Jan 17, 2024 AIP-210 Exam Crack Test Engine Dumps Training With 92 Questions [Q32-Q53]

NO.32 Why do data skews happen in the ML pipeline?

Test and evaluation data are designed incorrectly.

There Is a mismatch between live input data and offline data.

There is a mismatch between live output data and offline data.

There is insufficient training data for evaluation.

NO.33 Which of the following items should be included in a handover to the end user to enable them to use and run a trained model on their own system? (Select three.)

Information on the folder structure in your local machine

Intermediate data files

Link to a GitHub repository of the codebase

README document

Sample input and output data files

NO.34 Which of the following occurs when a data segment is collected in such a way that some members of the intended statistical population are less likely to be included than others?

Algorithmic bias

Sampling bias

Stereotype bias

Systematic value distortion

NO.35 A company is developing a merchandise sales application The product team uses training data to teach the AI model predicting sales, and discovers emergent bias. What caused the biased results?

The AI model was trained in winter and applied in summer.

The application was migrated from on-premise to a public cloud.

The team set flawed expectations when training the model.

The training data used was inaccurate.

NO.36 Which of the following is the correct definition of the quality criteria that describes completeness?

The degree to which all required measures are known.

The degree to which a set of measures are equivalent across systems.

The degree to which a set of measures are specified using the same units of measure in all systems.

The degree to which the measures conform to defined business rules or constraints.

NO.37 Which of the following methods can be used to rebalance a dataset using the rebalance design pattern?

Bagging

Boosting

Stacking

Weighted class

NO.38 What is the primary benefit of the Federated Learning approach to machine learning?

It does not require a labeled dataset to solve supervised learning problems.

It protects the privacy of the user’s data while providing well-trained models.

It requires less computation to train the same model using a traditional approach.

It uses large, centralized data stores to train complex machine learning models.

NO.39 Which of the following algorithms is an example of unsupervised learning?

Neural networks

Principal components analysis

Random forest

Ridge regression

NO.40 For each of the last 10 years, your team has been collecting data from a group of subjects, including their age and numerous biomarkers collected from blood samples. You are tasked with creating a prediction model of age using the biomarkers as input. You start by performing a linear regression using all of the data over the
10-year period, with age as the dependent variable and the biomarkers as predictors.
Which assumption of linear regression is being violated?

Equality of variance (Homoscedastidty)

Independence

Linearity

Normality

NO.41 Which two techniques are used to build personas in the ML development lifecycle? (Select two.)

Population estimates

Population regression

Population resampling

Population triage

Population variance

NO.42 An AI practitioner incorporates risk considerations into a deployment plan and decides to log and store historical predictions for potential, future access requests.
Which ethical principle is this an example of?

Fairness

Privacy

Safety

Transparency

NO.43 Workflow design patterns for the machine learning pipelines:

Aim to explain how the machine learning model works.

Represent a pipeline with directed acyclic graph (DAG).

Seek to simplify the management of machine learning features.

Separate inputs from features.

NO.44 You are building a prediction model to develop a tool that can diagnose a particular disease so that individuals with the disease can receive treatment. The treatment is cheap and has no side effects. Patients with the disease who don’t receive treatment have a high risk of mortality.
It is of primary importance that your diagnostic tool has which of the following?

High negative predictive value

High positive predictive value

Low false negative rate

Low false positive rate

NO.45 An organization sells house security cameras and has asked their data scientists to implement a model to detect human feces, as distinguished from animals, so they can alert th customers only when a human gets close to their house.
Which of the following algorithms is an appropriate option with a correct reason?

A decision tree algorithm, because the problem is a classification problem with a small number of features.

k-means, because this is a clustering problem with a small number of features.

Logistic regression, because this is a classification problem and our data is linearly separable.

Neural network model, because this is a classification problem with a large number of features.

NO.46 A healthcare company experiences a cyberattack, where the hackers were able to reverse-engineer a dataset to break confidentiality.
Which of the following is TRUE regarding the dataset parameters?

The model is overfitted and trained on a high quantity of patient records.

The model is overfitted and trained on a low quantity of patient records.

The model is underfitted and trained on a high quantity of patient records.

The model is underfitted and trained on a low quantity of patient records.

NO.47 Which of the following approaches is best if a limited portion of your training data is labeled?

Dimensionality reduction

Probabilistic clustering

Reinforcement learning

Semi-supervised learning

NO.48 Which three security measures could be applied in different ML workflow stages to defend them against malicious activities? (Select three.)

Disable logging for model access.

Launch ML Instances In a virtual private cloud (VPC).

Monitor model degradation.

Use data encryption.

Use max privilege to control access to ML artifacts.

Use Secrets Manager to protect credentials.

NO.49 Which of the following models are text vectorization methods? (Select two.)

Lemmatization

PCA

Skip-gram

TF-IDF

Tokenization

t-SNE

NO.50 An HR solutions firm is developing software for staffing agencies that uses machine learning.
The team uses training data to teach the algorithm and discovers that it generates lower employability scores for women. Also, it predicts that women, especially with children, are less likely to get a high-paying job.
Which type of bias has been discovered?

Automation

Emergent

Preexisting

Technical

NO.51 You are developing a prediction model. Your team indicates they need an algorithm that is fast and requires low memory and low processing power. Assuming the following algorithms have similar accuracy on your data, which is most likely to be an ideal choice for the job?

Deep learning neural network

Random forest

Ridge regression

Support-vector machine

NO.52 A market research team has ratings from patients who have a chronic disease, on several functional, physical, emotional, and professional needs that stay unmet with the current therapy. The dataset also captures ratings on how the disease affects their day-to-day activities.
A pharmaceutical company is introducing a new therapy to cure the disease and would like to design their marketing campaign such that different groups of patients are targeted with different ads. These groups should ideally consist of patients with similar unmet needs.
Which of the following algorithms should the market research team use to obtain these groups of patients?

k-means clustering

k-nearest neighbors

Logistic regression

Naive-Bayes

NO.53 Your dependent variable data is a proportion. The observed range of your data is 0.01 to 0.99. The instrument used to generate the dependent variable data is known to generate low quality data for values close to 0 and close to 1. A colleague suggests performing a logit-transformation on the data prior to performing a linear regression. Which of the following is a concern with this approach?
Definition of logit-transformation
If p is the proportion: logit(p)=log(p/(l-p))

After logit-transformation, the data may violate the assumption of independence.

Noisy data could become more influential in your model.

The model will be more likely to violate the assumption of normality.

Values near 0.5 before logit-transformation will be near 0 after.

Leave a Reply Cancel reply