📚 Data Science Riddle
You have a dataset with 1,000 samples and 10,000 features. What’s a common problem you might face when training a model on this data?
You have a dataset with 1,000 samples and 10,000 features. What’s a common problem you might face when training a model on this data?
Anonymous Quiz
20%
Underfitting
60%
Overfitting due to high dimensionality
7%
Data leakage
13%
Incorrect feature scaling
❤5😁2👍1
What is RAG? 🤖📚
RAG stands for Retrieval-Augmented Generation.
It’s a technique where an AI model first retrieves relevant info (like from documents or a database), and then generates an answer using that info.
🧠 Think of it like this:
Instead of relying only on what it "knows", the model looks things up first - just like you would Google something before replying.
🔍 Retrieval + 📝 Generation = Smarter, up-to-date answers!
RAG stands for Retrieval-Augmented Generation.
It’s a technique where an AI model first retrieves relevant info (like from documents or a database), and then generates an answer using that info.
🧠 Think of it like this:
Instead of relying only on what it "knows", the model looks things up first - just like you would Google something before replying.
🔍 Retrieval + 📝 Generation = Smarter, up-to-date answers!
❤4🔥3
Dropout Explained Simply
Neural networks are notorious for overfitting ( they memorize training data instead of generalizing).
One of the simplest yet most powerful solutions? Dropout.
During training, dropout randomly “drops” a percentage of neurons ( 20–50%). Those neurons temporarily go offline, meaning their activations aren’t passed forward and their weights aren’t updated in that round.
👉 What this does:
✔️ Forces the network to avoid relying on any single path.
✔️ Creates redundancy → multiple neurons learn useful features.
✔️ Makes the model more robust and less sensitive to noise.
When testing happens, dropout is turned off, and all neurons fire but now they collectively represent stronger, generalized patterns.
Imagine dropout like training with handicaps. It’s as if your brain had random “short blackouts” while studying, forcing you to truly understand instead of memorizing.
And that’s why dropout remains a go-to regularization technique in deep learning and even in advanced architectures.
Neural networks are notorious for overfitting ( they memorize training data instead of generalizing).
One of the simplest yet most powerful solutions? Dropout.
During training, dropout randomly “drops” a percentage of neurons ( 20–50%). Those neurons temporarily go offline, meaning their activations aren’t passed forward and their weights aren’t updated in that round.
👉 What this does:
✔️ Forces the network to avoid relying on any single path.
✔️ Creates redundancy → multiple neurons learn useful features.
✔️ Makes the model more robust and less sensitive to noise.
When testing happens, dropout is turned off, and all neurons fire but now they collectively represent stronger, generalized patterns.
Imagine dropout like training with handicaps. It’s as if your brain had random “short blackouts” while studying, forcing you to truly understand instead of memorizing.
And that’s why dropout remains a go-to regularization technique in deep learning and even in advanced architectures.
❤7
📚 Data Science Riddle
Which algorithm groups data into clusters without labels?
Which algorithm groups data into clusters without labels?
Anonymous Quiz
13%
Decision Tree
13%
Linear Regression
65%
K-Means
9%
Naive Bayes
❤2
📚 Data Science Riddle
In PCA, what do eigenvectors represent?
In PCA, what do eigenvectors represent?
Anonymous Quiz
47%
Directions of maximum variance
30%
Amount of variance captured
10%
Data reconstruction error
13%
Orthogonality of inputs
👍4
📚 Data Science Riddle
What metric is commonly used to decide splits in decision trees?
What metric is commonly used to decide splits in decision trees?
Anonymous Quiz
56%
Entropy
18%
Accuracy
6%
Recall
20%
Variance
❤4
📚 Data Science Riddle
Which algorithm is most sensitive to feature scaling?
Which algorithm is most sensitive to feature scaling?
Anonymous Quiz
26%
Decision Tree
24%
Random Forest
35%
KNN
16%
Naive Bayes