Hiroki Naganuma

Schedule

直近の Schedule

[今月](https://hiroki11x.github.io/posts/diary/2023_11/) ``` - 直近の締め切り/予定 - ASAP IBM CoverLetter - 11/21 Apple Research Paris 面接 - 11/22 NVIDIA 面接 - 11/22 ICLR Rebuttal End 面接 - 11/27 CVPR supplementary and Arxiv (Timm OOD) - 11/31 Arxiv (SAM-Focal) - 11/31 LinkedIn 返事 - 12/15 阪大 公募課題締め切り - 1/? ICML x4 (F2 Metric? / SAM-Focal / SAM x Auto-Tuning) - 1/? TMLR x1 (IRM Calibration) - 2/? KDD x1 (Calibration CTR) - 3/? TMLR x1 (TIC) ```

• 一時帰国関連

一日の流れ

• AWS で作業

• CVPR の Supplementary 提出
  • Anonymize 対応
  • Github 更新
  • Appndix 更新
  • OpenReview で提出
  • Arxiv を更新
• Borealis AI の Fellowship の書類準備して提出 ref
  • CV
  • Cover Letter
  • Research Proposal
  • 推薦状をお願いした
    • 横田先生
    • Ioannis に推薦状依頼
• Apple Paris 面接準備
  • 11/17 にまとめたやつ
  • 今日は、Coding じゃなくて、General ML Interview かな
  • 色々調べた
    • MCQ: multiple choice question (多項選択式の問題)
    • From Rejection to Offer: My Apple SE Internship Interview Journey Amid Mass Layoffs.
      • 9- Apple — Software Engineer at Siri Information Intelligence(SII) or SaLT Team — Dec 22
      • 5 mins intro
      • 5 mins behavioral and basic ML technical question
      • 20 min overview of my most recent research paper publication about Apple’s ASR system Siri’s Biased against, Women, Black and accented speakers, and people with disabilities. The recruiter had so many questions about that paper. Understandably so. Right?!
    • I have an Apple interview coming up (NLP-Deep Learning internship). Do you have any ideas on how to prepare? The recruiter just informed me it is non-technical, but I don’t know what to expect.
      • Try to make sure you understand precision, accuracy, recall and the ROC curve. Look into different model validation methods and more such fundamentals like optimization, information[3]and probability theory.
    • Apple Machine Learning Engineer Interview Guide
      • DSA: Data Structure and Algorithms
      • Apple MLE Round 1: Phone Screening
        • Why do you want to join Apple?
        • Why do you think you will be a good fit for the role?
        • What responsibilities do you expect to have from your job at Apple?
      • Apple MLE Round 2: Technical Interviews
        • ML Rounds: In this round, the questions will be specific to the hiring team’s domain. For example, candidates might be asked to solve a probability calculation question based on dynamic programming or answer questions related to NLP. The purpose of this round is to assess the candidate’s ability to apply their technical skills to real-world problems in the specific domain that the hiring team is focused on.
        • Coding Rounds: In this round, candidates will be asked Leetcode questions and their DSA skills will be evaluated. For example, one candidate reported being asked to implement a simpler version of Naive Bayes or Association Rules. Another candidate reported that the hiring team sent a link for audio-related data processing for a coding challenge. The purpose of this round is to assess the candidate’s coding skills and ability to solve problems efficiently.
        • Interview Questions
          • What’s the BERT model and why is it good?
          • Explain the concept of dynamic programming and how it can be used in probability calculations?
          • How would you approach solving a natural language processing problem? Can you give an example of a problem you’ve solved in this area?
          • Implement a simpler version of Naive Bayes.
          • How would you implement Association Rules? Can you walk me through the steps?
          • Given a dataset, implement a program to process the audio data.
          • Can you explain your approach and any challenges you faced?
• 学ぶこと
  • Naive Bayes
  • Association Rule
  • 分類のための指標(PrecisionとRecall)の解説
    • モデルの出力全体に対して得られる指標 (全件中何件正解したか)
    • Precision: 予測したもののうち、実際にあたっていた割合
    • Recall: 真値のうち、ただしく予測された割合
  • 正解率・適合率・再現率・F 値
  • ROCとAUCを超わかりやすく解説(超重要!!)【機械学習入門22】
  • Accuracy/Recall/Precisionってややこしくない??
  • バイアス・バリアンス　分解

所感

• 休めない

TODO

• CoverLetter
• 現状だと内容が甘い

• vomit draft をかく

• Apple 対策を Execute

IceBox

• マイルストーン設計
• それぞれを共著者に共有
  • Curvature
  • SAM AutoTune
• 阪大の申請書

• Cyberdyne TODO を Execute

• ICML / Curvature (w/ Lyle, Ioannis)
  • 相関をさっさと計算する
• Notion Research TODO

Reference

• TODO - 09/18
• Funding / その他応募締切 - 09/18
• 研究プロジェクト　Notion

Deadline

年間スケジュール

```
11月中旬 CVPR 締め切り x1 (Timm OOD)
[NeurIPS, New Orleans, 12/1-5?]
1月下旬　ICML 締め切り (F2 Metric? / TIC / SAM-Focal / SAM x Auto-Tuning)
2月上旬  KDD 締め切り (LogitNorm)
```