StudyFox
कीमतेंब्लॉगएक्सटेंशनमुफ़्त टूल
  • कीमतें
  • ब्लॉग
  • एक्सटेंशन
  • मुफ़्त टूल
  • शुरू करेंसाइन इन
StudyFox

अपने नोट्स, PDF और स्लाइड्स को फ़्लैशकार्ड, क्विज़, नोट्स और माइंड मैप में बदलें।

StudyFox खोलें

स्टडी टूल्स

  • होमवर्क हेल्पर
  • नोट मेकर
  • फ़्लैशकार्ड मेकर
  • क्विज़ मेकर
  • माइंड मैप मेकर

प्रोडक्ट

  • कीमतें
  • Chrome एक्सटेंशन
  • मुफ़्त एक्सटेंशन
  • स्टडी सेट देखें
  • ब्लॉग

सहायता

  • सहायता केंद्र
  • संपर्क करें
  • गोपनीयता नीति
  • सेवा की शर्तें

तुलना करें

  • Chegg का विकल्प
  • Course Hero का विकल्प
  • Coursology का विकल्प
  • Photomath का विकल्प
  • Gauthmath का विकल्प
  • PhotoSolve का विकल्प
  • CollegeTools का विकल्प
  • Transcript का विकल्प
  • SmartSolve का विकल्प
  • Homeworkify का विकल्प
  • AnswersAi का विकल्प
  • StudyX का विकल्प
  • Quizlet का विकल्प
  • Anki का विकल्प
  • Knowt का विकल्प
  • Brainscape का विकल्प
  • सभी तुलनाएँ

© 2026 StudyFox AI. सर्वाधिकार सुरक्षित।

  • X
  • Discord
  • Instagram
ExploreEngineering & TechData Science & AIDeep Learning Concepts and Diagrams Quiz

Quiz · Medium

Deep Learning Concepts and Diagrams Quiz

Activation functions, backpropagation, training problems, CNNs, RNNs and LSTMs, attention and transformers, autoencoders and GANs, with figures to read.

by Kill h.22 questionsCreated Nov 20, 2026, 08:26 PM
Data Science & AI· Deep LearningTags#deep-learning#neural-networks#quiz#diagrams

22 questions

  1. 1

    Why are nonlinear activation functions essential in a neural network?

    • A

      They make training faster on CPUs

    • Without them, any stack of layers is equivalent to a single linear transformation

    • C

      They remove the need for a loss function

    • D

      They keep the weights between 0 and 1

    Composing linear maps gives another linear map, so depth adds no expressive power. Nonlinearity lets the network model curved decision boundaries.

  2. 2

    A hidden layer should use which activation by default in a modern feed-forward or convolutional network?

    • A

      Step function

    • B

      Softmax

    • ReLU

    • D

      Sigmoid

    ReLU is cheap, does not saturate for positive inputs and mitigates vanishing gradients. Softmax is for the output layer of multiclass classifiers; sigmoid and step functions saturate or have zero gradients.

  3. 3

    A network outputs one probability per class for exactly one correct class. Which output activation and loss go together?

    • A

      Linear and binary cross-entropy

    • Softmax and categorical cross-entropy

    • C

      Sigmoid and MSE

    • D

      ReLU and hinge loss

    Softmax makes the outputs a probability distribution; categorical cross-entropy compares it with the one-hot label.

  4. 4

    In the figure, the sigmoid's derivative (dashed) peaks at about what value?

    • A

      0.1

    • B

      0.5

    • C

      1.0

    • 0.25

    is largest at x = 0 where it equals 0.5 x 0.5 = 0.25. Chaining many such factors (each at most 0.25) explains vanishing gradients.

  5. 5

    What does backpropagation compute?

    • A

      The best learning rate

    • B

      The validation accuracy

    • The gradient of the loss with respect to every parameter, using the chain rule

    • D

      The new weights after each update

    Backprop only produces gradients. The optimizer (SGD, Adam) uses them to update the weights.

  6. 6

    Why are LSTMs better than simple RNNs at long sequences?

    • A

      They process tokens in parallel

    • The additive cell-state path lets gradients flow over long spans

    • C

      They have fewer parameters

    • D

      They do not need training

    Simple RNNs multiply many Jacobians and the gradient vanishes. The LSTM's gated, mostly additive cell-state update mitigates this. LSTMs are still sequential, not parallel.

  7. 7

    Training loss keeps falling while validation loss rises. What should you do?

    • A

      Remove the validation set

    • B

      Train for many more epochs

    • Stop near the validation minimum (early stopping) or add regularization

    • D

      Increase the learning rate

    That pattern is overfitting. Early stopping at the lowest validation loss, plus dropout or weight decay, helps.

  8. 8

    During training the loss grows larger and larger and finally becomes NaN. The most likely cause is:

    • A learning rate that is too large (or exploding gradients)

    • B

      Using a softmax output

    • C

      Too much data

    • D

      A learning rate that is too small

    Overshooting steps make the loss diverge. Lower the learning rate and use gradient clipping. A tiny learning rate gives slow progress, not NaNs.

  9. 9

    What happens to a dropout layer at test (inference) time?

    • A

      Units are still dropped randomly

    • B

      The layer is replaced by batch norm

    • C

      Half of the weights are set to zero permanently

    • All units are used (no dropping)

    Dropout is applied only during training. With inverted dropout the activations were already scaled during training, so nothing else is needed at inference.

  10. 10

    A dense layer maps 100 inputs to 50 outputs (with biases). How many trainable parameters does it have?

    • A

      5,000

    • 5,050

    • C

      5,100

    • D

      150

    100 x 50 = 5,000 weights plus 50 biases = 5,050. Forgetting the biases gives 5,000.

  11. 11

    A 3x3 convolution with padding 1 and stride 1 is applied to a 64x64 image. What is the output size?

    • A

      32x32

    • B

      62x62

    • C

      66x66

    • 64x64

    . Padding 1 with a 3x3 kernel preserves the size; with no padding the output would be 62x62.

  12. 12

    What does the max-pooling operation in this figure produce from the 4x4 input?

    • A 2x2 map of block maxima: 6, 8 / 3, 4

    • B

      A 2x2 map of block averages

    • C

      A 4x4 map with negatives set to zero

    • D

      A 2x2 map of block minima

    Each 2x2 block is replaced by its largest value: 6, 8, 3, 4. Average pooling would give different numbers. Image: Aphex34 (derivative work: Olexa Riznyk), CC BY-SA 4.0, via Wikimedia Commons.

  13. 13

    Why do CNNs need far fewer parameters than fully connected networks for images?

    • A

      They skip the backward pass

    • They share the same small filters across all positions

    • C

      They use smaller images

    • D

      They have no bias terms

    Weight sharing and local connectivity: one 3x3 filter is reused at every location instead of having a separate weight for each pixel pair.

  14. 14

    What problem do residual (skip) connections mainly address?

    • Difficulty training very deep networks (degradation, vanishing gradients)

    • B

      Overfitting to noisy labels

    • C

      Too small a batch size

    • D

      Slow data loading

    y = F(x) + x gives gradients a direct path and lets blocks learn small corrections. They are not a regularizer.

  15. 15

    In the LSTM figure, the gate that decides what part of the old cell state to keep is the:

    • Forget gate

    • B

      Input gate

    • C

      Reset gate

    • D

      Output gate

    The forget gate multiplies the previous cell state c(t-1) by a value between 0 and 1. The input gate controls new information and the output gate controls what is exposed as h(t). Image: fdeloche, CC BY-SA 4.0, via Wikimedia Commons.

  16. 16

    In scaled dot-product attention, why divide by the square root of d_k?

    • A

      To reduce the number of heads

    • To keep the dot products from growing large and saturating the softmax

    • C

      To mask future tokens

    • D

      To make the weights sum to 1

    Dot products have variance that grows with dimension, pushing the softmax toward one-hot outputs with tiny gradients. Summing to 1 is the softmax's job.

  17. 17

    Which part of the transformer in this figure stops a position from attending to later tokens?

    • A

      Add & Norm

    • B

      The input embedding

    • C

      Positional encoding

    • Masked multi-head attention in the decoder

    The causal mask in the decoder's first attention block blocks future positions so generation matches training. Image: Yuening Jia, CC BY-SA 3.0, via Wikimedia Commons.

  18. 18

    Which model family is trained to predict the next token and is used for most chat language models?

    • Decoder-only (GPT-style)

    • B

      GAN

    • C

      Autoencoder

    • D

      Encoder-only (BERT-style)

    Decoder-only transformers do causal (next-token) language modeling. BERT-style encoders use masked language modeling and are used for understanding tasks.

  19. 19

    What is the role of the bottleneck 'code' in this autoencoder?

    • It forces the network to learn a compact representation of the input

    • B

      It adds noise to the output

    • C

      It stores the labels

    • D

      It is discarded before training

    Because the code is narrower than the input, the encoder must keep only the most informative features to reconstruct the input. Image: Michela Massi, CC BY-SA 4.0, via Wikimedia Commons.

  20. 20

    In a GAN, what is the generator's goal?

    • A

      Predict the next token

    • B

      Classify real vs fake images

    • C

      Minimize reconstruction error

    • Produce samples the discriminator labels as real

    The generator is trained against the discriminator: it improves by fooling it. The discriminator is the one that classifies real vs fake. Image: Mtanti, CC BY-SA 4.0, via Wikimedia Commons.

  21. 21

    True or False: With plain SGD, L2 regularization and weight decay are equivalent.

    • True

    • B

      False

    For plain SGD adding an L2 penalty to the loss is the same as decaying the weights each step. With adaptive optimizers like Adam they are not equivalent, hence AdamW.

  22. 22

    You have only 500 labeled images. Which strategy is usually most effective?

    • A

      Drop the validation set to get more training data

    • Fine-tune a network pre-trained on a large image data set

    • C

      Use a fully connected network on raw pixels

    • D

      Train a very deep CNN from scratch

    Transfer learning reuses general visual features learned from millions of images. Training a deep model from scratch on 500 images would overfit.