Skip to content

Latest commit

 

History

History
165 lines (107 loc) · 13.7 KB

data-science.md

File metadata and controls

165 lines (107 loc) · 13.7 KB

Data Science

LIFE CYCLE

Google’s famous MLops

ML systems is more than ML code.

Fast ai project checklist

When I used to do consulting, I’d always seek to understand an organization’s context for developing data projects, based on these considerations:

  • Strategy: What is the organization trying to do (objective) and what can it change to do it better (levers)?
  • Data: Is the organization capturing necessary data and making it available?
  • Analytics: What kinds of insights would be useful to the organization?
  • Implementation: What organizational capabilities does it have?
  • Maintenance: What systems are in place to track changes in the operational environment?
  • Constraints: What constraints need to be considered in each of the above areas?

WORKFLOWS

  1. kaggle

PLATFORMS

  1. Uber, google, netflix, airbnb, etc

STACK

  1. Medium on canonical stack

Data Science OKR KPI

  1. OKR vs KPI, strategic vs tactical
  2. Difference between KPI targets and goals
  3. Comet ml on medium

  1. For the Data Driven manager (not ds)
  2. Measuring DS business value
  3. Best KPIS for DS - the best is what not to do

Being a DS / Researcher

  1. A day in a life
  2. Advice for a ds, business kpi are not research kpi, etc
  3. Review of deep learning papers and co authorship
  4. Full stack DS Uri Weiss

Team Building / Group Cohesion

  1. DS vs DA vs MLE - the most intensive diagram post ever. This is the mother load of figure references.

References:

1, 2, 3, 4, 5, 6, 7, 8, 9, 10

Why data science needs generalists not specialists

  1. (good advice) Building a DS function (team)

Agile for data-science-research

  1. How to manage a data science research team using agile methodology, not scrum and not kanban
  2. Workflow for data science research projects
  3. Tips for data science research management
  4. IMO a really bad implementation of agile for data-science-projects

SOTA AND CURRENT TRENDS SUMMARIES

  1. ICLR 2019
  2. Medium
  3. State of ai, a yearly report

Building Data/DS teams

  1. (great) the data team a short story by erik bern

YOUTUBE COURSES

Deep learning Course

Kadenze - deep learning tensor flow - Histograms for (Image distribution - mean distribution) / std dev, are looking quite good.

Machine Learning Course

  1. Recommended: Udacity includes ML and DL
  2. Week1: Introduction Lesson 4: Supervised, unsupervised.
  3. Lesson 6: model regression, cost function
  4. Lesson 71: optimization objective, large margin classification
  5. PCA at coursera #1
  6. PCA at coursera #2
  7. PCA #3
  8. SVM at coursera #1 - simplified

Predictive Analytics Course

Syllabus

Week 2: Lesson 29: supervised learning

Lesson 36: From rules to trees

Lesson 43: overfitting, then validation, then accuracy

Lesson 46: bootstrap, bagging, boosting, random forests.

Lesson 52: NN

Lesson 55: Gradient Descent

Lesson 59: Logistic regression, SVM, Regularization, Lasso, Ridge regression

Lesson 64: gradient descent, stochastic, parallel, batch.


Unsupervised: Lesson X K-means, DBscan

BOOKS & NOTEBOOKS

  1. Machine learning design patterns, git notebooks!, medium
    1. DP1 - transform Moving an ML model to production is much easier if you keep inputs, features, and transforms separate
    2. DP2 - checkpoints Saving the intermediate weights of your model during training provides resilience, generalization, and tunability
    3. DP3 - virtual epochs Base machine learning model training and evaluation on total number of examples, not on epochs or steps
    4. DP4 - keyed predictions Export your model so that it passes through client keys
    5. DP5 - repeatable sampling use the hash of a well distributed column to split your data into training, validation, and testing
  2. Gensim notebooks - from w2v, doc2vec to nmf, lda, pca, sklearn api, cosine, topic modeling, tsne, etc.
  3. Deep learning with python - francois chollet, deep learning & vision git notebooks!, official notebooks.
  4. Yandex school, nlp notebooks
  5. Machine learning engineering book (i.e., data science)
  6. Interpretable Machine Learning book

COST

  1. GPT2/3

Patents

  1. Method Patent Exceptionalism

General Advice

(really good) Practical advice for analysis of large, complex data sets - distributions, outliers, examples, slices, metric significance, consistency over time, validation, description, evaluation, robustness in measurement, reproducibility, etc.