Scribd
Upload
4 views2 pages

Supervised vs Unsupervised Learning Guide

The document outlines key concepts in machine learning, differentiating between supervised and unsupervised learning, and explaining optimization algorithms like stochastic gradient descent. It covers various neural network architectures, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), detailing features such as parameter sharing, pooling layers, and bidirectional processing. Additionally, it discusses the importance of validation sets, estimators, and nonlinearity functions in model training and performance.

Uploaded by

jayaprakash9223
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
4 views2 pages

Supervised vs Unsupervised Learning Guide

The document outlines key concepts in machine learning, differentiating between supervised and unsupervised learning, and explaining optimization algorithms like stochastic gradient descent. It covers various neural network architectures, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), detailing features such as parameter sharing, pooling layers, and bidirectional processing. Additionally, it discusses the importance of validation sets, estimators, and nonlinearity functions in model training and performance.

Uploaded by

jayaprakash9223
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

1. Differentiate supervised and unsupervised learning.

• Supervised Learning → Model is trained with labeled data (input + output).

• Unsupervised Learning → Model is trained with only input data (no labels).

2. What is stochastic gradient descent?


Stochastic Gradient Descent (SGD) is an optimization algorithm that updates model
parameters using gradients calculated from a single training example or small batch, instead
of the full dataset. It speeds up training.

3. Define estimator.
An estimator is a mathematical function or model parameter that estimates some property of
data (like mean, variance, or model weights).

4. What are validation sets and why are they important?


A validation set is a part of data (separate from training) used to tune hyperparameters and
check model performance. It prevents overfitting.

5. What are sparse interactions in a convolutional neural network?


In CNNs, each neuron is connected to only a small local region (receptive field) of the input,
not the whole input. This reduces computation and parameters → called sparse interactions.

6. Present an outline of pooling layer in convolutional neural network.


Pooling layer reduces the spatial dimensions (height × width) of feature maps.

• Max Pooling → selects max value.

• Average Pooling → takes average.

• Benefits → reduces computation, controls overfitting, preserves important features.

7. Explain feature map.


A feature map is the output of convolution operation on the input image using a filter. It
represents extracted features like edges, textures, shapes.

8. Discuss about parameter sharing in neural network.


In CNN, the same filter weights are reused across the entire input image. This reduces the
number of parameters and improves efficiency → called parameter sharing.
9. Define a recurrent neural network.
RNN is a neural network designed for sequential data (e.g., text, speech). It has feedback
loops where output at one step is fed as input to the next step.

10. What is Bidirectional Recurrent Neural Network?


A Bidirectional RNN processes input in both forward and backward directions. This captures
past and future context in sequences.

11. Mention any two nonlinearity functions used in CNN.

• ReLU (Rectified Linear Unit)

• Sigmoid / Tanh

12. What is the role of optimizers in CNN training?


Optimizers update the weights of the network to minimize loss function efficiently. Examples:
SGD, Adam, RMSProp.

13. Mention any two RNN design patterns.

• Many-to-One (e.g., sentiment classification)

• Many-to-Many (e.g., language translation, video captioning)

14. What is bidirectional RNN?


A Bidirectional RNN has two RNN layers: one processes input sequence forward, the other
backward, combining both contexts.

15. Compare LSTM with vanilla RNN.

• Vanilla RNN → suffers from vanishing gradient, cannot handle long-term dependencies.

• LSTM (Long Short-Term Memory) → has memory cells and gates (input, output, forget)
to retain long-term dependencies effectively.

You might also like