机器学习原理

题集来源:Alireza Dirafzoon(alirezadir)收录项目:AIMLInterviews查看原题集

题目目录

42 道原题
  1. Can you explain the backpropagation algorithm and its role in training neural networks?

    原题
  2. What is a convolutional neural network (CNN) and how does it work?

    原题
  3. What is transfer learning and how is it used in practice?

    原题
  4. Walk through a transformer block. Why divide attention scores by √dₖ?

    原题
  5. What is the KV cache and why does it matter for inference? How does its memory scale?

    原题
  6. Compare MHA, MQA, and GQA. Why did Llama adopt GQA?

    原题
  7. What is RoPE and why are rotary embeddings preferred over learned positional embeddings?

    原题
  8. When would you choose RAG vs fine-tuning vs long-context?

    原题
  9. Compare SFT, DPO, GRPO, and RLVR — data requirements, what extra models each needs, and when to use each.

    原题
  10. Why does GRPO drop the critic network, and how does it estimate advantage?

    原题