Top 10 ML MCQs on SVM Concepts (2025 Edition)

Top 10 New MCQs on SVM Concepts (2025 Edition)

1. Which of the following best describes the margin in an SVM classifier?

A. Distance between two closest support vectors
B. Distance between support vectors of opposite classes
C. Distance between decision boundary and the nearest data point of any class
D. Width of the separating hyperplane

Answer: C

Explanation: The margin is the perpendicular distance from the decision boundary to the closest data point (this data point is called support vector) of any class. SVM aims to maximize this margin.

---

2. In soft-margin SVM, the penalty parameter C controls what?

A. The kernel function complexity
B. The balance between margin width and classification errors
C. The learning rate during optimization
D. The dimensionality of transformed space

Answer: B

Explanation: Parameter C determines how much misclassification is tolerated. A large C → fewer violations, smaller margin; a small C → allows more violations, larger margin.

A Soft Margin SVM is a type of Support Vector Machine that allows some misclassification or margin violations in order to achieve better generalization when data is not perfectly linearly separable. In simple words, Soft Margin SVM finds the best possible separating hyperplane that balances maximum margin and minimum classification error.

Overfitting vs Underfitting: Large C - risk of overfitting, small C - risk of underfitting.

---

3. Which of the following statements about the kernel trick in SVM is true?

A. It explicitly computes higher-dimensional feature mappings
B. It avoids computing transformations by using inner products in the feature space
C. It can only be applied to linear SVMs
D. It reduces the number of support vectors required

Answer: B

Explanation: The kernel trick enables SVMs to work in high-dimensional spaces without explicitly computing the transformed features. It uses kernel functions to calculate inner products in that space, making non-linear separation computationally efficient.

What is kernel trick?

A Soft Margin SVM is a type of Support Vector Machine that allows some misclassification or margin violations in order to achieve better generalization when data is not perfectly linearly separable. In simple words: Soft Margin SVM finds the best possible separating hyperplane that balances maximum margin and minimum classification error.

---

4. Which step is unique to non-linear SVMs?

A. Feature normalization
B. Slack variable introduction
C. Kernel trick application
D. Margin maximization

Answer: C

Explanation: The kernel trick allows mapping non-linearly separable data into a higher-dimensional space without explicitly computing transformations — used only in non-linear SVMs.

Why kernel trick is unique to non-linear SVMs only?

The kernel trick is used only in non-linear SVMs because linear SVMs already work directly in the input space — there’s no need to map the data to a higher-dimensional space to make it linearly separable.

The kernel trick is used only in non-linear SVMs because linear SVMs already find a separating hyperplane in the original space, whereas non-linear data needs implicit mapping to higher dimensions.

In other words:

• Linear SVM → data is already separable by a straight hyperplane → no mapping needed.

• Non-linear SVM → data is not separable in the original space → kernel trick is applied to find a linear boundary in a higher-dimensional feature space.

---

5. If the data is perfectly linearly separable, what is the ideal value of C?

A. Very small (close to 0)
B. Moderate (around 1)
C. Very large (→ ∞)
D. Exactly equal to margin value

Answer: C

Explanation: A large C ensures no margin violations (hard-margin SVM), which is suitable when data is perfectly separable.

What are the problems with very large C value?

The SVM with very large C value will try to classify every training point correctly, no matter how narrow or overfitted the margin becomes. This lead to;

• Very small margin (The model sacrifices margin width to perfectly fit all training points.)
• Overfitting risk (It fits even noisy or outlier points, harming generalization to new data.)
• Hard margin behavior (the Soft Margin SVM effectively becomes a Hard Margin SVM, demanding perfect separation.)
• Unstable model (Small changes in data may cause large shifts in the decision boundary)

---

6. Which optimization problem does SVM solve during training?

A. Minimization of loss function via gradient descent
B. Maximization of likelihood function
C. Quadratic optimization with linear constraints
D. Linear programming without constraints

Answer: C

Explanation: SVM training is a quadratic optimization problem where a convex quadratic function is minimized under linear constraints.

Why the optimization problem is "Quadratic optimization with linear constraints"?

During SVM training, the goal is to find the best separating hyperplane between two classes. That means we need to find w and b such that the classifier f(x) = w⋅x + b correctly separates the classes with the largest possible margin.

The objective is to maximize the margin (2/||w||). To maximize this margin, we need to minimize L2 norm squared of the weight vector w (i.e. ∣∣w∣∣²). So the optimization problem becomes a quadratic function of w (i.e., ∣∣w∣∣²/2 ). Hence the name quadratic optimization.

To ensure correct classification, every data point (x_i, y_i) must satisfy: y_i(w.x_i + b) >= 1. This means:

• Points of the positive class lie on one side of the hyperplane.

• Points of the negative class lie on the other side, at least one margin unit away.

These constraints are linear in terms of $w$ and $b$

---

7. What is the primary reason for using a kernel function in SVM?

A. To increase training speed
B. To handle non-linear relationships efficiently
C. To reduce the number of features
D. To minimize overfitting automatically

Answer: B

Explanation: Kernels implicitly project input data into a higher-dimensional space where linear separation becomes possible.

---

8. In SVM, support vectors are:

A. All training samples
B. Only samples lying on the margin boundaries
C. Samples inside the margin or misclassified
D. Both B and C

Answer: D

Explanation: Support vectors are data points that either lie on the margin boundaries or violate the margin (inside it or misclassified).

Some points lie exactly on the margin boundary. These are support vectors on the margin.

Some points may lie inside the margin or even be misclassified (in soft margin SVM). These are support vectors violating the margin

---

9. When the gamma (γ) parameter of an RBF kernel is too high, what typically happens?

A. The decision boundary becomes smoother
B. Model generalizes better
C. Model overfits by focusing on nearby points
D. Model underfits with large bias

Answer: C

Explanation: High γ makes each point’s influence very localized, leading to an overly complex boundary and overfitting.

What is gamma (γ) parameter?

γ (gamma) is a hyperparameter that controls how quickly similarity decreases with distance.

Small γ - Large influence radius (Each data point affects a wide region, leading to smoother, more general decision boundaries)

Large γ - Small influence radius (Each data point affects only nearby points, leading to tighter, more complex decision boundaries)

---

10. Which of the following metrics is most relevant for evaluating SVM on imbalanced datasets?

A. Accuracy
B. Precision and Recall
C. Log-loss
D. Margin width

Answer: B

Explanation: On imbalanced data, accuracy can be misleading. Precision and recall (or F1-score) better reflect SVM’s real performance.

---

 

Top 10 Machine Learning Testing Stage MCQs with Answers (2025 Updated)

Top 10 Machine Learning Testing Stage MCQs with Answers (2025 Updated)

1. What is the primary purpose of the testing stage in a machine learning workflow?

A. To tune model hyperparameters
B. To evaluate model performance on unseen data
C. To collect additional labeled data
D. To select the best optimization algorithm

Answer: B

Explanation: It helps confirm whether your model has truly learned general patterns — not just memorized the training examples. The testing stage is needed to verify the model’s reliability, fairness, and readiness for deployment ensuring that what you built in training will work in the real world.

---

2. During testing, why must the test dataset remain untouched during training and validation?

A. It helps speed up model convergence
B. It ensures the model learns from all available data
C. It prevents data leakage and gives an unbiased estimate of performance
D. It improves the model’s interpretability

Answer: C

Explanation: The test dataset must remain completely untouched during training and validation because its sole purpose is to measure how well your trained model performs on new, unseen data just like in the real world.

If the test data is used (even indirectly) during training or validation, the model may “learn” patterns or information from it. This is called data leakage, and it causes the model to appear more accurate than it truly is — leading to overestimated performance.

---

3. If a model performs well on validation data but poorly on test data, what does this most likely indicate?

A. Data leakage in training
B. Overfitting to the validation set
C. Underfitting to the training set
D. Insufficient regularization in test data

Answer: B

Explanation: When a model performs well on validation data but poorly on test data, it usually means that the model has overfitted to the validation set. That is, it has learned patterns that are too specific to the validation data, instead of learning general patterns that apply to new, unseen data.

Analogy: Imagine preparing for an exam by practicing only past question papers (validation set). You ace those, but when you get new questions (test set), you struggle because you memorized patterns, not concepts.

Training data: To train the model — adjust weights and learn relationships between features and target.

Validation data: To fine-tune hyperparameters, choose model configurations, and decide when to stop training.

Testing data: To evaluate the final model’s performance on completely unseen data.

---

4. Which metric is least suitable for evaluating a classification model on an imbalanced test set?

A. Precision
B. Recall
C. Accuracy
D. F1-score

Answer: C

Explanation: When a classification dataset is imbalanced, meaning one class (say, “negative”) has far more samples than the other (“positive”), then accuracy becomes a misleading metric.

It may show high values even when the model completely fails to detect the minority class.

We may use Precision, Recall, F1-score, or AUC instead for fair evaluation.

Example: Suppose you have a dataset of 10,000 samples with two classes — “Yes” and “No.” If the dataset is imbalanced with 9,900 “Yes” samples and only 100 “No” samples, a model that simply predicts “Yes” for every instance will achieve an accuracy of 99%.

At first glance, that seems excellent — but in reality, the model fails to detect even a single “No” case. This means it completely ignores the minority class, even though the reported accuracy looks perfect.

---

5. In model evaluation, what does a large difference between training and test accuracy typically indicate?

A. The model is well-calibrated
B. The model is overfitting
C. The model is generalizing well
D. The dataset is balanced

Answer: B

Explanation: A large difference between training and test accuracy (especially when training accuracy is much higher) signals overfitting. This means that the model has learned patterns specific to the training data instead of general trends that apply to new data.

Overfitting = Model performs much better on training data than on unseen data. This means the model memorized rather than learned.

Will there be a possibility that test accuracy is much higher than training accuracy?

No. A large gap, where the test accuracy is significantly higher than training accuracy is not possible and if happens that’s a red flag and usually means something is wrong. But a small difference (test slightly higher) is possible and sometimes expected (due to dropout, regularization, or randomness).

---

6. Which of the following statements about test data is TRUE?

A. Test data should be augmented the same way as training data
B. Test data should be collected after the model is deployed
C. Test data should be used for hyperparameter tuning
D. Test data should come from the same distribution as training data but remain unseen

Answer: D

Explanation: Test data should come from the same distribution as training data because,

Generalization - The goal of machine learning is to generalize, i.e., perform well on new data drawn from the same population as the training data. If the test data is from a different distribution, you’re not measuring generalization — you’re measuring domain shift or transfer performance (a different problem). For example, if you train a machine learning model to predict the height of the person using Indian data and test the model with European data the accuracy drops. This drop is not due to bad model but due to the distribution differs (human biological variation).

Fair performance estimation - Using data from the same distribution ensures the test accuracy reflects how the model will behave on future, similar data (i.e., from the same source). If distributions differ, test results may underestimate or overestimate performance — giving a false impression of model quality.

Same distribution ensures test data represents the same problem domain.

Remain unseen ensures unbiased, realistic evaluation of model generalization.

---

7. In cross-validation, what plays the role of the test set in each fold?

A. The validation split of each fold
B. The training split of each fold
C. The combined training and validation splits
D. A completely new dataset

Answer: A

Explanation: In cross-validation, each fold’s validation split acts as the test set for that round, giving a fair way to test every data point exactly once.

Cross-validation: Cross-validation (often k-fold cross-validation) is a technique to evaluate a model’s performance more reliably, especially when the dataset is small. Instead of having one fixed “train-test” split, cross-validation reuses the data multiple times by dividing it into k parts (called folds).

It is called validation split because in each iteration, the fold that is left out is not used for  training. The model is trained on the remaining folds and evaluated on this left-out fold. This left-out fold acts like a test set in that iteration

---

8. Which evaluation method best simulates real-world testing conditions for time-series models?

A. Random K-fold cross-validation
B. Leave-one-out validation
C. Rolling window validation
D. Stratified sampling

Answer: C

Explanation: In time-series problems (example: stock prices by date, weather readings etc.), data points are ordered in time. So, future values depend on past values. This means you can’t randomly shuffle the data or use ordinary k-fold cross-validation (which mixes past and future samples).

Rolling Window Validation (also called Walk-Forward Validation) is designed specifically for time-series models. It simulates how models are used in the real world: The model is trained on past data, Then tested on future data that occurs later in time.

---

9. Why is the test stage essential before model deployment in real applications?

A. It confirms that the model architecture is optimal
B. It ensures low training loss
C. It verifies generalization ability under unseen scenarios
D. It automatically adjusts hyperparameters

Answer: C

Explanation: The test stage is the final evaluation phase of a machine learning workflow. After a model is trained (and tuned using validation data), it’s tested on a completely unseen dataset called the test set.

This stage checks how well the model will perform on new, real-world data that it hasn’t seen during training or validation.

---

10. What is a common mistake made during the testing phase of ML models?

A. Using standard metrics like RMSE
B. Using separate data splits
C. Measuring inference speed
D. Using test data for model selection

Answer: D

Explanation: The most common mistake during the testing phase is using the test data to make modeling decisions (model selection or hyperparameter tuning).

This leads to data leakage and overestimates true performance.

The test phase is the final, unbiased evaluation of your trained model. It measures how well your model generalizes to unseen data. The test set is not supposed to influence the model in any way.

Model selection means deciding on which model architecture to use (e.g., Random Forest vs. Neural Network) and which hyperparameters perform best (e.g., learning rate, number of layers, etc.). This selection process should happen during validation, not testing.

However, a common mistake is: Checking performance on the test set repeatedly while tuning models, and then picking the one that performs best on the test set.

This seems harmless — but it’s data leakage.

---

 

Machine Learning Training Phase MCQs with Answers [2025 Updated]

 

 Top 10 MCQs on Training of Machine Learning Models with Answers | Gradient Descent & Optimization Explained

 

1. Loss Function Purpose

In supervised training, what is the primary role of the loss function?

A. To measure model speed
B. To measure how far predictions deviate from true labels
C. To determine the optimal learning rate
D. To normalize feature values

Answer: B
 

Explanation: The loss function quantifies prediction error, guiding weight adjustments during training. The loss function is the core compass that guides a model during training — without it, the model has no direction or measure of how well it’s performing.

Loss function is crucial

• Gives feedback to the model
• Shapes the optimization landscape
• Controls bias/variance tradeoff 

---

 

2. Gradient Calculation

In gradient-based optimization, the gradient of the loss function represents:

A. The direction of the steepest descent
B. The direction of the steepest ascent
C. The curvature of the loss surface
D. The absolute value of the error

Answer: B
 

Explanation: The gradient points toward the steepest increase in loss; we move in the opposite direction to minimize it.

What does the gradient tell us?

When we train a model using gradient-based optimization (like gradient descent), we want to minimize the loss function — that is, make the model’s error as small as possible.

To do that, we need to know how the loss changes with respect to the model’s parameters (weights).

That’s exactly what the gradient tells us.

Why do we want to minimize the loss function here?

The gradient itself points toward the direction of maximum increase in the function (loss). But in gradient descent, we want to minimize the loss — so we move in the opposite direction of the gradient.

That’s why the update rule in gradient descent is:

$$w_{new} = w_{old} - \eta \times \nabla L(w)$$ 

---

 

3. Backpropagation Core Idea

What is the main purpose of backpropagation in neural network training?

A. To store intermediate outputs
B. To propagate input forward
C. To compute gradients of weights using the chain rule
D. To normalize activations

Answer: C
 

Explanation: Backpropagation efficiently calculates partial derivatives of the loss with respect to each weight via the chain rule.

Backpropagation (Backward Propagation of Errors) is the algorithm used to train neural networks by adjusting their weights based on the error (loss) between predicted and true outputs.

It’s how the network learns from its mistakes.

 

---

 

4. Mini-Batch Training Advantage

Why is mini-batch gradient descent often preferred over batch or stochastic gradient descent?

A. It eliminates gradient noise completely
B. It balances computational efficiency with gradient stability
C. It always converges faster than batch descent
D. It uses no randomness

Answer: B
 

Explanation: Mini-batches provide more stable updates than stochastic GD and require less computation than full-batch GD.

What is mini-batch gradient descent?

Mini-batch gradient descent is a variant of gradient descent where the training dataset is divided into small batches (subsets) of data. The model updates its weights after processing each mini-batch, rather than after every single example or after the entire dataset. 

Mini-batch gradient descent is chosen over SGD or Batch gradient descent because of the characteristics faster training, stable convergence, memory efficient and GPU optimization. 

---

 

5. Weight Update Rule

In standard gradient descent, how are model weights updated?

A. $w_{new} = w_{old} + \eta \times \nabla L(w)$
B. $w_{new} = w_{old} - \eta \times \nabla L(w)$
C. $w_{new} = w_{old} \times \nabla L(w)$
D. $w_{new} = \eta \times w_{old}$

Answer: B
 

Explanation: We subtract the gradient scaled by the learning rate to move toward lower loss.

When training a model, the goal is to minimize the loss function $L(w)$, which measures how far the model’s predictions are from the true outputs.

• The weights $w$ of the model determine its predictions.

• To reduce the loss, we need to adjust these weights in the “right direction.”

The gradient of the loss function w.r.t. the weights, $\nabla L(w)$, tells us:

• Direction: The direction in which the loss increases fastest.

• Magnitude: How steeply the loss increases along each weight.

So if we follow the gradient as-is, we’d increase the loss — which is the opposite of what we want.

 

---

 

6. Vanishing Gradient Problem

Which activation function is most likely to cause the vanishing gradient problem?

A. ReLU
B. Leaky ReLU
C. Sigmoid
D. ELU

Answer: C
 

Explanation: Sigmoid saturates for large inputs, causing gradients to approach zero and slowing learning.

What is vanishing gradient problem?

When training deep neural networks using gradient-based optimization, the model updates its weights using gradients calculated via backpropagation. In some cases, the gradient becomes extremely small (approaching zero) as it propagates backward through the layers. Due to this, the weights in the earlier layers hardly update and the learning slows dramatically or stops. This is called the vanishing gradient problem.

It often happens with activation functions that “saturate” — i.e., functions whose output flattens for large positive or negative inputs. 

---

 

7. Convergence in Training

Which of the following best indicates training convergence?

A. The validation loss starts increasing
B. The training loss becomes zero
C. The change in loss across epochs becomes negligible
D. The learning rate decreases automatically

Answer: C
 

Explanation: Convergence occurs when further training no longer significantly changes the loss.

Training convergence?

Training convergence refers to the point during the training of a machine learning model where:

• The loss function stops decreasing significantly.

• The model parameters (weights) stabilize.

• Further training does not improve performance on the training data (and ideally on validation data).

In simple words: the model has “learned as much as it can” from the data. 

---

 

8. Optimizer Momentum

What is the role of momentum in optimization algorithms like SGD with momentum?

A. To adapt the learning rate per parameter
B. To average losses across epochs
C. To accelerate convergence by smoothing gradient updates
D. To prevent overfitting

Answer: C
 

Explanation: Momentum accumulates past gradients to keep moving in consistent directions, improving speed and stability.

What is momentum in optimization algorithm?

Momentum is a technique used in gradient-based optimization (like stochastic gradient descent) to accelerate training and improve convergence, especially in deep neural networks. It helps the optimizer move faster in the right direction and smooth out oscillations. Think of it as adding “inertia” to the weight updates. 

Why momentum in optimization algorithm?

During training, gradient descent can face problems like Oscillations in narrow valleys (Gradients may point in zig-zag directions, slowing convergence) and/or Slow progress in shallow regions (Gradients are small so tiny updates; hence slow learning). Momentum solves both by accumulating past gradients and using them to influence the current update. 

---

 

9. Learning Rate Scheduler

Why might we use a learning rate scheduler during training?

A. To gradually reduce learning rate to fine-tune convergence
B. To reduce overfitting by randomizing learning rates
C. To restart training from previous checkpoints
D. To ensure constant learning rate

Answer: A
 

Explanation: Decaying the learning rate allows large early steps and fine adjustments later for stable convergence.

What is learning rate scheduler and why is needed?

A learning rate scheduler is a strategy to change the learning rate dynamically during training rather than keeping it constant. Typically, the learning rate starts larger at the beginning (It allows faster learning). Then it gradually decreases (allows smaller, precise steps to fine-tune convergence near minima).

Faster initial learning, Stable convergence, and Better final performance are the reasons for using a learning rate scheduler. 

---

 

10. Batch Normalization Effect

How does batch normalization help during training?

A. By eliminating the need for bias terms
B. By increasing model capacity
C. By forcing all activations to zero
D. By reducing vanishing/exploding gradients and speeding up convergence

Answer: D
 

Explanation: Batch normalization standardizes layer inputs, stabilizing gradient flow and allowing faster, more reliable training.

---

 

 

 

 

10 Hot Decision Tree MCQs: Gain Ratio, Continuous Attributes & Tie-Breaking

10 Hot Decision Tree MCQs: Gain Ratio, Continuous Attributes & Tie-Breaking

1. The root node in a decision tree is selected based on:

A) Minimum entropy

B) Maximum information gain

C) Minimum Gini

D) Random initialization

Answer: B

Explanation: The root node is the first split in the tree. The goal is to reduce uncertainty in the dataset as much as possible. Decision tree algorithms (like ID3, C4.5) calculate information gain for all attributes. The attribute with the highest information gain is chosen as the root because it splits the data in the best way, creating the purest child nodes.

The root node is selected by picking the attribute that gives the largest reduction in entropy — i.e., the highest information gain.

---

2. If a dataset has 100% identical attribute values for all samples but mixed labels, the information gain of any attribute will be:

A) 0

B) 1

C) Undefined

D) Negative

Answer: A

Explanation: If all samples have the same attribute values, splitting on any attribute does not reduce uncertainty. Child nodes after the split are exactly the same as the parent in terms of class distribution. Therefore, the weighted entropy of children = entropy of parent. So, the information gain = 0.

---

3. In a two-class problem, Gini Index = 0.5 represents:

A) Maximum impurity

B) Pure split

C) Perfect classification

D) Minimum impurity

Answer: A

Explanation: Gini = 0 → node is pure (all samples belong to one class). Gini = 0.5 → node is maximally impure in a two-class problem (50%-50% split). Gini Index = 0.5 means the node is completely mixed, with an equal number of samples from both classes.

---

4. A pruned decision tree generally has:

A) Higher accuracy on training data but lower on test data

B) Lower training accuracy but better generalization

C) Equal accuracy everywhere

D) Random performance

Answer: B

Explanation: Pruning sacrifices some training accuracy to avoid overfitting. Pruning simplifies the tree. Slightly worse on training data but much better on new/unseen data.

Option A: NO - this is an overfitted tree, not a pruned one.

Option C: NO - Rare in practice

Option D: NO - Pruning is systematic not random.

---

5. In manual decision tree construction, if an attribute gives 0 information gain, what should you do?

A) Still choose it

B) Pplit based on it partially

C) Skip it for splitting 

D) Replace missing values

Answer: C

Explanation: If an attribute gives 0 information gain, it cannot help separate classes, so you ignore it and choose a better attribute for splitting.

---

6. In a decision tree, if a node contains only one sample, what is its entropy?

A) 0
B) 0.5
C) 1
D) Cannot be calculated

Answer: A

Explanation: A single sample belongs to a single class → node is perfectly pure → entropy = 0.

---

7. Which splitting criterion can be used for multi-class problems besides binary classification?

A) Gini Index
B) Entropy / Information Gain
C) Gain Ratio
D) All of the above

Answer: D

Explanation: All these measures can handle more than two classes; they just compute probabilities for each class.

---

8. Which of the following is most likely to cause overfitting in a decision tree?

A) Shallow tree
B) Large minimum samples per leaf
C) Very deep tree with small leaves
D) Using pruning

Answer: C

Explanation: Deep trees with tiny leaves memorize training data → overfit → poor generalization. 

---

9. In manual construction of a decision tree, what is the first step?

A) Calculate child node entropy
B) Select root attribute based on information gain
C) Split dataset randomly
D) Prune unnecessary branches

Answer: B

Explanation: The root is chosen to maximize information gain, which reduces the initial uncertainty the most.

---

10. If a node’s children after a split all have entropy = 0.3 and the parent has entropy = 0.3, what does it indicate?

A) Maximum information gain
B) Node is pure
C) Overfitting
D) No  information gain

Answer: D

Explanation: Information gain = Parent entropy − Weighted child entropy = 0 → the split did not improve purity.

---