3 · Tier 2 | 3/5 | 3.1% |
| SVM introduction and properties · Unit 2 · Tier 2 | 3/5 | 2.9% |
| Hyperplane and decision surface · Unit 2 · Tier 3 | 3/5 | 2.8% |
| Bayesian belief networks · Unit 2 · Tier 3 | 2/5 | 1.6% |
| Markov decision process · Unit 5 · Tier 3 | 2/5 | 1.6% |
| Inductive bias and inference · Unit 3 · Tier 3 | 2/5 | 1.5% |
| Gradient descent and delta rule · Unit 4 · Tier 3 | 3/5 | 1.5% |
| Logistic regression · Unit 1 · Tier 3 | 2/5 | 1.4% |
| Locally weighted regression · Unit 3 · Tier 3 | 1/5 | 1.4% |
| RL applications · Unit 5 · Tier 3 | 1/5 | 1.3% |
| Deep learning case studies · Unit 4 · Tier 3 | 1/5 | 1.1% |
| Genetic algorithm basics · Unit 5 · Tier 3 | 1/5 | 1.1% |
| Crossover and mutation · Unit 5 · Tier 3 | 1/5 | 1.1% |
| EM algorithm · Unit 2 · Tier 3 | 1/5 | 1% |
| Generalization · Unit 4 · Tier 3 | 1/5 | 0.8% |
| SOM algorithm · Unit 4 · Tier 3 | 1/5 | 0.5% |
| Deep learning introduction · Unit 4 · Tier 3 | 1/5 | 0.4% |
3.Paper pattern (2025-26)
| Section | Questions | Rule |
|---|---|---|
| A | 7 × 2 marks | Attempt all |
| B | 5 × 7 marks | Attempt any 3 |
| C | 10 × 7 marks | Attempt any 1 |
4.In the syllabus, never asked yet
Low priority, but read them once — a new paper can always start here.
- ML approaches overview · U1
- Issues in machine learning · U1
- Bayes optimal classifier · U2
- SVM kernels · U2
- Radial basis function networks · U3
- Deep Q learning · U5
- Genetic programming and evolution · U5
5.Study order
- 1Types of learning (Unit 1, Tier 1)
- 2Issues in decision trees (Unit 3, Tier 1)
- 3Bayes theorem and Naive Bayes (Unit 2, Tier 1)
- 4GA components and cycle (Unit 5, Tier 1)
- 5Multilayer networks and backpropagation (Unit 4, Tier 1)
- 6Q-learning (Unit 5, Tier 1)
- 7CNN layers and architecture (Unit 4, Tier 1)
- 8Linear regression (Unit 1, Tier 1)
- 9History, applications, ML vs AI/DS (Unit 1, Tier 1)
- 10Instance-based and lazy learning (Unit 3, Tier 2)
- 11Reinforcement learning basics (Unit 5, Tier 2)
- 12Perceptron and activation (Unit 4, Tier 2)
- 13k-Nearest Neighbour (Unit 3, Tier 2)
- 14Well-defined problems and design (Unit 1, Tier 2)
- 15Concept learning (Unit 2, Tier 2)
- 16Entropy and information gain (Unit 3, Tier 2)
- 17Decision tree and ID3 (Unit 3, Tier 2)
- 18SVM introduction and properties (Unit 2, Tier 2)
FAQ
Which units of BCS055 carry the most marks?
Unit 3 (, 24% of past marks) and Unit 1 (, 21% of past marks).
Do AKTU Machine Learning Techniques questions repeat?
Yes. On average 53% of the marks in each newer BCS055 paper came from questions already asked in an earlier paper, and 23 questions have appeared in more than one paper.
How is this list made?
Every question from the 5 papers (2021-22–2025-26) was mapped to the current AKTU syllabus and grouped with the same question from other years. Questions are ranked by how many papers they appeared in, then by how recently.
Study this subject
Note: this is based on past AKTU papers, not a guarantee. Questions are mapped to the current syllabus; topics that are no longer in the syllabus are left out.