SlideShare a Scribd company logo
1 of 34
Download to read offline
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Challenges for Machine Learning
Systems toward Continuous
Improvement
IBIS 2019 機械学習工学セッション @ Nagoya
2019.11.22
Arm Treasure Data
Aki Ariga
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
[Sculley, 2015] より翻訳をして引用
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
●
●
●
https://sites.google.com/view/sig-mlse/wg
試行錯誤から本番適用にもっていくにあたり、継続的に改善を続
けていく上で適切な機械学習システムを作るため
- 論文になりづらいLesson Learntを収集し
- アーキテクチャパターンを体系化したい
(鷲崎先生の取り組み [Washizaki 2019]以外、まだ十分に整備されていない)
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
● Project process, organization structure/management
● Machine Learning systems for distributed training
● Hardware acceleration e.g. GPU, TPU, FPGA, etc...
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Typical steps for a Machine Learning project
1. ビジネス課題を
2. 類似の課題を、論文を中心にサーベイ
3. 方法を考える
4. システム設計
5. 学習データ(特徴量+ラベル)の設計
6. 実データの収集と前処理をする
7. 探索的データ分析とアルゴリズムの選定
8. 学習・パラメータチューニング
9. システムに組み込む
10. 予測精度・ビジネス指標をモニタリング
実験ループ:
5〜8を繰り返し
本番ループ:
8〜10を繰り返し
(4に戻ることも)
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
1. ビジネス課題を
2. 類似の課題を、論文を中心にサーベイ
3. 機械学習をしない方法を考える
4. システム設計
5. 学習
6. 実データの収集と前処理
7. 探索的データ分析とアルゴリズムの選定
8. 学習・パラメータチューニング
9. システムに組み込む
10. をモニタリング
1. を書く
2. コードを書く
3. Pull Request/CIでの
4. コードレビュー、マージ
5. コード/バイナリのビルド、デプロイ
6. を と
共にモニタリング
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
1. ビジネス課題を
2. 類似の課題を、論文を中心にサーベイ
3. 機械学習をしない方法を考える
4. システム設計
5. 学習
6. 実データの収集と前処理
7. 探索的データ分析とアルゴリズムの選定
8. 学習・パラメータチューニング
9. システムに組み込む
10. をモニタリング
1. を書く
2. コードを書く
3. Pull Request/CIでの
4. コードレビュー、マージ
5. コード/バイナリのビルド、デプロイ
6. を と
共にモニタリング
● に振る舞いが決まる
● 仕様を網羅して ができる
● エラーは一意に定義でき、コードの
ロジックを することができ
る
● 入力 に対して にに振る
舞いが決まる
● 仕様を網羅できず、
しかできない
● エラーかどうかの検出が難しく、モ
デル更新など しかでき
ない
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
● “Data defines everything”
● “Change Anything Changes Everything” [Sculley, 2015]
● データが振る舞いを決めるため、確率的にエラーを抑えることし
かできない
● 確率的な挙動をするため、変更の影響範囲を事前に予見できない
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
● した音声認識システム
○ リリース後、 に → 予測性能が劣化
● 工場の画像検査システムで、一ヶ月以上現場の画像で学習
○ が変化 → 予測性能が劣化
● ウェブサイトのランキングシステム
○ により検索語のトレンドが変遷 → 性能が劣化
[Batch] より翻訳をして引用
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
[Ziobaite, 2010]より引用
Production models should be
retrained continuously to avoid
concept drift
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
[CD4ML]より引用
モデルを本番に学習・デプロイし続けるには、
チームをまたいだ協力が必要
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
🤔
● : ”Infrastructure as Code” made Ops of Dev
○ which improves
● : Isolate DS/Researcher’s code from production system
○ which aims to improve
○ researcher’s code by Dev is
■ Those Dev can be DS/researcher and get high salary!
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
[Ariga] より
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Preprocessing
EDA
Training
Deploy/Serving
Audit
[Ariga] より
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
[Marsden, 2019]より翻訳して引用
●
○ 9ヶ月前に学習したモデルが全く同じ環境で、同じデータで再学習でき、
ほぼ同じ(数%以内の差)の精度を得られるべきである
●
○ 本番で稼働しているどのモデルも、作成時のパラメータと学習データ、
更に生データまでトレースできるべきである
●
○ 他の同僚の作ったモデルを本人に聞くことなく改善でき、
非同期で改善とコードやデータのマージができるべきである
●
○ 手動での作業0でモデルはデプロイできるべき。
統計的にモニタリングできるべき
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
1. Data/Model management
2. Experiment Tracking
3. Reproducible experiment
4. Pipeline management
5. ML framework abstraction
6. Model Serving/Deployment
7. Testing and quality check
8. Explaining model/data
9. Monitoring/Observability
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Reproducible / Accountable
Collaborative / Continuous1. Data/Model management
How do you manage 10000+ models
in production?
● Phase
○ EDA, Training, Audit
● Related concepts
○ Data/Model Versioning,
Data/Model Lineage,
Metadata management,
Feature store
● OSS
○ ModelDB, pachyderm, DVC,
ML Metadata, Feast
Example of Data Lineage. [Uber] より引用
EDA Training
Audit
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Reproducible / Accountable
Collaborative / Continuous2. Experiment tracking
What parameter/condition did you choose
with the production model?
● Phase
○ Training, Audit
● Related concepts
○ Parameter management, Artifact
management
Model tracking example
[MLFlow tracking]
より引用
● OSS
○ kubeflow pipelines, MLflow
tracking, polyaxon, comet.ml
Training Audit
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
3. Reproducible experiments
Reproducible / Accountable
Collaborative / Continuous
Is you notebook can reproducible in
your peer’s environment?
● Phase
○ EDA, Training, Audit
● Related concepts
○ Reproducible notebook,
○ Dependency management
● OSS
○ Jupyter notebook, Polynote,
Docker
https://twitter.com/keigohtr/status/1197321232800071680
もちろん、実装によるが...
EDA Training
Audit
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Reproducible / Accountable
Collaborative / Continuous4. Pipeline management
How do you manage consistent
end-to-end ML pipeline?
● Phase
○ Preprocessing, Training,
Deploy/Serving, Audit
● Related concepts
○ Pipeline versioning, Workflow,
CI
● OSS
○ kubeflow, Argo, ArgoCD,
MLFlow Projects, Tensorflow
Transform, [General workflow
engines]
Training
Audit
Preprocessing
Deploy/Serving
[Kubeflow pipeline] より引用
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Reproducible / Accountable
Collaborative / Continuous5. ML framework abstraction
How can Dev make DS’s code deployable
without understanding the code?
● Phase
○ Preprocessing, Training, Deploy
● Related concepts
○ Deploy with function/decorator,
Configuration based training,
AutoML
● OSS & Known frameworks
○ TensorFlow Transform,
Metaflow
○ (FBLeaner Flow, Overton,
Bighead, Boson, Metaflow)
TrainingPreprocessing
Deploy/Serving
Metaflow enables to deploy first models within a week for most
of projects [Metaflow]
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Overton: A Data System for Monitoring and Improving
Machine-Learned Products [Overton]
“domain engineers should not be
forced to write traditional deep learning
modeling code”
“Engineers are Comfortable with
Automatic Hyperparameter Tuning”
Apple implemented ML system with weak
supervision and slicing.
Engineers are required to
1) create/select schema and input payload.
2) Add slices or labeling functions, or synthetic
examples
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Reproducible / Accountable
Collaborative / Continuous6. Testing and quality check
What metrics should we track for
quality?
● Phase
○ Training, Deploy
● Related concepts
○ Data validation, Component
integration validation, Model
quality check, Adversarial
example detection
● OSS
○ TensorFlow Data Validation,
Deequ
Training Deploy/Serving
[TFDV] Understand data from stats
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Data validation: Testing toward train-serve skew
教師あり学習の予測モデルは、学
習データの分布に近いという仮定
学習データの仮定をvalidationす
れば良い
● Check with schema
○ Categorical variable
○ Numerical value range
○ Similarity of
distribution
[Baylor, 2017]より引用
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
7. Explaining model/data
Can you convince your boss why we
should choose this model?
● Phase
○ Training, Audit
● Related concepts
○ Model explainability,
Bias/fairness and ethics
check
● OSS
○ TensorFlow Model Analysis,
Facets, LIME, ELI5, SHAP,
Manifold
Reproducible / Accountable
Collaborative / Continuous
Training Audit
Manifold can compare two sliced population with multiple
models/features [Manifold]
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
“Explain what’s happening and why. There is still
significant fear, uncertainty and doubt (FUD) about AI. I have
seen that providing a basic education — along the lines of
the AI for Everyone curriculum — eases these conversations.
Other tactics including explainability, visualization,
rigorous testing, and auditing also help build trust in an AI
system and convince our customers (and ourselves!) that it
really works.”
[Batch 2] より引用。下線部は筆者強調
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
8. Model Serving/Deployment
How rapidly can you deploy a model to production?
● Phase
○ Deploy/Serving
● Related concepts
○ Rollout (Canary rollouts, A/B testing, Shadowing)
○ Model serialization/export
● OSS
○ TF Serving, Seldon, KFServing, MLFlow Model registry, Clipper
○ PMML, PFA, ONNX, Menoh
Reproducible / Accountable
Collaborative / Continuous
Deploy/Serving
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
“Time is money” - impact of serving latency [Bernardi 2019]
+30% latency reduces 0.5% CVR at
Booking.com
Techniques to reduce latency:
● Model Redundancy
● In-house developed Linear Prediction
engine
● Sparse models
● Precomputation and caching
● Bulking
● Minimum Feature Transformation
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
9. Monitoring/Observability
How do you notice production
model/data is corrupt?
● Phase
○ Deploy/Serving
● Related concepts
○ Data Validation, Outlier
detection, Concept drift
detection, Delayed feedback
● OSS
○ [General Monitoring tools],
[General Workflow engine],
TFDV
Reproducible / Accountable
Collaborative / Continuous
Deploy/Serving
“Smooth bimodal distributions with one clear
stable point are signs of a model that successfully
distinguishes two classes” [Bernardi 2019]
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
Challenges Phase
Reprod
ucible
Accou
ntable
Collab
orative
Contin
uous
Data/Model management EDA/Training/Audit ✅ ✅ ✅
Experiment Tracking Training/Audit ✅ ✅ ✅
Reproducible experiment EDA/Training/Audit ✅ ✅ ✅
Pipeline management
Preprocessing/Training/S
erving/Deploy/Audit ✅ ✅ ✅ ✅
ML framework abstraction
Preprocessing/Training/D
eploy ✅ ✅ ✅
Testing and quality check Training/Deploy ✅ ✅
Explaining model/data Training/Audit ✅
Model Serving/Deployment Serving/Deploy ✅
Monitoring/Observability Serving/Deploy ✅ ✅ ✅
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
● Introduced the importance of Continuous Improvement for machine learning
system
● Introduced 9 Challenges for ML systems in production
● Let’s join MLSE slack and collect/share knowledge on GitHub Wiki!
https://github.com/chezou/ml_in_production/wiki
○ Planning for the Workshop for Machine Learning system in production
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
● https://github.com/EthicalML/awesome-production-machine-learning
● https://hackernoon.com/why-is-devops-for-machine-learning-so-different-384z32f1
● https://martinfowler.com/articles/cd4ml.html
Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved.
● [Sculley, 2015] Sculley, David, et al. "Hidden technical debt in machine learning systems." Advances in neural information processing systems.
2015.
● [Washizaki 2019] Washizaki, Hironori, et al. "Studying Software Engineering Patterns for Designing Machine Learning Systems." arXiv
preprint arXiv:1910.04736 (2019).
● [Batch] The Batch Nov. 6, 2019
https://info.deeplearning.ai/the-batch-deepmind-masters-starcraft-2-ai-attacks-on-amazon-a-career-in-robot-management-banks-embrace-b
ots-1
● [Batch 2] The Batch Nov. 20, 2019
https://info.deeplearning.ai/the-batch-artificial-noses-surveillance-on-wheels-unwelcome-researchers-privacy-problems-beyond-
bounding-boxes
● [Kubeflow pipeline] https://www.kubeflow.org/docs/pipelines/overview/pipelines-overview/
● [TFDV] https://www.tensorflow.org/tfx/data_validation/get_started
● [Zliobaite, 2010] Žliobaitė, Indrė. "Learning under concept drift: an overview." arXiv preprint arXiv:1010.4784 (2010).
● [Marsden, 2019] Marsden, Luke. "The Future of MLOps"
https://docs.google.com/presentation/d/17RWqPH8nIpwG-jID_UeZBCaQKoz4LVk1MLULrZdyNCs/edit#slide=id.p
● [Baylor, 2017] D. Baylor et al., “TFX: A tensorflow-based production-scale machine learning platform,” in Proceedings of the 23rd acm
sigkdd international conference on knowledge discovery and data mining, 2017, pp. 1387–1395
● [CD4ML] D. Sato, et al. “Continuous Delivery for Machine Learning” , https://martinfowler.com/articles/cd4ml.html
● [Uber] “Databook: Turning big data into knowledge with metadata at uber.” https://eng.uber.com/databook/
● [Manifold] “Manifold: A Model-Agnostic Visual Debugging Tool for Machine Learning at Uber” https://eng.uber.com/manifold/
● [Overton] Ré, Christopher & Niu, Feng & Gudipati, Pallavi & Srisuwananukorn, Charles. (2019). Overton: A Data System for Monitoring and
Improving Machine-Learned Products.
● [Bernardi 2019] Bernardi, Lucas, et al. "150 Successful Machine Learning Models: 6 Lessons Learned at Booking. com." Proceedings of
the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. ACM, 2019.
● [Metaflow] Julie Pitt, et al. “A Human-Friendly Approach to MLOps”, MLOps NYC19, https://youtu.be/fOSZuONmLbA
● [Ariga] “MLOpsの歩き方”, n月刊ラムダノートvol.1 no.1, 2019
Confidential © Arm 2017Confidential © Arm 2017Confidential © Arm 2019
Thank You!
Danke!
Merci!
谢谢!
ありがとう!
Gracias!
Kiitos!

More Related Content

What's hot

What's hot (20)

Ml system in_python
Ml system in_pythonMl system in_python
Ml system in_python
 
Kaggle Happywhaleコンペ優勝解法でのOptuna使用事例 - 2022/12/10 Optuna Meetup #2
Kaggle Happywhaleコンペ優勝解法でのOptuna使用事例 - 2022/12/10 Optuna Meetup #2Kaggle Happywhaleコンペ優勝解法でのOptuna使用事例 - 2022/12/10 Optuna Meetup #2
Kaggle Happywhaleコンペ優勝解法でのOptuna使用事例 - 2022/12/10 Optuna Meetup #2
 
AIのラボからロボティクスへ --- 東大松尾研究室のWRS2020パートナーロボットチャレンジへの挑戦
AIのラボからロボティクスへ --- 東大松尾研究室のWRS2020パートナーロボットチャレンジへの挑戦AIのラボからロボティクスへ --- 東大松尾研究室のWRS2020パートナーロボットチャレンジへの挑戦
AIのラボからロボティクスへ --- 東大松尾研究室のWRS2020パートナーロボットチャレンジへの挑戦
 
MLOps入門
MLOps入門MLOps入門
MLOps入門
 
(修正)機械学習デザインパターン(ML Design Patterns)の解説
(修正)機械学習デザインパターン(ML Design Patterns)の解説(修正)機械学習デザインパターン(ML Design Patterns)の解説
(修正)機械学習デザインパターン(ML Design Patterns)の解説
 
MLOps に基づく AI/ML 実運用最前線 ~画像、動画データにおける MLOps 事例のご紹介~(映像情報メディア学会2021年冬季大会企画セッショ...
MLOps に基づく AI/ML 実運用最前線 ~画像、動画データにおける MLOps 事例のご紹介~(映像情報メディア学会2021年冬季大会企画セッショ...MLOps に基づく AI/ML 実運用最前線 ~画像、動画データにおける MLOps 事例のご紹介~(映像情報メディア学会2021年冬季大会企画セッショ...
MLOps に基づく AI/ML 実運用最前線 ~画像、動画データにおける MLOps 事例のご紹介~(映像情報メディア学会2021年冬季大会企画セッショ...
 
東大大学院 電子情報学特論講義資料「深層学習概論と理論解析の課題」大野健太
東大大学院 電子情報学特論講義資料「深層学習概論と理論解析の課題」大野健太東大大学院 電子情報学特論講義資料「深層学習概論と理論解析の課題」大野健太
東大大学院 電子情報学特論講義資料「深層学習概論と理論解析の課題」大野健太
 
機械学習で泣かないためのコード設計 2018
機械学習で泣かないためのコード設計 2018機械学習で泣かないためのコード設計 2018
機械学習で泣かないためのコード設計 2018
 
先駆者に学ぶ MLOpsの実際
先駆者に学ぶ MLOpsの実際先駆者に学ぶ MLOpsの実際
先駆者に学ぶ MLOpsの実際
 
機械学習デザインパターンおよび機械学習システムの品質保証の取り組み
機械学習デザインパターンおよび機械学習システムの品質保証の取り組み機械学習デザインパターンおよび機械学習システムの品質保証の取り組み
機械学習デザインパターンおよび機械学習システムの品質保証の取り組み
 
プロジェクトマネージャのための機械学習工学入門
プロジェクトマネージャのための機械学習工学入門プロジェクトマネージャのための機械学習工学入門
プロジェクトマネージャのための機械学習工学入門
 
Direct Sparse Odometryの解説
Direct Sparse Odometryの解説Direct Sparse Odometryの解説
Direct Sparse Odometryの解説
 
ブレインパッドにおける機械学習プロジェクトの進め方
ブレインパッドにおける機械学習プロジェクトの進め方ブレインパッドにおける機械学習プロジェクトの進め方
ブレインパッドにおける機械学習プロジェクトの進め方
 
機械学習の力を引き出すための依存性管理
機械学習の力を引き出すための依存性管理機械学習の力を引き出すための依存性管理
機械学習の力を引き出すための依存性管理
 
ゼロから始める転移学習
ゼロから始める転移学習ゼロから始める転移学習
ゼロから始める転移学習
 
全力解説!Transformer
全力解説!Transformer全力解説!Transformer
全力解説!Transformer
 
Optimizer入門&最新動向
Optimizer入門&最新動向Optimizer入門&最新動向
Optimizer入門&最新動向
 
2018年01月27日 TensorBoardによる学習の可視化
2018年01月27日 TensorBoardによる学習の可視化2018年01月27日 TensorBoardによる学習の可視化
2018年01月27日 TensorBoardによる学習の可視化
 
BigQuery MLの行列分解モデルを 用いた推薦システムの基礎
BigQuery MLの行列分解モデルを 用いた推薦システムの基礎BigQuery MLの行列分解モデルを 用いた推薦システムの基礎
BigQuery MLの行列分解モデルを 用いた推薦システムの基礎
 
「機械学習:技術的負債の高利子クレジットカード」のまとめ
「機械学習:技術的負債の高利子クレジットカード」のまとめ「機械学習:技術的負債の高利子クレジットカード」のまとめ
「機械学習:技術的負債の高利子クレジットカード」のまとめ
 

Similar to Challenges for machine learning systems toward continuous improvement

The Strategic Role of the Enterprise Application Framework
The Strategic Role of the Enterprise Application FrameworkThe Strategic Role of the Enterprise Application Framework
The Strategic Role of the Enterprise Application Framework
Jean-Marc Desvaux
 
Mobile to Mainframe - the Challenges of Enterprise DevOps Adoption
Mobile to Mainframe - the Challenges of Enterprise DevOps AdoptionMobile to Mainframe - the Challenges of Enterprise DevOps Adoption
Mobile to Mainframe - the Challenges of Enterprise DevOps Adoption
Sanjeev Sharma
 
IBM Rational Developer for System z Quick Start Sales Presentation
IBM Rational Developer for System z Quick Start Sales PresentationIBM Rational Developer for System z Quick Start Sales Presentation
IBM Rational Developer for System z Quick Start Sales Presentation
IBM Rational software
 

Similar to Challenges for machine learning systems toward continuous improvement (20)

20210428 - Sustainable Engineering practices & API Communities: Adoption Best...
20210428 - Sustainable Engineering practices & API Communities: Adoption Best...20210428 - Sustainable Engineering practices & API Communities: Adoption Best...
20210428 - Sustainable Engineering practices & API Communities: Adoption Best...
 
Datarobot, 자동화된 분석 적용 시 분석 절차의 변화 및 효용 - 홍운표 데이터 사이언티스트, DataRobot :: AWS Sum...
Datarobot, 자동화된 분석 적용 시 분석 절차의 변화 및 효용 - 홍운표 데이터 사이언티스트, DataRobot :: AWS Sum...Datarobot, 자동화된 분석 적용 시 분석 절차의 변화 및 효용 - 홍운표 데이터 사이언티스트, DataRobot :: AWS Sum...
Datarobot, 자동화된 분석 적용 시 분석 절차의 변화 및 효용 - 홍운표 데이터 사이언티스트, DataRobot :: AWS Sum...
 
Data meets AI - ATP Roadshow India
Data meets AI - ATP Roadshow IndiaData meets AI - ATP Roadshow India
Data meets AI - ATP Roadshow India
 
What We're Learning Adopting Spring Boot and PCF for Dell.com's eCommerce
What We're Learning Adopting Spring Boot and PCF for Dell.com's eCommerceWhat We're Learning Adopting Spring Boot and PCF for Dell.com's eCommerce
What We're Learning Adopting Spring Boot and PCF for Dell.com's eCommerce
 
SRE (service reliability engineer) on big DevOps platform running on the clou...
SRE (service reliability engineer) on big DevOps platform running on the clou...SRE (service reliability engineer) on big DevOps platform running on the clou...
SRE (service reliability engineer) on big DevOps platform running on the clou...
 
Pre-Con Education: Introduction to Mainframe Academy With CA Technologies
Pre-Con Education: Introduction to Mainframe Academy With CA TechnologiesPre-Con Education: Introduction to Mainframe Academy With CA Technologies
Pre-Con Education: Introduction to Mainframe Academy With CA Technologies
 
How_to_build_your_cloud_enablement_engine_with_the_people_you_already_have
How_to_build_your_cloud_enablement_engine_with_the_people_you_already_haveHow_to_build_your_cloud_enablement_engine_with_the_people_you_already_have
How_to_build_your_cloud_enablement_engine_with_the_people_you_already_have
 
The Strategic Role of the Enterprise Application Framework
The Strategic Role of the Enterprise Application FrameworkThe Strategic Role of the Enterprise Application Framework
The Strategic Role of the Enterprise Application Framework
 
Empowering your Process Automation with Machine Learning
Empowering your Process Automation with Machine LearningEmpowering your Process Automation with Machine Learning
Empowering your Process Automation with Machine Learning
 
Methods Over Madness 2003 Ver.
Methods Over Madness 2003 Ver.Methods Over Madness 2003 Ver.
Methods Over Madness 2003 Ver.
 
Train once, deploy anywhere on the cloud and at the edge with Neo - AIM301 - ...
Train once, deploy anywhere on the cloud and at the edge with Neo - AIM301 - ...Train once, deploy anywhere on the cloud and at the edge with Neo - AIM301 - ...
Train once, deploy anywhere on the cloud and at the edge with Neo - AIM301 - ...
 
IBM RACEv - Ferramenta para análise técnico e financeira para Plataforma Main...
IBM RACEv - Ferramenta para análise técnico e financeira para Plataforma Main...IBM RACEv - Ferramenta para análise técnico e financeira para Plataforma Main...
IBM RACEv - Ferramenta para análise técnico e financeira para Plataforma Main...
 
Mobile to Mainframe - the Challenges of Enterprise DevOps Adoption
Mobile to Mainframe - the Challenges of Enterprise DevOps AdoptionMobile to Mainframe - the Challenges of Enterprise DevOps Adoption
Mobile to Mainframe - the Challenges of Enterprise DevOps Adoption
 
Good Design is Good Business: Business Design with RSA and SA
Good Design is Good Business: Business Design with RSA and SAGood Design is Good Business: Business Design with RSA and SA
Good Design is Good Business: Business Design with RSA and SA
 
IBM Rational Developer for System z Quick Start Sales Presentation
IBM Rational Developer for System z Quick Start Sales PresentationIBM Rational Developer for System z Quick Start Sales Presentation
IBM Rational Developer for System z Quick Start Sales Presentation
 
Kochi Mulesoft Meetup #10 - MuleSoft Composer: Connect apps and data easily w...
Kochi Mulesoft Meetup #10 - MuleSoft Composer: Connect apps and data easily w...Kochi Mulesoft Meetup #10 - MuleSoft Composer: Connect apps and data easily w...
Kochi Mulesoft Meetup #10 - MuleSoft Composer: Connect apps and data easily w...
 
mulecomposer.pdf
mulecomposer.pdfmulecomposer.pdf
mulecomposer.pdf
 
Global Squad Leader - Resume
Global Squad Leader - ResumeGlobal Squad Leader - Resume
Global Squad Leader - Resume
 
Building a Mass Migration Program Moving Through Vision to Experiment and Sca...
Building a Mass Migration Program Moving Through Vision to Experiment and Sca...Building a Mass Migration Program Moving Through Vision to Experiment and Sca...
Building a Mass Migration Program Moving Through Vision to Experiment and Sca...
 
Drive Digital Transformation using Machine Learning
Drive Digital Transformation using Machine LearningDrive Digital Transformation using Machine Learning
Drive Digital Transformation using Machine Learning
 

More from Aki Ariga

Julia 100 exercises #JuliaTokyo
Julia 100 exercises #JuliaTokyoJulia 100 exercises #JuliaTokyo
Julia 100 exercises #JuliaTokyo
Aki Ariga
 

More from Aki Ariga (20)

Managing Machine Learning workflows on Treasure Data
Managing Machine Learning workflows on Treasure DataManaging Machine Learning workflows on Treasure Data
Managing Machine Learning workflows on Treasure Data
 
仕事ではじめる機械学習
仕事ではじめる機械学習仕事ではじめる機械学習
仕事ではじめる機械学習
 
主人が外資系IT企業に転職して4ヶ月が過ぎました
主人が外資系IT企業に転職して4ヶ月が過ぎました主人が外資系IT企業に転職して4ヶ月が過ぎました
主人が外資系IT企業に転職して4ヶ月が過ぎました
 
R&D at Foodtech company - #CookpadTechConf 2016
R&D at Foodtech company - #CookpadTechConf 2016R&D at Foodtech company - #CookpadTechConf 2016
R&D at Foodtech company - #CookpadTechConf 2016
 
Why I started Machine Learning Casual Talks? #MLCT
Why I started Machine Learning Casual Talks? #MLCTWhy I started Machine Learning Casual Talks? #MLCT
Why I started Machine Learning Casual Talks? #MLCT
 
クックパッドサマーインターン2015 機械学習・自然言語処理 実習課題
クックパッドサマーインターン2015 機械学習・自然言語処理 実習課題クックパッドサマーインターン2015 機械学習・自然言語処理 実習課題
クックパッドサマーインターン2015 機械学習・自然言語処理 実習課題
 
Rubyistがgemの前にPypiデビューするのは間違っているだろうか
Rubyistがgemの前にPypiデビューするのは間違っているだろうかRubyistがgemの前にPypiデビューするのは間違っているだろうか
Rubyistがgemの前にPypiデビューするのは間違っているだろうか
 
Machine Learning Casual Talks Intro #MLCT
Machine Learning Casual Talks Intro #MLCTMachine Learning Casual Talks Intro #MLCT
Machine Learning Casual Talks Intro #MLCT
 
Make Julia more popular in Japan!!1 #JuliaTokyo
Make Julia more popular in Japan!!1 #JuliaTokyoMake Julia more popular in Japan!!1 #JuliaTokyo
Make Julia more popular in Japan!!1 #JuliaTokyo
 
Refrection of kawasaki.rb
Refrection of kawasaki.rbRefrection of kawasaki.rb
Refrection of kawasaki.rb
 
Introduction and benchmarking of MeCab.jl #JapanR
Introduction and benchmarking of MeCab.jl  #JapanRIntroduction and benchmarking of MeCab.jl  #JapanR
Introduction and benchmarking of MeCab.jl #JapanR
 
Recommendation for iruby #tqrk08
Recommendation for iruby #tqrk08Recommendation for iruby #tqrk08
Recommendation for iruby #tqrk08
 
The book that changed me
The book that changed meThe book that changed me
The book that changed me
 
Introduction of Mecab.jl #JuliaTokyo
Introduction of Mecab.jl #JuliaTokyoIntroduction of Mecab.jl #JuliaTokyo
Introduction of Mecab.jl #JuliaTokyo
 
Introduction to Kanagawa Ruby Kaigi01 #kana01
Introduction to Kanagawa Ruby Kaigi01 #kana01Introduction to Kanagawa Ruby Kaigi01 #kana01
Introduction to Kanagawa Ruby Kaigi01 #kana01
 
Julia 100 exercises #JuliaTokyo
Julia 100 exercises #JuliaTokyoJulia 100 exercises #JuliaTokyo
Julia 100 exercises #JuliaTokyo
 
Machine Learning Casual Talks opening talk
Machine Learning Casual Talks opening talkMachine Learning Casual Talks opening talk
Machine Learning Casual Talks opening talk
 
Gong anyware
Gong anywareGong anyware
Gong anyware
 
gsub with ActiveSupport::SafeBuffer
gsub with ActiveSupport::SafeBuffergsub with ActiveSupport::SafeBuffer
gsub with ActiveSupport::SafeBuffer
 
はじめて翻訳記事を書いたら300ブクマ超えた話
はじめて翻訳記事を書いたら300ブクマ超えた話はじめて翻訳記事を書いたら300ブクマ超えた話
はじめて翻訳記事を書いたら300ブクマ超えた話
 

Recently uploaded

UNIT-V FMM.HYDRAULIC TURBINE - Construction and working
UNIT-V FMM.HYDRAULIC TURBINE - Construction and workingUNIT-V FMM.HYDRAULIC TURBINE - Construction and working
UNIT-V FMM.HYDRAULIC TURBINE - Construction and working
rknatarajan
 
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
9953056974 Low Rate Call Girls In Saket, Delhi NCR
 
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
dollysharma2066
 
result management system report for college project
result management system report for college projectresult management system report for college project
result management system report for college project
Tonystark477637
 

Recently uploaded (20)

The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
 
UNIT-V FMM.HYDRAULIC TURBINE - Construction and working
UNIT-V FMM.HYDRAULIC TURBINE - Construction and workingUNIT-V FMM.HYDRAULIC TURBINE - Construction and working
UNIT-V FMM.HYDRAULIC TURBINE - Construction and working
 
Vivazz, Mieres Social Housing Design Spain
Vivazz, Mieres Social Housing Design SpainVivazz, Mieres Social Housing Design Spain
Vivazz, Mieres Social Housing Design Spain
 
Double rodded leveling 1 pdf activity 01
Double rodded leveling 1 pdf activity 01Double rodded leveling 1 pdf activity 01
Double rodded leveling 1 pdf activity 01
 
CCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete Record
CCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete RecordCCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete Record
CCS335 _ Neural Networks and Deep Learning Laboratory_Lab Complete Record
 
Call for Papers - International Journal of Intelligent Systems and Applicatio...
Call for Papers - International Journal of Intelligent Systems and Applicatio...Call for Papers - International Journal of Intelligent Systems and Applicatio...
Call for Papers - International Journal of Intelligent Systems and Applicatio...
 
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
 
Unit 1 - Soil Classification and Compaction.pdf
Unit 1 - Soil Classification and Compaction.pdfUnit 1 - Soil Classification and Compaction.pdf
Unit 1 - Soil Classification and Compaction.pdf
 
UNIT-IFLUID PROPERTIES & FLOW CHARACTERISTICS
UNIT-IFLUID PROPERTIES & FLOW CHARACTERISTICSUNIT-IFLUID PROPERTIES & FLOW CHARACTERISTICS
UNIT-IFLUID PROPERTIES & FLOW CHARACTERISTICS
 
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
 
data_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdfdata_management_and _data_science_cheat_sheet.pdf
data_management_and _data_science_cheat_sheet.pdf
 
chapter 5.pptx: drainage and irrigation engineering
chapter 5.pptx: drainage and irrigation engineeringchapter 5.pptx: drainage and irrigation engineering
chapter 5.pptx: drainage and irrigation engineering
 
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
 
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
 
Online banking management system project.pdf
Online banking management system project.pdfOnline banking management system project.pdf
Online banking management system project.pdf
 
UNIT-III FMM. DIMENSIONAL ANALYSIS
UNIT-III FMM.        DIMENSIONAL ANALYSISUNIT-III FMM.        DIMENSIONAL ANALYSIS
UNIT-III FMM. DIMENSIONAL ANALYSIS
 
Generative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTGenerative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPT
 
Intze Overhead Water Tank Design by Working Stress - IS Method.pdf
Intze Overhead Water Tank  Design by Working Stress - IS Method.pdfIntze Overhead Water Tank  Design by Working Stress - IS Method.pdf
Intze Overhead Water Tank Design by Working Stress - IS Method.pdf
 
Coefficient of Thermal Expansion and their Importance.pptx
Coefficient of Thermal Expansion and their Importance.pptxCoefficient of Thermal Expansion and their Importance.pptx
Coefficient of Thermal Expansion and their Importance.pptx
 
result management system report for college project
result management system report for college projectresult management system report for college project
result management system report for college project
 

Challenges for machine learning systems toward continuous improvement

  • 1. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Challenges for Machine Learning Systems toward Continuous Improvement IBIS 2019 機械学習工学セッション @ Nagoya 2019.11.22 Arm Treasure Data Aki Ariga
  • 2. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. [Sculley, 2015] より翻訳をして引用
  • 3. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. ● ● ● https://sites.google.com/view/sig-mlse/wg 試行錯誤から本番適用にもっていくにあたり、継続的に改善を続 けていく上で適切な機械学習システムを作るため - 論文になりづらいLesson Learntを収集し - アーキテクチャパターンを体系化したい (鷲崎先生の取り組み [Washizaki 2019]以外、まだ十分に整備されていない)
  • 4. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. ● Project process, organization structure/management ● Machine Learning systems for distributed training ● Hardware acceleration e.g. GPU, TPU, FPGA, etc...
  • 5. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Typical steps for a Machine Learning project 1. ビジネス課題を 2. 類似の課題を、論文を中心にサーベイ 3. 方法を考える 4. システム設計 5. 学習データ(特徴量+ラベル)の設計 6. 実データの収集と前処理をする 7. 探索的データ分析とアルゴリズムの選定 8. 学習・パラメータチューニング 9. システムに組み込む 10. 予測精度・ビジネス指標をモニタリング 実験ループ: 5〜8を繰り返し 本番ループ: 8〜10を繰り返し (4に戻ることも)
  • 6. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. 1. ビジネス課題を 2. 類似の課題を、論文を中心にサーベイ 3. 機械学習をしない方法を考える 4. システム設計 5. 学習 6. 実データの収集と前処理 7. 探索的データ分析とアルゴリズムの選定 8. 学習・パラメータチューニング 9. システムに組み込む 10. をモニタリング 1. を書く 2. コードを書く 3. Pull Request/CIでの 4. コードレビュー、マージ 5. コード/バイナリのビルド、デプロイ 6. を と 共にモニタリング
  • 7. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. 1. ビジネス課題を 2. 類似の課題を、論文を中心にサーベイ 3. 機械学習をしない方法を考える 4. システム設計 5. 学習 6. 実データの収集と前処理 7. 探索的データ分析とアルゴリズムの選定 8. 学習・パラメータチューニング 9. システムに組み込む 10. をモニタリング 1. を書く 2. コードを書く 3. Pull Request/CIでの 4. コードレビュー、マージ 5. コード/バイナリのビルド、デプロイ 6. を と 共にモニタリング ● に振る舞いが決まる ● 仕様を網羅して ができる ● エラーは一意に定義でき、コードの ロジックを することができ る ● 入力 に対して にに振る 舞いが決まる ● 仕様を網羅できず、 しかできない ● エラーかどうかの検出が難しく、モ デル更新など しかでき ない
  • 8. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. ● “Data defines everything” ● “Change Anything Changes Everything” [Sculley, 2015] ● データが振る舞いを決めるため、確率的にエラーを抑えることし かできない ● 確率的な挙動をするため、変更の影響範囲を事前に予見できない
  • 9. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. ● した音声認識システム ○ リリース後、 に → 予測性能が劣化 ● 工場の画像検査システムで、一ヶ月以上現場の画像で学習 ○ が変化 → 予測性能が劣化 ● ウェブサイトのランキングシステム ○ により検索語のトレンドが変遷 → 性能が劣化 [Batch] より翻訳をして引用
  • 10. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. [Ziobaite, 2010]より引用 Production models should be retrained continuously to avoid concept drift
  • 11. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. [CD4ML]より引用 モデルを本番に学習・デプロイし続けるには、 チームをまたいだ協力が必要
  • 12. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. 🤔 ● : ”Infrastructure as Code” made Ops of Dev ○ which improves ● : Isolate DS/Researcher’s code from production system ○ which aims to improve ○ researcher’s code by Dev is ■ Those Dev can be DS/researcher and get high salary!
  • 13. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. [Ariga] より
  • 14. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Preprocessing EDA Training Deploy/Serving Audit [Ariga] より
  • 15. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. [Marsden, 2019]より翻訳して引用 ● ○ 9ヶ月前に学習したモデルが全く同じ環境で、同じデータで再学習でき、 ほぼ同じ(数%以内の差)の精度を得られるべきである ● ○ 本番で稼働しているどのモデルも、作成時のパラメータと学習データ、 更に生データまでトレースできるべきである ● ○ 他の同僚の作ったモデルを本人に聞くことなく改善でき、 非同期で改善とコードやデータのマージができるべきである ● ○ 手動での作業0でモデルはデプロイできるべき。 統計的にモニタリングできるべき
  • 16. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. 1. Data/Model management 2. Experiment Tracking 3. Reproducible experiment 4. Pipeline management 5. ML framework abstraction 6. Model Serving/Deployment 7. Testing and quality check 8. Explaining model/data 9. Monitoring/Observability
  • 17. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Reproducible / Accountable Collaborative / Continuous1. Data/Model management How do you manage 10000+ models in production? ● Phase ○ EDA, Training, Audit ● Related concepts ○ Data/Model Versioning, Data/Model Lineage, Metadata management, Feature store ● OSS ○ ModelDB, pachyderm, DVC, ML Metadata, Feast Example of Data Lineage. [Uber] より引用 EDA Training Audit
  • 18. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Reproducible / Accountable Collaborative / Continuous2. Experiment tracking What parameter/condition did you choose with the production model? ● Phase ○ Training, Audit ● Related concepts ○ Parameter management, Artifact management Model tracking example [MLFlow tracking] より引用 ● OSS ○ kubeflow pipelines, MLflow tracking, polyaxon, comet.ml Training Audit
  • 19. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. 3. Reproducible experiments Reproducible / Accountable Collaborative / Continuous Is you notebook can reproducible in your peer’s environment? ● Phase ○ EDA, Training, Audit ● Related concepts ○ Reproducible notebook, ○ Dependency management ● OSS ○ Jupyter notebook, Polynote, Docker https://twitter.com/keigohtr/status/1197321232800071680 もちろん、実装によるが... EDA Training Audit
  • 20. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Reproducible / Accountable Collaborative / Continuous4. Pipeline management How do you manage consistent end-to-end ML pipeline? ● Phase ○ Preprocessing, Training, Deploy/Serving, Audit ● Related concepts ○ Pipeline versioning, Workflow, CI ● OSS ○ kubeflow, Argo, ArgoCD, MLFlow Projects, Tensorflow Transform, [General workflow engines] Training Audit Preprocessing Deploy/Serving [Kubeflow pipeline] より引用
  • 21. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Reproducible / Accountable Collaborative / Continuous5. ML framework abstraction How can Dev make DS’s code deployable without understanding the code? ● Phase ○ Preprocessing, Training, Deploy ● Related concepts ○ Deploy with function/decorator, Configuration based training, AutoML ● OSS & Known frameworks ○ TensorFlow Transform, Metaflow ○ (FBLeaner Flow, Overton, Bighead, Boson, Metaflow) TrainingPreprocessing Deploy/Serving Metaflow enables to deploy first models within a week for most of projects [Metaflow]
  • 22. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Overton: A Data System for Monitoring and Improving Machine-Learned Products [Overton] “domain engineers should not be forced to write traditional deep learning modeling code” “Engineers are Comfortable with Automatic Hyperparameter Tuning” Apple implemented ML system with weak supervision and slicing. Engineers are required to 1) create/select schema and input payload. 2) Add slices or labeling functions, or synthetic examples
  • 23. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Reproducible / Accountable Collaborative / Continuous6. Testing and quality check What metrics should we track for quality? ● Phase ○ Training, Deploy ● Related concepts ○ Data validation, Component integration validation, Model quality check, Adversarial example detection ● OSS ○ TensorFlow Data Validation, Deequ Training Deploy/Serving [TFDV] Understand data from stats
  • 24. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Data validation: Testing toward train-serve skew 教師あり学習の予測モデルは、学 習データの分布に近いという仮定 学習データの仮定をvalidationす れば良い ● Check with schema ○ Categorical variable ○ Numerical value range ○ Similarity of distribution [Baylor, 2017]より引用
  • 25. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. 7. Explaining model/data Can you convince your boss why we should choose this model? ● Phase ○ Training, Audit ● Related concepts ○ Model explainability, Bias/fairness and ethics check ● OSS ○ TensorFlow Model Analysis, Facets, LIME, ELI5, SHAP, Manifold Reproducible / Accountable Collaborative / Continuous Training Audit Manifold can compare two sliced population with multiple models/features [Manifold]
  • 26. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. “Explain what’s happening and why. There is still significant fear, uncertainty and doubt (FUD) about AI. I have seen that providing a basic education — along the lines of the AI for Everyone curriculum — eases these conversations. Other tactics including explainability, visualization, rigorous testing, and auditing also help build trust in an AI system and convince our customers (and ourselves!) that it really works.” [Batch 2] より引用。下線部は筆者強調
  • 27. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. 8. Model Serving/Deployment How rapidly can you deploy a model to production? ● Phase ○ Deploy/Serving ● Related concepts ○ Rollout (Canary rollouts, A/B testing, Shadowing) ○ Model serialization/export ● OSS ○ TF Serving, Seldon, KFServing, MLFlow Model registry, Clipper ○ PMML, PFA, ONNX, Menoh Reproducible / Accountable Collaborative / Continuous Deploy/Serving
  • 28. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. “Time is money” - impact of serving latency [Bernardi 2019] +30% latency reduces 0.5% CVR at Booking.com Techniques to reduce latency: ● Model Redundancy ● In-house developed Linear Prediction engine ● Sparse models ● Precomputation and caching ● Bulking ● Minimum Feature Transformation
  • 29. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. 9. Monitoring/Observability How do you notice production model/data is corrupt? ● Phase ○ Deploy/Serving ● Related concepts ○ Data Validation, Outlier detection, Concept drift detection, Delayed feedback ● OSS ○ [General Monitoring tools], [General Workflow engine], TFDV Reproducible / Accountable Collaborative / Continuous Deploy/Serving “Smooth bimodal distributions with one clear stable point are signs of a model that successfully distinguishes two classes” [Bernardi 2019]
  • 30. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. Challenges Phase Reprod ucible Accou ntable Collab orative Contin uous Data/Model management EDA/Training/Audit ✅ ✅ ✅ Experiment Tracking Training/Audit ✅ ✅ ✅ Reproducible experiment EDA/Training/Audit ✅ ✅ ✅ Pipeline management Preprocessing/Training/S erving/Deploy/Audit ✅ ✅ ✅ ✅ ML framework abstraction Preprocessing/Training/D eploy ✅ ✅ ✅ Testing and quality check Training/Deploy ✅ ✅ Explaining model/data Training/Audit ✅ Model Serving/Deployment Serving/Deploy ✅ Monitoring/Observability Serving/Deploy ✅ ✅ ✅
  • 31. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. ● Introduced the importance of Continuous Improvement for machine learning system ● Introduced 9 Challenges for ML systems in production ● Let’s join MLSE slack and collect/share knowledge on GitHub Wiki! https://github.com/chezou/ml_in_production/wiki ○ Planning for the Workshop for Machine Learning system in production
  • 32. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. ● https://github.com/EthicalML/awesome-production-machine-learning ● https://hackernoon.com/why-is-devops-for-machine-learning-so-different-384z32f1 ● https://martinfowler.com/articles/cd4ml.html
  • 33. Copyright 1995-2019 Arm Limited (or its affiliates). All rights reserved. ● [Sculley, 2015] Sculley, David, et al. "Hidden technical debt in machine learning systems." Advances in neural information processing systems. 2015. ● [Washizaki 2019] Washizaki, Hironori, et al. "Studying Software Engineering Patterns for Designing Machine Learning Systems." arXiv preprint arXiv:1910.04736 (2019). ● [Batch] The Batch Nov. 6, 2019 https://info.deeplearning.ai/the-batch-deepmind-masters-starcraft-2-ai-attacks-on-amazon-a-career-in-robot-management-banks-embrace-b ots-1 ● [Batch 2] The Batch Nov. 20, 2019 https://info.deeplearning.ai/the-batch-artificial-noses-surveillance-on-wheels-unwelcome-researchers-privacy-problems-beyond- bounding-boxes ● [Kubeflow pipeline] https://www.kubeflow.org/docs/pipelines/overview/pipelines-overview/ ● [TFDV] https://www.tensorflow.org/tfx/data_validation/get_started ● [Zliobaite, 2010] Žliobaitė, Indrė. "Learning under concept drift: an overview." arXiv preprint arXiv:1010.4784 (2010). ● [Marsden, 2019] Marsden, Luke. "The Future of MLOps" https://docs.google.com/presentation/d/17RWqPH8nIpwG-jID_UeZBCaQKoz4LVk1MLULrZdyNCs/edit#slide=id.p ● [Baylor, 2017] D. Baylor et al., “TFX: A tensorflow-based production-scale machine learning platform,” in Proceedings of the 23rd acm sigkdd international conference on knowledge discovery and data mining, 2017, pp. 1387–1395 ● [CD4ML] D. Sato, et al. “Continuous Delivery for Machine Learning” , https://martinfowler.com/articles/cd4ml.html ● [Uber] “Databook: Turning big data into knowledge with metadata at uber.” https://eng.uber.com/databook/ ● [Manifold] “Manifold: A Model-Agnostic Visual Debugging Tool for Machine Learning at Uber” https://eng.uber.com/manifold/ ● [Overton] Ré, Christopher & Niu, Feng & Gudipati, Pallavi & Srisuwananukorn, Charles. (2019). Overton: A Data System for Monitoring and Improving Machine-Learned Products. ● [Bernardi 2019] Bernardi, Lucas, et al. "150 Successful Machine Learning Models: 6 Lessons Learned at Booking. com." Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. ACM, 2019. ● [Metaflow] Julie Pitt, et al. “A Human-Friendly Approach to MLOps”, MLOps NYC19, https://youtu.be/fOSZuONmLbA ● [Ariga] “MLOpsの歩き方”, n月刊ラムダノートvol.1 no.1, 2019
  • 34. Confidential © Arm 2017Confidential © Arm 2017Confidential © Arm 2019 Thank You! Danke! Merci! 谢谢! ありがとう! Gracias! Kiitos!