Discuss UiPath UiPath-SAIv1 Exam Topic 8 Question 37

When creating a training dataset, what is the recommended number of samples for the Classification fields?

A5-10 document samples from each class.

B10-20 document samples from each class.

C20-50 document samples from each class.

D50-200 document samples from each class.

Suggested Answer: C

According to the UiPath documentation, the recommended number of samples for the classification fields depends on the number of document types and layouts that you want to classify. The more document types and layouts you have, the more samples you need to cover the diversity of your data.However, a general guideline is to have at least 20-50 document samples from each class, as this would provide enough data for the classifiers to learn from12.A large number of samples per layout is not mandatory, as the classifiers can generalize from other layouts as well3.

References:1: Document Classification Training Overview2: Document Classification Training Related Activities3: Training High Performing Models

Get Premium UiPath-SAIv1 Questions as Interactive Web-Based Practice Test or PDF Real Exam Format Detailed Explanations Instant Access Get Your Practice Test

13 comments · newest first

Ailene 10 months ago

More samples usually mean better accuracy, right?

Coleen 10 months ago

Definitely agree with option C!

Kenneth 10 months ago

Wait, 50-200 samples? That seems excessive!

Teddy 10 months ago

I thought 10-20 was enough for basic models.

Veronique 11 months ago

I've read that 20-50 samples is a good range.

Graciela 11 months ago

I recall reading that 20-50 samples is often recommended, but I wonder if it varies by the complexity of the classes.

Garry 11 months ago

I’m a bit confused; I thought it was around 10-20 samples, but that seems low for effective training.

Julio 11 months ago

I practiced a question similar to this, and I feel like 50-200 samples might be the right range for good classification.

Lawana 12 months ago

I think I remember something about needing at least 20 samples per class, but I'm not entirely sure.

Queenie 12 months ago

Okay, I've got this. The key is to have enough samples to capture the variability in each class, but not so many that the model gets too complex. I'd say 10-20 samples per class is a solid sweet spot.

Tyisha 12 months ago

I'm a bit confused on this one. I know more data is generally better, but I'm not sure if that applies the same way for classification tasks. I'll have to review my notes.

Carlota 1 year ago

I'm pretty sure the recommended range is 20-50 samples per class. That seems like a good balance between having enough data and not overfitting.

Alease 1 year ago

Hmm, this is a tricky one. I'll need to think carefully about the trade-offs between sample size and model complexity.

UiPath-SAIv1 Exam - Topic 8 Question 37 Discussion