alleryStart learning →Image similarity in AlexNet's last hidden layer is measured by
ACosine similarity over raw pixel patches
BEuclidean distance over 4096-D feature vectors
CHamming distance over hashed token codes
DCross-entropy between predicted class distributions
Answer & Solution
Correct answer: B. Euclidean distance over 4096-D feature vectors
Images with similar L2 distance between their 4096-D last-FC-layer activations are considered semantically similar — even when their pixels differ substantially (e.g., dogs in different poses).
Keep practising AlexNet CNN ImageNet
Browse all AlexNet CNN ImageNet questions for Claude, or explore more Aifoundations topics.
Related questions
Fill in the AlexNet numbersThe SGD update rule for AlexNet usesRemoving any single middle convolutional layer of AlexNet costs roughlyAlexNet was trained for roughly how many epochs on ImageNet?At test time dropout is approximated byAlexNet uses overlapping pooling withAlexNet's second form of data augmentation isAlexNet's first form of data augmentation is