Main principles of Data Science and Machine Learning
0 likes98 views
The document provides an overview of machine learning, deep learning, and big data, covering essential concepts, principles, and methodologies. It discusses types of machine learning, risks and challenges, and various datasets sources, along with practical materials for further learning. Key speaker Nikolay Karelin shares insights and resources for both academic and practical applications in AI.
Main principles of Data Science and Machine Learning
- 2. Introduction to Deep Learning and Big Data 2
- 3. These slides online Web version: https://www.slideshare.net/NikolayKarelin 3
- 4. Contents • Introduction • Machine Learning Principle • Deep Learning • Big Data • Risks and Challenges • Practical Materials • Community • Academic Resources • Software & Hardware • Books & Courses 4
- 5. Speaker Nikolay Karelin, PhD (optics) Head of AI, SilkData.ai Past experience: • VPIphotonics.com • Institute of Physics, National Academy of Sciences 5
- 7. ML Rule of Thumb If a typical person can do a mental task with less than one second of thought, we can probably automate it using AI either now or in the near future. Andrew Ng, Stanford Uni, • Fraud detection • Recommendations • Customer segmentation • Text analysis • Text summarization Source: What Artificial Intelligence Can and Can’t Do Right Now 7
- 8. Buzzwords Data science Big data Data mining Synthetic intelligence Cognitive computing Affective computing Deep tech Medtech Wikipedia: Outline of artificial intelligence Image source: What’s the Difference Between Artificial Intelligence, Machine Learning, and Deep Learning?, NVIDIA 8
- 9. TL;DR What is / Why use ML? TL;DR: • Software tool for getting answers and decision making • Mainly based on data, not rules/algorithms • Two-step: • Automatic extraction of “knowledge” (patterns) from past data • Prediction on new data Which data? • See below… 9
- 11. Classical Software vs. Machine Learning Idea from François Chollet, Deep Learning with Python, Manning, 2017 Data Software Answers Expert Rules Software New data Model Answers Past data & answers Model Patterns found in past data 11
- 12. Curve Fitting Example f(x) = x3 - 4x2 + 3x – 2 + N(0, 0.25) Problem: find best polynomial approximation from few (noisy) samples Why curve fitting??? Known as ‘regression’ in ML Adapted from: Nikolenko et al., Deep Learning, 2017, in Russian 12
- 13. Dataset, Samples and Features Dataset (samples) Input variable (data features / attributes) Output variable (data labels) ML task: find (approximation) of input → output mapping 13
- 14. Model and (Hyper-)Parameters f(x) = ΣN an xn Model Hyperparameter (model order) Parameters (fitting/optimization) 14
- 15. Under- and Overfitting Different polynomial approximations f(x) = a0 + a1 x f(x) = a0 + a1 x + a2 x2 + a3 x3 f(x) = ΣN an xn, N = 9 15
- 16. Machine Learning Types Supervised • Requires past data (labeled dataset) • Mostly, predictions Unsupervised • Produces output from unlabeled data • Example: clustering Semi-supervised • For expensive data Manifold learning • Example: dimensionality reduction Reinforcement learning • Learning ‘on-the-fly’ • Example: self-driving cars, videogame bots 16
- 17. (Classical) ML Types Source: https://scikit-learn.org/stable/tutorial/machine_learning_map/index.html 17
- 18. Datasets, Features and Labels Images / image sequences • Features: image pixels • Titles • Objects & boundaries Texts • Features: words, symbols, sentences • Keywords (named entities) • Sentiment / rating • Translation • Database request • Similar text Structured data • Example: database report • Customer preferences 18
- 19. Datasets: Where to find them • Web crawling (Copyright?) • APIs: • Wikipedia, Google, Twitter, openstreetmap.org, … • Government (open) data • Wikipedia (dump) • https://www.kaggle.com/ (quality varies) • Google Dataset Search: https://toolbox.google.com/datasetsearch • ‘Awesome lists’: https://github.com/awesomedata/awesome-public- datasets 19
- 20. Data Obfuscation: Example Allstate’s contest on Kaggle: • https://www.kaggle.com/c/allstate-claims-severity/data • “Each row in this dataset represents an insurance claim. You must predict the value for the 'loss' column. Variables prefaced with 'cat' are categorical, while those prefaced with 'cont' are continuous.” Obfuscation • No real names / categorical values • Possible linear mix / rescaling of numerical values • Patterns still remain and can be extracted by a model id cat1 cat2 cat3 cat4 cat5 cat6 cat7 … cat116 1 A B A A A A B … LB cont1 cont2 cont3 cont4 cont5 cont6 cont7 cont8 … loss 0.7263 0.2459 0.1875 0.7896 0.31006 0.71836 0.3350 0.3026 … 2213.18 20
- 21. Multi-iterative process More details of ML pipeline Data collection Exploration Visualization Data cleaning Dataset splitting Feature selection Model selection Model training Testing & Evaluation 21
- 22. Feature Selection Only subset of data attributes • Improves model stability Criteria • Maximal correlation between feature and output • Adding of feature improves model (frequently, model training starts from feature subset) Example: Driving history • Important for insurance • Less important for favorite routes 22
- 23. Feature Engineering Convert text etc. to numbers Example: categorical feature (aka ‘one-hot encoding’) Male Female Male Male Female Female Male Male Female Female Gender 1 0 1 1 0 0 1 1 0 0 0 1 0 0 1 1 0 0 1 1 Male Female 23
- 24. Data Cleaning Missing data Incorrect values Non-relevant data combination Dataset imbalance • Example: fraud • Normally a small fraction of transactions • Remove most of ‘other’ data or special ML techniques 24
- 25. Data Cleaning: Example NYC Taxi Dataset Tipping analysis Tips payed in cash excluded Adapted from Real-World Machine Learning, by H. Brink, J. W. Richards, M. Fetherolf (2016), https://github.com/brinkar/real-world-machine-learning Tip amount (USD) Count Count 25
- 26. Domain knowledge It’s important to understand what’s possible / probable Ways to organize work: • Cooperation with domain experts • Mutual education • Data as way to insight "The purpose of our research is insight, not numbers" Numerical Methods for Scientists and Engineers Richard Hamming, reprint of the McGraw-Hill, 1973 26
- 27. Model Training Principle Parameters (weights) Prediction Input data Model True labels (ground truth) Loss function Optimizer Loss score (error) Data Algorithm Train data Validation data Test data ~70-80% < 10-15% < 10-15% 27
- 28. Model evaluation • Metric • Baseline model 28 Source: https://www.dataoptimal.com/churn-prediction-with-r/
- 29. Deep Learning 29
- 30. Deep Learning Principle: CNN Image features calculation Decision 30 Source: Jason's Machine Learning 101 presentation, Dec 2017
- 31. Nearly human performance in • Image recognition • Object detection: • Images • Video • Translation Deep Learning: revolution Image source: Xavier Giro-o-Nieto 2012 contest 31
- 32. Deep learning: Style transfer https://deepdreamgenerator.com/ddream/mpjrb6uisml 32
- 33. Deep Learning: Neuron Source: http://cs231n.github.io/neural-networks-1/ 33
- 34. Deep Learning: Activation Source: http://cs231n.github.io/neural-networks-1/ tanh() Sigmoid ReLU ReLU tanh() 34
- 35. Deep Learning: Key points • Error back-propagation • Big data • GPUs / Clusters • Exploding / Vanishing gradients • CNN for image recognition 35
- 36. Deep Learning: Building Blocks Source: http://cs231n.github.io/convolutional-networks/ Convolution Pooling 36
- 37. Deep Learning: Building Blocks Source: http://cs231n.github.io/convolutional-networks/ 37
- 38. Deep Learning: Network Examples AlexNet Source: https://www.analyticsvidhya.com/blog/2017/08/10-advanced-deep-learning-architectures-data-scientists/ VGGNet 38
- 39. Deep Learning: Network Examples Source: https://www.analyticsvidhya.com/blog/2017/08/10-advanced-deep-learning-architectures-data-scientists/ GoogleNet (or Inception Network) 39
- 40. Deep Learning: Network Examples Source: https://www.analyticsvidhya.com/blog/2017/08/10-advanced-deep-learning-architectures-data-scientists/ 40
- 41. Deep Learning: Network Examples Source: https://www.analyticsvidhya.com/blog/2017/08/10-advanced-deep-learning-architectures-data-scientists/ YOLO (You Only Look Once) 41
- 42. Deep Learning: Network Examples Source: https://www.analyticsvidhya.com/blog/2017/08/10-advanced-deep-learning-architectures-data-scientists/ SegNet 42
- 43. Deep Learning: Network Examples Source: http://www.cs.toronto.edu/~urtasun/courses/CSC2541_Winter17/Deep_generative_models.pdf Generative Adversarial Network (GAN) 43
- 44. Deep Learning: GAN Example Source: https://medium.com/@kcimc/how-to-recognize-fake-ai-generated-images-4d1f6f9a2842 Rus: https://habr.com/post/432580/ Generative Adversarial Network (GAN) 2014 2017 44
- 45. Google Translate (Model) arXiv:1609.08144 Blog: https://ai.googleblog.com/2016/11/zero-shot-translation-with-googles.html 45
- 46. Deep Learning: Size Source: https://medium.com/@culurciello/analysis-of-deep-neural-networks-dcf398e71aae 46
- 47. Big Data 47
- 48. What is BIG? Areas • Health • IoT • Media • Science Image source: https://en.wikipedia.org/wiki/Big_data 48
- 49. Big Data: Perspectives • Business • Social • Technical • Clusters • Cloud 49
- 50. Apache Hadoop Software framework for distributed storage and processing of big data using the MapReduce programming model. https://en.wikipedia.org/wiki/Apache_Hadoop https://www.tutorialspoint.com/hadoop/hadoop_mapreduce.htm 50
- 51. Apache Spark and more Open-source distributed general-purpose cluster- computing framework. Source: https://databricks.com/spark/about 51
- 52. ML pipelines Medium: https://medium.com/@kt.era.ee/the-data-science-workflow-43859db0415 Comparison table: https://docs.google.com/spreadsheets/d/1ZHLzlMdBEHDorCUEeeiSx3WAgfLGw34pC08C9T0IMkE/edit#gid=0 Ideas from Google: https://developers.google.com/machine-learning/guides/rules-of-ml/ 52
- 54. Types of Risks Errors in ML Understanding ML Data risks • Quantity • Quality • Cost (esp. labels) • Security (incl. international issues) Often, feasibility study is required Data ML approach Business task 54
- 55. DL example: Google Translate … ыво алолд фыалллл ооооооооооо оооооооооооо фЛОЛД игишщгпр оооооооооооо шгигигшщр оооооооооо “Mongolian” ? German ооооо злолтщшолд оооооооооооооооооооооооооооооооо ваоифтваомщофит оооооооооооооооооооощватмшщтощт щт “Mongolian” ? English More examples: https://twitter.com/enneuralmachine 55
- 56. Google Translate fault (or joke…) ыво алолд фыалллл ооооооооооо оооооооооооо фЛОЛД игишщгпр оооооооооооо шгигигшщр оооооооооо “Mongolian” Du musst dich selbst töten warf seinen Kopf auf seine Knie German ооооо злолтщшолд оооооооооооооооооооооооооооооооо ваоифтваомщофит оооооооооооооооооооощватмшщтощт щт “Mongolian” Have you forgotten that you have not forgiven yourself yet? English 56
- 58. Example: Change of Lighting Conditions Why a self-driving car chose wrong path? (simulation) DeepXplore: Automated Whitebox Testing of Deep Learning Systems, K. Pei et al., 2017, arXiv:1705.06640 58
- 59. Example: Object Mis-classification Lack of context Accuracy vs. performance trade-off https://habrahabr.ru/post/346140/ https://youtu.be/VOC3huqHrss 59
- 60. Deep Dream Source: https://pikabu.ru/story/deep_dream_generator_i_legkaya_narkomaniya_4057583 Look for animals 60
- 61. Image Confidence Normal 0.88 Face Detection https://hackernoon.com/capsule-networks-are-shaking-up-ai-heres-how-to-use-them-c233a0971952 61
- 62. Image Confidence Normal 0.88 Rotated 0.79 Face Detection https://hackernoon.com/capsule-networks-are-shaking-up-ai-heres-how-to-use-them-c233a0971952 62
- 63. Image Confidence Normal 0.88 Rotated 0.79 “Distorted” 0.90 Face Detection https://hackernoon.com/capsule-networks-are-shaking-up-ai-heres-how-to-use-them-c233a0971952 63
- 67. Materials • Community • Academic resources • Software & Hardware • Books & Courses 67
- 68. Community 68
- 69. ods.ai Some stats: 20 407 participants, 3.06М posts 69
- 70. ods.ai • Заполните форму (подтверждение может занять несколько дней) • На кого ссылаться: nkarelin • После подтверждения: рекомендуется написать о себе в канале #welcome • Обсуждение встреч в Минске: #_meetings_minsk • Канал для БААИ: #baai 70
- 72. Belarus Big Data User Group Facebook: https://www.facebook.com/groups/big.data.nerds.minsk/ 72
- 75. Towards Data Science Source: https://towardsdatascience.com/ 75
- 77. Academic resources Google Scholar arXiv.org arxiv-sanity.org DataFest: https://datafest.by/ Presentation: https://github.com/karelin/sci_papers_talk Video (Ru): https://youtu.be/HuJyLIis6ts 77
- 78. Google scholar Main page: https://scholar.google.com/ 78
- 83. Github Code • https://github.com/tensorflow/tensorflow Data • Example: https://github.com/Ricardo-Liu/Dateset ‘Awesome’ / SOTA lists / web sites • https://github.com/sebastianruder/NLP-progress • https://github.com/topics/awesome-list • https://github.com/ChristosChristofidis/awesome-deep-learning • https://project-awesome.org/ 83
- 84. Programming languages Image source: Wikipedia 84
- 85. Python numerical stack https://www.scipy.org/ 85
- 87. Anaconda https://www.anaconda.com/ 1,400+ data science packages for Python/R and manage your packages, dependencies, and environments 87
- 88. PyCharm Community vs. Pro vs. Education https://www.jetbrains.com/pycharm/ 88
- 89. GPU Image source: https://www.techpowerup.com/gpu-specs/titan-v-ceo-edition.c3277 https://www.nvidia.com/en-us/data-center/tesla-v100/ CUDA = Compute Unified Device Architecture 89
- 90. • Runtime hierarchy • Thread • Warp • Block • Grid • Memory hierarchy • Registers • Local • Shared • … • Global • Host (CPU) GPU: Architecture (Device) Grid Constant Memory Texture Memory Global Memory Block (0,0) Shared Memory Local Memory Thread (0,0) Registers Local Memory Thread (1,0) Registers Block (1,0) Shared Memory Local Memory Thread (0,0) Registers Local Memory Thread (1,0) Registers Host Source: David Kirk/NVIDIA and Wen-mei W. Hwu, 2007-2010, ECE 498AL, University of Illinois, Urbana-Champaign 90
- 91. • https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units • Most comprehensive, though unofficial GPU: details 91
- 92. • https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units • Most comprehensive, though unofficial GPU: details Tesla K40 / K80 GeForce Titan Chip / Codename GK110B GK110-400-A1 CUDA Cores 2880 2688 Clock rate 745 / 875 MHz 837 / 876 MHz Memory size 12 Gb 6144 Mb DRAM Type GDDR5 GDDR5 Double precision 1430–1680 GFLOPS 1300-1499.9* GFLOPS 92
- 94. DL frameworks Main source: https://developer.nvidia.com/deep-learning-frameworks 94
- 95. Why parallel? Source: https://www.karlrupp.net/2018/02/42-years-of-microprocessor-trend-data/ Data: https://github.com/karlrupp/microprocessor-trend-data 95
- 96. Google Colab https://colab.research.google.com/ 96
- 97. Google Colab • GPU: NVIDIA Tesla K80 / TPU • 12-hours limit • Example: • https://github.com/glebsteron/colab-turi-style-transfer-coreml https://colab.research.google.com/ 97
- 98. Cloud solutions • Google cloud ($300 at start) • AWS (Amazon, 1-year free) • Microsoft Azure • Paperspace • IBM Cloud • FloydHub https://medium.com/@rupak.thakur/aws-vs-paperspace-vs-floydhub-choosing-your-cloud-gpu-partner-350150606b39 https://www.nvidia.com/en-us/data-center/gpu-cloud-computing/ 98
- 100. Deep Learning book Ian Goodfellow, Yoshua Bengio, and Aaron Courville, Deep Learning, MIT Press, 2016 Web site: http://www.deeplearningbook.org/ 100
- 101. Artificial Intelligence: A Modern Approach Artificial Intelligence: A Modern Approach, Third edition, 2016, by Stuart Russell and Peter Norvig Web page: http://aima.cs.berkeley.edu/ 101
- 102. Глубокое обучение Николенко С. И., Кадурин А. А., Архангельская Е. О., Глубокое обучение, Питер-пресс, сер. Библиотека программиста, 2019, 480 с. https://www.piter.com/product/glubokoe-obuchenie 102
- 103. Deep Learning with Python François Chollet, Deep Learning with Python, November 2017, ISBN 9781617294433, 384 p. https://www.manning.com/books/deep-learning-with- python Code: https://github.com/fchollet/deep-learning-with- python-notebooks 103
- 104. Deep Learning with Keras Antonio Gulli, Sujit Pal, Deep Learning with Keras, April 2017 https://www.packtpub.com/big-data-and-business- intelligence/deep-learning-keras Code: https://github.com/agulli/kerasBook 104
- 105. Large Scale Machine Learning with Python Bastiaan Sjardin, Luca Massaron, Alberto Boschetti, Large Scale Machine Learning with Python, August 2016 https://www.packtpub.com/big-data-and-business- intelligence/large-scale-machine-learning-python Code: https://github.com/PacktPublishing/Large-Scale- Machine-Learning-With-Python 105
- 106. Real-World Machine Learning Henrik Brink, Joseph W. Richards, and Mark Fetherolf, Real-World Machine Learning, September 2016 https://www.manning.com/books/real-world-machine- learning Code: https://github.com/brinkar/real-world-machine- learning 106
- 107. Python Machine Learning Sebastian Raschka, Vahid Mirjalili, Python Machine Learning - Second Edition, September 2017 https://www.packtpub.com/big-data-and-business- intelligence/python-machine-learning-second-edition Code: https://github.com/rasbt/python-machine-learning- book-2nd-edition 107
- 108. Machine Learning Yearning Andrew Y. Ng, "Machine Learning Yearning" Book web site: http://www.mlyearning.org/ 108
- 109. Courses • http://course.fast.ai/ • https://mlcourse.ai/ • Stanford • http://cs231n.stanford.edu/ / http://cs231n.github.io/ (Convolutional Neural Networks for Visual Recognition), http://svl.stanford.edu/teaching/ • http://web.stanford.edu/class/cs224n/ (Natural Language Processing with Deep Learning) • https://cs.stanford.edu/academics/courses • Coursera • https://www.coursera.org/learn/machine-learning • https://www.coursera.org/specializations/machine-learning-data-analysis (МФТИ) 109
- 110. mlcourse.ai (En + Ru) Resources: https://mlcourse.ai/resources (Eng + Ru) 110
- 111. Stepik course (Ru) https://stepik.org/course/150 111
- 112. Questions? 112
- 113. These slides online Web version: https://www.slideshare.net/NikolayKarelin 113
- 114. Headquarters: Silk Data sp. z o.o. Domaniewska 17/19-133 02-672 Warsaw Poland Partner in Germany: SilkCode GmbH Luisenstraße 62 D-47799 Krefeld Germany www.silkdata.tech [email protected] +48 452 380 167