ML
ML Interview Roadmap
← Study plan·Deep Learning

Day 45 of 133

Transfer learning for vision + DSA Tries finish

Freeze vs fine-tune; AutoAugment vs RandAugment.

DSA · NeetCode Tries

  • Word Search IIDSA · Tries

    Interview questions to prep

    1. Why a trie over a hash map — what queries does the trie make cheaper?
    2. What's the time/space trade-off vs storing all suffixes?
    3. How would you support deletion or wildcard matching efficiently?

DL · Transfer learning for vision

  • Feature extraction vs fine-tuningDeep Learningfast.ai

    Interview questions to prep

    1. When do you freeze the backbone vs fine-tune end-to-end?
    2. Why do you usually fine-tune with a lower learning rate than training from scratch?
  • Augmentation strategies (RandAugment, AutoAugment)Deep LearningRandAugment

    Interview questions to prep

    1. Compare AutoAugment vs RandAugment — what's the simplification?
    2. When does heavy augmentation hurt instead of help — what's the failure mode?

References & further reading