Consider this Python code snippet using PyTorch:

• A.
• B.torch.Size ([32, 5121). The issue is a dimension mismatch.
• C.torch.Size([256, 512]). This implementation is correct and efficient for cross-modal attention.
• D.torch.Size ([32, 32]). This is correct and computes attention weights for each text-image pair in the batch independently
• E.Error. The transpose operation is incorrect for achieving cross-modal attention.

Comment: *

Name: *

Email: *

Verification: *

Consider a multimodal emotion recognition system that uses both facial expressions (images) and speech (audio). You want to fuse the information from these two modalities at the decision level. Which of the following techniques would be MOST suitable for decision-level fusion?

• A.Concatenate the feature vectors extracted from the images and audio, then train a single classifier.
• B.Train separate classifiers for images and audio, then average their output probabilities for each emotion class.
• C.Train separate classifiers for images and audio, then use a weighted average of their output probabilities based on the confidence scores of each classifier.
• D.Train separate classifiers for images and audio, then use the output of the image classifier as input to the audio classifier-
• E.Train a single transformer to process both images and audio in sequence.

Comment: *

Name: *

Email: *

Verification: *

You are evaluating a multimodal model that generates descriptions for video clips. You have human ratings for the relevance, fluency, and coherence of the generated descriptions. Which statistical test is MOST appropriate for determining if there is a statistically significant difference in the median ratings for each of these criteria (relevance, fluency, coherence) between two different versions of your model?

• A.T-test
• B.ANOVA
• C.Mann-Whitney U test
• D.Kruskal-Wallis test
• E.Friedman Test

Comment: *

Name: *

Email: *

Verification: *

Which visualization technique is suitable for representing the distribution of performance scores for different multimodal ML models over different modalities?

• A.Heatmap
• B.Histogram
• C.Box plot
• D.Pie chart

Comment: *

Name: *

Email: *

Verification: *

You are working with a multimodal generative model that combines text and image inputs. The model's performance is suboptimal when generating images conditioned on complex text descriptions. Which data analysis technique would be MOST effective in identifying the root cause of this issue?

• A.Calculating the average image resolution in the dataset.
• B.Performing a sentiment analysis of the text descriptions to identify potential biases.
• C.Analyzing the correlation between the complexity of the text descriptions (e.g., number of words, sentence structure) and the quality of the generated images.
• D.Measuring the frequency of different objects appearing in the images.
• E.Calculating the mean pixel intensity of the images.

Comment: *

Name: *

Email: *

Verification: *