AI0-001 Machine Learning and Deep Learning Practice Test 2 — 15 Questions

Question 1

A team trains a random forest model on a dataset with 50 features. The model's performance on the test set is significantly worse than on the training set. Which technique is most appropriate to address this issue?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Apply cross-validation to tune hyperparameters and reduce overfitting. The model's significantly worse test performance compared to training performance is a classic symptom of overfitting. Cross-validation is the most appropriate technique to address this because it evaluates the model's performance on multiple subsets of the data, providing a more robust estimate of generalization. Tuning hyperparameters (e.g., max_depth, min_samples_split) using cross-validation directly reduces overfitting by constraining model complexity, which is the core issue here.

Answer

Increase the number of trees in the forest

Answer

Use feature scaling

Answer

Perform PCA to reduce dimensions

Question 2

A data scientist is using an ensemble method to combine multiple models. Which three statements about bagging (Bootstrap Aggregating) are true? (Select THREE.)

Accepted Answer

It reduces variance without increasing bias. Bagging reduces variance by training models on different bootstrap samples of the data and averaging their predictions. Since each model is trained independently on a random sample with replacement, the ensemble's variance decreases without introducing additional bias, as the expected prediction remains unbiased. This is a key property that distinguishes bagging from boosting, which reduces both bias and variance.

Answer

It requires the base models to be of different types

Answer

It reduces the error by combining weak learners

Question 3

Refer to the exhibit. A developer is using the above configuration for a multi-class classification task. The model performs well on training data but poorly on validation data. Which modification could help?

Accepted Answer

Add L2 regularization to the dense layers. The model is overfitting, as indicated by good performance on training data but poor on validation data. Adding L2 regularization to the dense layers penalizes large weights, reducing model complexity and improving generalization. This directly addresses overfitting without disrupting the training dynamics.

Answer

Remove dropout

Answer

Increase the dropout rate

Answer

Increase the learning rate

Question 4

Refer to the exhibit. The training pod is using 2 GPUs. During training, the GPU utilization is only 30% each. What is the most likely cause?

Accepted Answer

The batch size is too small to fully utilize GPUs. With a batch size that is too small, each GPU receives insufficient data per training step to saturate its compute cores. This leads to low utilization (e.g., 30%) because the GPUs spend most of their time idling while waiting for the next batch to be loaded and processed. Increasing the batch size allows each GPU to process more data in parallel, improving throughput and utilization.

Answer

The learning rate is too high

Answer

The image is missing CUDA libraries

Answer

The number of epochs is too high

Question 5

Refer to the exhibit. A compliance audit requires that model predictions be explainable for regulatory reasons. Which setting in the deployment configuration supports this requirement?

Accepted Answer

explainability: "required". The 'explainability' setting in the deployment configuration directly enables model interpretability features, such as SHAP or LIME, which generate human-readable explanations for individual predictions. This is essential for compliance audits that require transparency into how a model arrived at a specific output, satisfying regulatory requirements like GDPR or financial industry standards.

Answer

target_latency: 100

Answer

data_retention: "90 days"

Answer

drift_detection: true

Question 6

A team trains a deep learning model for image classification with 1000 classes. The training loss decreases but validation loss starts increasing after 10 epochs. What should they do first?

Accepted Answer

Add dropout layers. The scenario describes overfitting: training loss decreases while validation loss increases. Adding dropout layers is a regularization technique that randomly drops neurons during training, preventing co-adaptation and improving generalization. This directly addresses the validation loss increase without altering the learning dynamics or data distribution.

Answer

Use data augmentation

Answer

Increase batch size

Answer

Reduce learning rate

Question 7

A deep learning model for sentiment analysis has millions of parameters and is trained on a small dataset. Which technique can help prevent overfitting?

Accepted Answer

Dropout. Dropout is a regularization technique that randomly drops a fraction of neurons during training, which prevents the model from relying too heavily on any single feature and forces it to learn more robust representations. This is particularly effective when the model has millions of parameters but is trained on a small dataset, as it reduces co-adaptation of neurons and mitigates overfitting.

Answer

Learning rate scheduling

Answer

Batch normalization

Answer

Early stopping

Question 8

Which TWO techniques are commonly used for feature scaling? (Choose two.)

Accepted Answer

Standardization. Standardization (A) is a common feature-scaling technique that transforms each feature by subtracting its mean and dividing by its standard deviation, producing values with mean 0 and variance 1 (z-scores). Min-Max scaling (C) is also a standard scaling technique that rescales each feature to a fixed range, typically [0, 1], using (x - min) / (max - min). One-hot encoding (B) is a categorical encoding method that creates binary indicator columns, not a scaling technique. Normalization (D) is a distractor here because, although the term is sometimes loosely used for scaling, it more precisely refers to adjusting vector magnitudes (e.g., to unit norm) and is not one of the two canonical scaling techniques being tested. PCA (E) is a dimensionality-reduction technique based on projecting data onto principal components, not a feature-scaling method.

Answer

One-hot encoding

Answer

Normalization

Answer

PCA

Question 9

Which THREE are common activation functions used in neural networks? (Choose three.)

Accepted Answer

Sigmoid. Sigmoid is a common activation function in neural networks because it maps any real-valued input to a value between 0 and 1, making it useful for binary classification outputs. It introduces non-linearity and has a smooth gradient, though it suffers from vanishing gradient issues in deep networks.

Answer

K-means

Answer

Softmax

Question 10

Based on the exhibit, what does this indicate about the model?

Accepted Answer

The model has high precision but low recall, missing many positives. The exhibit shows a confusion matrix or performance metrics where the model correctly identifies some positives but has a high number of false negatives, indicating low recall. This means the model misses many actual positive cases, which aligns with high precision (few false positives) but low recall. Option D correctly describes this imbalance.

Answer

The model has balanced performance

Answer

The model is underfitting

Answer

The model is overfitting

Question 11

A data scientist is using a gradient boosting model (XGBoost) for a regression task and observes that the model's performance on the training set is much better than on the test set. Which hyperparameter tuning strategy would most effectively reduce overfitting?

Accepted Answer

Reduce the maximum depth of trees. Reducing the maximum depth of trees limits the complexity of individual trees, preventing them from learning overly specific patterns in the training data. In XGBoost, deeper trees can capture noise and lead to high variance, so lowering depth directly reduces overfitting by enforcing simpler models.

Answer

Increase the number of boosting rounds

Answer

Increase the learning rate

Answer

Subsample less than 1.0

Question 12

A deep learning model for autonomous vehicle perception uses a large convolutional neural network. During deployment, the model misclassifies a stop sign that has a small sticker on it. This is likely an example of what type of vulnerability, and which defense is most appropriate?

Accepted Answer

Adversarial attack; implement adversarial training. The misclassification of a stop sign with a small sticker is a classic example of an adversarial attack, where imperceptible perturbations (like a sticker) cause the model to output an incorrect prediction. Adversarial training, which augments the training data with adversarial examples, is the most appropriate defense because it improves the model's robustness against such perturbations by explicitly learning to resist them.

Answer

Model inversion; add differential privacy

Answer

Data poisoning; use robust aggregation

Answer

Transfer learning; use domain adaptation

Question 13

A company is deploying a machine learning model that predicts customer churn. The model currently has high variance. Which THREE actions should the data scientist take to reduce variance? (Select THREE.)

Accepted Answer

Reduce model complexity (e.g., fewer features, simpler model).. High variance means the model is overfitting the training data and is too sensitive to noise, so reducing model complexity (A) helps by limiting the model's capacity to fit spurious patterns, e.g., using fewer features or a simpler algorithm. Regularization (B) directly penalizes large weights (L1/L2, dropout, etc.), constraining the hypothesis space and thus lowering variance. Adding more training data (C) gives the learner more evidence, which reduces the variance component of expected error and stabilizes the learned function. Removing outliers (D) is a data-cleaning step that can reduce the influence of extreme points but does not systematically address the bias-variance tradeoff and is not a standard variance-reduction remedy. Increasing model complexity (E) is the opposite of what is needed, since it would raise variance further and worsen overfitting.

Answer

Remove outliers from the training data.

Answer

Increase model complexity.

Question 14

A healthcare startup is developing a diagnostic system using medical images. The team has collected 10,000 labeled images of skin lesions. They plan to train a convolutional neural network (CNN) from scratch. However, training converges slowly, and the validation accuracy plateaus at 70%. The data scientist suspects overfitting. The dataset contains 8,000 images of benign lesions and 2,000 of malignant. The team has limited GPU resources. Which of the following is the MOST effective course of action to improve validation accuracy? A. Reduce the number of convolutional layers. B. Apply transfer learning using a pre-trained model on ImageNet. C. Increase the learning rate by a factor of 10. D. Add more dropout after every convolutional layer.

Accepted Answer

Apply transfer learning using a pre-trained model on ImageNet.. Transfer learning leverages a model pre-trained on a large dataset (e.g., ImageNet), which provides useful features for medical images and reduces the need for large amounts of data and computational resources. It is particularly effective when the dataset is small and imbalanced. Option A (increasing learning rate) might cause divergence or overshoot minima. Option B (reducing layers) may reduce capacity and underfit. Option C (adding dropout) can help with overfitting but is unlikely to jump from 70% to a significantly higher accuracy given limited data; transfer learning provides a stronger boost.

Answer

Increase the learning rate by a factor of 10.

Answer

Reduce the number of convolutional layers.

Answer

Add more dropout after every convolutional layer.

Question 15

A financial institution uses a random forest model to approve loan applications. Recently, the model's false positive rate has increased, leading to more defaults. The data science team reviews the feature importance and finds that the model heavily relies on a feature 'zip code' which correlates with income. The company is concerned about fairness. The regulatory team requires that the model's predictions are not biased against protected groups. Which action BEST addresses the fairness concern while maintaining predictive performance? A. Remove the 'zip code' feature and retrain the model. B. Use adversarial debiasing to train a model that is invariant to protected attributes. C. Add more training data from underrepresented zip codes. D. Apply a post-processing technique that adjusts thresholds for different groups.

Accepted Answer

Use adversarial debiasing to train a model that is invariant to protected attributes.. Adversarial debiasing trains a model to be invariant to protected attributes by simultaneously optimizing for predictive accuracy and minimizing the ability to predict the protected attribute from the model's predictions. This directly addresses fairness while maintaining performance, unlike removing a feature or post-processing.

Answer

Apply a post-processing technique that adjusts thresholds for different groups.

Answer

Remove the 'zip code' feature and retrain the model.

Answer

Add more training data from underrepresented zip codes.