Start learning →

Claude AlexNet CNN ImageNet — practice questions

Explore 16 free AlexNet CNN ImageNet questions in Claude Aifoundations. Open each question or exercise to read its content, reference answers and available explanation.

Practise AlexNet CNN ImageNet with Quill →

AlexNet CNN ImageNet questions and exercises with answers

Showing questions 1–16 of 16.

AlexNet, per Krizhevsky et al., consists of how many learned layers?Read the answer and worked solution →The nonlinearity used after every convolutional and FC layer isRead the answer and worked solution →Which technique randomly zeroes hidden neurons during training?Read the answer and worked solution →AlexNet's input image size isRead the answer and worked solution →The final softmax layer in AlexNet outputs distributions overRead the answer and worked solution →AlexNet has roughly how many learnable parameters?Read the answer and worked solution →Why are ReLU networks faster to train than tanh networks?Read the answer and worked solution →AlexNet's first form of data augmentation isRead the answer and worked solution →AlexNet's second form of data augmentation isRead the answer and worked solution →AlexNet uses overlapping pooling withRead the answer and worked solution →At test time dropout is approximated byRead the answer and worked solution →AlexNet was trained for roughly how many epochs on ImageNet?Read the answer and worked solution →Removing any single middle convolutional layer of AlexNet costs roughlyRead the answer and worked solution →The SGD update rule for AlexNet usesRead the answer and worked solution →Image similarity in AlexNet's last hidden layer is measured byRead the answer and worked solution →Fill in the AlexNet numbersRead the answer and worked solution →

Worked practice questions

AlexNet, per Krizhevsky et al., consists of how many learned layers?

  1. Five layers in total
  2. Eight layers in total
  3. Sixteen layers in total
  4. Three layers in total
Show answer and solution

Correct answer: B. Eight layers in total

**AlexNet has 8 learned layers:** 5 convolutional + 3 fully-connected. The final FC layer feeds a 1000-way softmax for ImageNet classification.

Open this question →

The nonlinearity used after every convolutional and FC layer is

  1. The ReLU activation function
  2. The hyperbolic tangent function
  3. The sigmoid logistic function
  4. The identity function with bias
Show answer and solution

Correct answer: A. The ReLU activation function

AlexNet was a major argument for ReLU over saturating nonlinearities. ReLU networks trained several times faster than equivalent tanh or sigmoid networks.

Open this question →

Which technique randomly zeroes hidden neurons during training?

  1. Weight decay on the L2 norm
  2. Stochastic depth of residual blocks
  3. Batch normalisation across the layer
  4. Dropout with probability 0.5
Show answer and solution

Correct answer: D. Dropout with probability 0.5

Dropout sets each hidden neuron's output to zero with probability 0.5 during training. It forces neurons to learn robust features that cooperate with random subsets of others.

Open this question →

How to practise

  1. Choose a AlexNet CNN ImageNet question and try to work out your answer.
  2. Open the question to check the correct answer and read its worked solution.
  3. Continue in the app for bite-sized lessons, explanations and review with Quill. You can start without creating an account.

More Aifoundations topics

Many Shot in Context LearningLinear Regression and Supervised LearningMath for AI LA ProbabilityInstructGPT RLHFMLP and ActivationsTransformer AttentionWhat Is AI Intro