openskills.info
← Data Scientist career guide

Data and AI

Data Scientist timeline

How the field of a Data Scientist got here: the sourced milestones from every published course of this career, on one axis.

1713 to 2025

287 milestones across 30 courses

  • Core
  • Supporting
  • Specialization

The axis stretches toward recent years, where most milestones fall. Bars count milestones per year; dots are single milestones.

Before 1990

71 milestones

  1. 1713

    Bernoulli publishes the law of large numbers

    Ars Conjectandi introduced a result connecting repeated observations with long-run frequency. It established a mathematical bridge between chance models and empirical regularity.

  2. 1763

    Bayes's essay introduces inverse probability

    The posthumous essay gave a rule for updating a probability after observing evidence. It became the basis for the prior, likelihood, and posterior structure used in Bayesian inference.

  3. 1786

    Playfair publishes The Commercial and Political Atlas

    William Playfair used statistical graphics to present commercial and political data. The work established charts as a form for comparing quantitative change and magnitude.

  4. 1805

    Legendre publishes the method of least squares

    Legendre's appendix on determining comet orbits introduced the least squares criterion for combining observations, the estimation rule every regression fit still uses.

  5. 1809

    Gauss connects least squares to probability theory

    Gauss's Theoria motus developed the method independently and grounded it in the theory of errors, giving least squares the probabilistic foundation that underpins regression inference.

  6. 1812

    Laplace publishes Théorie analytique des probabilités

    Laplace consolidated probability methods and their applications to observational error and inference. The work helped establish probability as a general analytical tool rather than a gambling topic.

  7. 1855

    Snow's cholera map appears in the St James report

    The report included a map of cholera cases around the Broad Street pump. It made spatial distribution part of an argument about a public-health question.

  8. 1869

    Minard publishes the map of Napoleon's Russian campaign

    Charles Joseph Minard combined geographic position, army size, direction, and temperature in one statistical graphic. It remains a reference for combining several variables without treating a chart as decoration.

  9. 1886

    Galton names "regression" from heredity studies

    Galton's analysis of hereditary stature observed offspring heights regressing toward the population mean, and the term became the name of the modeling method that grew out of that work.

  11. 1900

    Pearson introduces the chi-squared goodness-of-fit test

    Pearson's criterion compared observed and expected frequencies. It supplied a reusable test statistic for categorical data and model fit.

  12. 1900

    Pearson publishes the chi-squared goodness-of-fit test

    Pearson's paper supplied a formal way to compare observed and expected frequencies. It established a foundational pattern for judging a model or hypothesis against held-out observations.

  13. 1901

    Pearson publishes principal component analysis

    Pearson describes fitting lines and planes of closest fit to points. The work established the geometric basis for reducing correlated measurements to principal components.

  14. 1908

    Gosset publishes Student's t distribution

    Working as "Student" at Guinness, Gosset derived the exact small-sample behavior of estimates, giving regression coefficients their correct standard errors and intervals at realistic sample sizes.

  15. 1922

    Fisher formalizes calculation of P from contingency tables

    Fisher's paper treated the interpretation of chi-squared statistics and calculation of P. It shaped the significance-testing vocabulary later used in applied hypothesis testing.

  16. 1922

    Fisher formalizes the theory of statistical estimation

    Fisher's paper on the mathematical foundations of theoretical statistics established the estimation framework, including consistency and efficiency, that regression's optimality claims rest on.

  19. 1933

    Neyman and Pearson publish the testing framework

    Their work defined a decision framework for testing statistical hypotheses. It made error rates and power central design quantities.

  20. 1935

    Fisher publishes The Design of Experiments

    Fisher's book presented experimental procedure and significance testing together. It became a reference point for randomization, replication, and planned inference.

  21. 1936

    Fisher publishes linear discriminant analysis

    Fisher described a rule for separating groups from measured features. Classification evaluation later inherited the need to count correct and incorrect decisions from such rules.

  23. 1943

    McCulloch and Pitts publish a logical neuron model

    The paper represented neural activity with threshold units and logical operations. It supplied a mathematical starting point for networks of simplified neurons.

  24. 1946 Jun

    Plackett and Burman publish multifactorial designs

    The paper described optimum multifactorial designs for studying many component changes against a measured characteristic. It provided a compact screening approach when a full factorial is too large.

  25. 1948 Oct 30

    The MRC streptomycin trial is published

    The British Medical Journal published a controlled tuberculosis trial with concealed random allocation. The design demonstrated randomized comparison in a high-stakes medical setting.

  26. 1950

    Turing publishes "Computing Machinery and Intelligence"

    Alan Turing proposed the imitation game — now called the Turing test — as a criterion for machine intelligence, grounding the field's central question in the ability to process and generate language.

  27. 1950 Oct

    Turing publishes Computing Machinery and Intelligence

    Turing proposed the imitation game as a way to discuss machine intelligence. The paper framed a long-running question about machine behavior rather than a specific implementation method.

  29. 1951 Jan

    Box and Wilson publish response-surface methodology

    Box and Wilson described sequential experiments for finding settings that optimize a quantitative response. The method made experimentation an iterative route toward operating conditions rather than a single comparison.

  30. 1954

    Georgetown-IBM machine translation demonstration

    Researchers at Georgetown University and IBM automatically translated 60 Russian sentences into English using six grammar rules, sparking decades of optimism — and eventual disillusionment — about fully automatic translation.

  32. 1954

    Signal detectability formalizes threshold decisions

    Peterson, Birdsall, and Fox analyzed detection as a decision under signal and noise. This provided the threshold trade-off framing that later supports ROC analysis.

  34. 1957

    Bellman publishes Dynamic Programming

    Bellman’s book established recursive value-based methods for multistage decision problems. Its optimality principle became the mathematical basis for Bellman equations in reinforcement learning.

  35. 1957

    Sokal and Michener formalize numerical taxonomy

    Sokal and Michener publish a quantitative approach to classifying organisms from many characters. It helped establish systematic similarity-based grouping as a practical task.

  37. 1958

    Rosenblatt introduces the perceptron

    The perceptron established a trainable linear classifier with adjustable weights. It made learning from examples a concrete procedure for a neural model.

  38. 1958

    Rosenblatt publishes the Perceptron

    Frank Rosenblatt described the perceptron, a trainable linear classifier that adjusts its weights from labeled examples. It established the template this course's "core idea" section still follows: choose a model, train it on data, evaluate it on new inputs.

  39. 1966

    ELIZA chatbot demonstrated at MIT

    Joseph Weizenbaum's ELIZA simulated a psychotherapist using pattern matching and substitution, demonstrating that simple rules could produce convincing-seeming conversation — and that users would attribute understanding where none existed.

  40. 1967

    Bertin publishes Semiology of Graphics

    Jacques Bertin described visual variables and their use in graphic representation. The work gave later visualization practice a vocabulary for marks and visual encodings.

  41. 1967

    Cover and Hart analyze nearest-neighbor classification

    The paper grounded prediction by nearby labeled examples as a supervised method.

  42. 1967

    Cover and Hart formalize nearest-neighbor classification

    Cover and Hart proved that the nearest-neighbor decision rule's error rate is bounded relative to the best possible classifier, giving the simple "classify by closest examples" idea a theoretical foundation. The rule is this course's kNN algorithm.

  43. 1967

    MacQueen presents K-means

    MacQueen describes methods for classifying multivariate observations, including the iterative approach now known as K-means. It made center-based partitioning a standard baseline.

  44. 1970

    Box and Jenkins publish time-series analysis methods

    Established AR, MA, and ARIMA modeling as the dominant paradigm for time-series forecasting, grounding the lag and window concepts that later feature engineering builds on.

  45. 1970

    Codd publishes the relational model

    E. F. Codd published A Relational Model of Data for Large Shared Data Banks. It established the table-and-relation model that SQL later made usable as a query language.

  46. 1970

    SHRDLU understands natural language commands

    Terry Winograd's SHRDLU system understood and executed English commands within a simulated blocks world, showing that structured natural language input was tractable in sufficiently constrained domains.

  47. 1970

    Tukey circulates a preliminary Exploratory Data Analysis edition

    Tukey's preliminary edition established exploratory analysis as a distinct approach before its later book publication.

  48. 1971

    SMART retrieval system

    SMART documented weighted term representations for automatic document retrieval. That established an early practical pattern for converting text into numeric inputs.

  49. 1973

    Anscombe shows identical fits, different data

    Anscombe's constructed quartets produced the same fitted lines, R-squared, and intervals from visibly different relationships, making residual plotting a mandatory step rather than an optional extra.

  50. 1974

    Stone formalizes cross-validatory choice and assessment

    Stone applied cross-validation to both selecting and assessing statistical predictions. The distinction underlies the course's separation of model tuning from final performance estimation.

  51. 1976

    Rubin formalizes inference with missing data

    Rubin distinguished mechanisms that generate missing values and their consequences for inference. Missingness became a modeling and study- design issue rather than a data-cleaning afterthought.

  52. 1977

    Exploratory Data Analysis is published

    Tukey's book collected graphical and resistant techniques under the EDA name, giving practitioners a shared vocabulary for probing data before confirmation.

  53. 1977

    Tukey publishes Exploratory Data Analysis

    John Tukey presented exploratory data analysis as a way to use displays to investigate data before formal conclusions. This reinforced the distinction between exploratory and communicative views.

  54. 1979

    Efron introduces bootstrap methods

    The bootstrap used resampling to estimate uncertainty when a closed- form sampling distribution is inconvenient. It made computational inference practical across a wider range of estimators.

  55. 1979

    Hartigan and Wong publish an efficient K-means algorithm

    Hartigan and Wong publish an algorithm for K-means clustering. It became a widely used practical implementation path for iterative partitioning.

  56. 1979

    Oracle V2 commercializes SQL

    Relational Software, Inc. released Oracle V2 as a commercial SQL-based relational database. SQL moved from research systems into a product category that competing database vendors adopted.

  57. 1980 Apr

    Taguchi publishes the loss function for quality variation

    Taguchi related deviation from a target to loss rather than treating every in-specification result as equally good. This sharpened the practical focus on reducing variation in designed processes.

  58. 1982

    ROC area receives a practical interpretation

    Hanley and McNeil explained the meaning and use of the area under an ROC curve. AUC became a compact measure of ranking performance across thresholds.

  59. 1983

    Tufte publishes The Visual Display of Quantitative Information

    Edward Tufte collected principles and examples for displaying quantitative information. The book made graphical integrity and information-dense displays central topics in visualization design.

  60. 1984

    Breiman, Friedman, Olshen, and Stone publish CART

    Classification and Regression Trees formalized recursive if-then splitting on feature values as a single method for both classification and regression, the decision-tree algorithm this course covers as an interpretable but overfitting-prone baseline.

  61. 1984 Dec

    PC-MATLAB debuts publicly

    PC-MATLAB appeared at the IEEE Conference on Decision and Control, moving the matrix-calculator idea into a commercial product for personal computers.

  62. 1985

    Pro-MATLAB reaches Unix workstations

    The Unix workstation version followed PC-MATLAB and expanded the commercial environment beyond the personal-computer release.

  63. 1985

    Stata 1.0 is released

    Stata shipped its first version in January 1985, joining a statistical-software market that put least squares fits and their diagnostics on professional desktops rather than mainframes.

  64. 1986

    ANSI publishes SQL-86

    ANSI published the first SQL standard. A standard core gave database products a common reference point even as their dialects continued to differ.

  65. 1986

    Back-propagating errors popularizes multilayer training

    The paper demonstrated training internal representations by propagating error through multilayer networks.

  66. 1986

    Backpropagation enables representation learning in multilayer networks

    Rumelhart, Hinton, and Williams described propagating error derivatives through layers to adjust internal weights. The method made gradient-based training practical for multilayer networks.

  67. 1986

    Backpropagation popularizes gradient learning in networks

    Rumelhart, Hinton, and Williams described learning internal representations by propagating error derivatives through layered networks. This made multilayer parameter updates a practical training pattern.

  68. 1986

    Quinlan publishes ID3

    Quinlan's ID3 algorithm built decision trees by choosing, at each split, the feature that most reduced uncertainty about the label. It made tree induction from labeled examples a practical, repeatable procedure rather than a hand-built rule set.

  69. 1986 Oct

    Back-propagation paper describes learning representations

    Rumelhart, Hinton, and Williams described back-propagation for networks of neuron-like units. The method adjusted connection weights to reduce output error.

  70. 1987

    ISO publishes ISO/IEC 9075

    ISO adopted SQL as ISO/IEC 9075. The paired ANSI and ISO work made SQL an international database-language standard.

  71. 1987

    Signal Processing Toolbox and ODE solvers arrive

    MATLAB added a specialist signal-processing toolbox and ordinary differential equation capability, extending the matrix language toward engineering domains.

1990s

33 milestones

  1. 1990

    Elman network introduces neural word embeddings

    Jeffrey Elman's recurrent network encoded each word in a training set as a fixed vector — the first practical word embedding — and the vocabulary as a vector database, enabling sequence predictions beyond the reach of simple perceptrons.

  2. 1990

    R begins at the University of Auckland

    Robert Gentleman and Ross Ihaka begin the informal project that becomes R. The work starts as an effort to build an S-like environment for statistical computing.

  4. 1991

    Python becomes publicly available

    Python was created in the early 1990s as a successor to ABC. Its early releases established the language that later became community-maintained.

  5. 1991 Feb

    Python is posted to Usenet

    Python's first public release established the language that later became the host environment for numerical and tabular libraries.

  6. 1992

    Q-learning receives its convergence result

    Watkins and Dayan showed that Q-learning converges to optimal action values under specified assumptions. The result made an off-policy, model-free control method a central reference point.

  7. 1992

    REINFORCE formalizes policy-gradient learning

    Williams described statistical gradient-following algorithms for reinforcement learning. This provided a direct route to optimizing a parameterized policy rather than first estimating action values.

  8. 1992

    Relief feature weighting

    Relief introduced nearest-instance feature weighting for supervised learning. It made feature relevance a computable property rather than a manual ranking alone.

  9. 1992

    Sparse matrices and Simulink appear

    Sparse-matrix support made storage and computation depend on nonzero entries, while Simulink added a companion simulation environment.

  10. 1992

    SQL-92 expands the standard

    ANSI and ISO published the revision commonly called SQL-92. It became a long-lived baseline for the portable syntax developers still recognize.

  12. 1993

    IBM alignment models for statistical machine translation

    IBM's fertility-based alignment models replaced hand-coded translation rules with statistics learned from parallel corpora, inaugurating the statistical machine translation era and proving that data-driven methods could outperform linguistic expertise at scale.

  13. 1994

    GroupLens describes collaborative filtering for Netnews

    GroupLens used ratings to predict which Netnews articles a reader might like. It documented an early open architecture for gathering and distributing collaborative-filtering signals.

  14. 1994

    Hyndman introduces the forecast package for R

    Made exponential smoothing and ARIMA modeling accessible through R, popularizing the idea that statistical time-series tools could be packaged for broad practitioner use.

  15. 1995

    Benjamini and Hochberg formalize false-discovery-rate control

    The procedure supplied a less conservative error policy for decisions spanning many hypotheses, such as feature-level drift tests.

  16. 1995

    Cortes and Vapnik publish Support-Vector Networks

    Cortes and Vapnik described the support-vector machine, which finds the hyperplane that maximizes the margin between classes and extends to non-separable data. It is the margin-based algorithm this course lists as effective in high-dimensional spaces.

  18. 1995

    Support-vector networks are published

    Support-vector networks formalized maximum-margin classification with kernels.

  19. 1995 May

    Ringo evaluates social information filtering

    The Ringo work applied social information filtering to music recommendations and compared recommendation algorithms. It made personalized recommendations a concrete interactive-system problem.

  20. 1996

    DBSCAN introduces density-based clustering with noise

    Ester and colleagues publish DBSCAN for finding dense clusters in spatial databases while identifying noise. It replaced a required cluster count with density-connected regions.

  21. 1996

    MATLAB 5 adds cell arrays and structures

    Cell arrays and dot-notation structures broadened the language beyond homogeneous numeric arrays while retaining indexed access.

  22. 1996

    R is described as a language for data analysis and graphics

    Ihaka and Gentleman publish their paper on R. The public description establishes the language and graphics focus that still frames R practice.

  23. 1997

    AdaBoost establishes adaptive boosting

    AdaBoost combined weak classifiers by repeatedly emphasizing difficult examples.

  24. 1997

    Freund and Schapire publish AdaBoost

    Freund and Schapire's decision-theoretic generalization of on-line learning introduced AdaBoost, which combines many weak learners into one accurate classifier by reweighting misclassified examples on each round — the boosting idea this course's gradient-boosting entry builds on.

  25. 1997

    Long short-term memory addresses long sequence dependencies

    LSTM introduced a recurrent architecture with controlled information flow. It became a durable reference point for sequence modeling before attention-based architectures.

  26. 1997

    The R Core group is formalized

    The project moves from an informal effort to a Core group with responsibility for the source and supporting infrastructure. This gives R a durable open-source maintenance structure.

  28. 1998

    LeNet demonstrates convolutional document recognition

    LeCun and colleagues applied gradient-based learning to a convolutional network for handwritten character recognition. The work connected learned local features with a useful vision task.

  29. 1998

    LeNet-5 applies gradient-based learning to documents

    LeCun and colleagues described a convolutional network for document recognition. The work linked learned local filters and pooling with a practical image task.

  30. 1998 May

    Microsoft Research compares predictive algorithms

    This report compared correlation, vector-similarity, and Bayesian approaches across datasets and evaluation protocols. It included a metric for the utility of a ranked suggestion list.

  31. 1999

    Non-negative matrix factorization gains a parts-based interpretation

    Lee and Seung show non-negative matrix factorization learning additive parts-based representations. It gave factorization an interpretation for nonnegative images and text counts.

  32. 1999

    SQL:1999 adds object and procedural features

    SQL:1999 added features including recursive queries and standardized procedural extensions. It widened SQL beyond basic table queries while retaining the relational core.

  33. 1999 Sep

    SIFT Feature Detection

    Lowe introduces scale-invariant feature transforms that remain robust across scale, rotation, and illumination change. SIFT became the default local descriptor for image matching and object recognition for over a decade.

2000s

37 milestones

  1. 2000

    R 1.0.0 is released

    R reached version 1.0 in February 2000, consolidating an open statistical environment whose lm fits and formula interface became a default route to regression in research and teaching.

  2. 2000

    The MATLAB desktop becomes the integrated workspace

    The desktop brought the command window, editor, workspace view, and related panels into the working environment learners recognize today.

  3. 2000 Feb 29

    R 1.0.0 is released

    R reaches its first official version. The release marks the transition from the early development period to a stable public platform for contributors and users.

  4. 2000 Jun

    PEP 1 defines the proposal process

    PEP 1 established the Python Enhancement Proposal process. Language and community proposals gained a documented route for discussion and decision.

  5. 2000 Oct

    Python 2.0 is released

    Python 2.0 added list comprehensions and cyclic garbage collection. Both changes became part of the practical language baseline for later Python users.

  6. 2001

    Breiman publishes Random Forests

    Breiman formalized random forests: many decision trees trained on randomly resampled data and feature subsets, with predictions averaged across the ensemble. It is this course's stated technique for reducing overfitting while keeping a tree-based model's flexibility.

  7. 2001

    Gradient boosting is framed as function approximation

    Friedman described fitting additive models by gradient descent in function space.

  8. 2001

    HCTSA extracts thousands of time-series features

    The Highly Comparative Time-Series Analysis framework automated extraction of over seven thousand features, establishing the computational feature-extraction paradigm that tsfresh later scaled.

  10. 2001

    Random Forests formalizes tree averaging

    Breiman described randomized tree predictors and their generalization behavior.

  11. 2001 Jul

    Viola-Jones Face Detection

    A cascade of simple Haar-like features selected by AdaBoost achieves real-time face detection at 15 fps on consumer hardware. This was the first practical object detector widely deployed in cameras, phones, and security systems.

  12. 2002

    SVM recursive feature elimination

    Guyon and colleagues applied recursive feature elimination with support vector machines to gene-expression classification. The method became a representative wrapper approach for selecting variables.

  13. 2003

    Bengio introduces the neural probabilistic language model

    Yoshua Bengio and co-authors showed that learning distributed word representations jointly with a sequence model produced better generalization than n-grams, laying the theoretical groundwork for all subsequent neural language models.

  14. 2003

    Matplotlib enters its early public era

    Matplotlib's original 2003-2008 project identity marks the emergence of a Python plotting layer for array-based analysis.

  15. 2003

    SQL:2003 standardizes window functions

    SQL:2003 added window functions and SQL/XML work. Window functions made ranking and calculations across related result rows part of standard SQL practice.

  16. 2003

    The R Foundation supports the project

    The R Foundation is established to support R, hold its software and documentation copyright, and provide a point of contact for institutions and commercial organizations.

  17. 2003

    Variable and feature selection survey

    Guyon and Elisseeff surveyed feature construction, ranking, multivariate selection, search, and validity assessment. It consolidated the questions that modern selection workflows still address.

  18. 2003 Feb

    Amazon publishes item-to-item collaborative filtering

    Amazon described an item-to-item method whose online computation scaled independently of customer and catalog size. The paper tied recommendation quality to real-time delivery in a large catalog.

  20. 2004

    Function handles and parallel computing expand program structure

    Function handles and parallel-computing support made reusable program units and larger computational workloads more central to MATLAB practice.

  21. 2005

    NumPy is created

    NumPy combined earlier Numeric and numarray work into the array library that provides the vectorized numerical base used by pandas.

  22. 2005

    Wilkinson publishes The Grammar of Graphics

    Leland Wilkinson described graphics as composable elements rather than a fixed menu of chart types. This model informed declarative charting systems that separate data, transformations, marks, scales, and guides.

  23. 2006

    Deep belief nets renew interest in training deep models

    Hinton, Osindero, and Teh presented a fast learning algorithm for deep belief nets. The paper made layer-wise representation learning a major research direction.

  25. 2006 Mar

    Microsoft forms the Experimentation Platform team

    Microsoft formed ExP as an incubation project for controlled experiments across online properties. Experimentation began moving from isolated analyses toward a shared service.

  27. 2007

    Google publishes a controlled-experiment case study for advertisers

    Google's case study used a statistically designed experiment to assess new advertising features. It documented controlled experimentation as a product-development evaluation method at online scale.

  28. 2007

    scikit-learn begins as a Google Summer of Code project

    The project that became scikit-learn started in 2007 and grew into the Python workbench whose linear-model module sits beside regularized variants of least squares.

  30. 2007

    scikit-learn first released

    David Cournapeau's Google Summer of Code project becomes scikit-learn, establishing the fit/predict/score API pattern that gradient boosting libraries would later adopt and that remains the standard for classical ML in Python.

  32. 2008

    t-SNE is introduced for high-dimensional visualization

    van der Maaten and Hinton publish t-SNE for placing high-dimensional observations in two or three dimensions. Its local-neighborhood focus made it influential for exploratory embedding plots.

  33. 2008 Jul 30

    Controlled experiments on the web survey is published online

    The survey organized web experimentation terminology, architecture, statistical analysis, and practical pitfalls into a common reference. Its 2009 journal issue connected industrial practice with experimental design.

  34. 2008 Dec

    Python 3.0 is released

    Python 3.0 made incompatible language and library changes rather than preserving every Python 2 behavior. Porting became an explicit concern for projects spanning the transition.

  36. 2009

    The R Journal documents the expanding ecosystem

    The R Journal publishes an account of the project's organization and trajectory, including the growth of contributed packages. R practice increasingly includes work built and shared through packages.

  37. 2009 Aug

    Matrix factorization becomes a core recommendation method

    The review explained matrix factorization models for recommendations with explicit and implicit feedback, temporal effects, and confidence levels. It connected the Netflix Prize era to latent-factor practice.

2010s

103 milestones

  1. 2010

    Cawley and Talbot analyze selection bias

    Their study showed how using the same data for model selection and assessment can produce optimistic performance estimates. Nested cross-validation became a standard response when tuning is substantial.

  2. 2010

    GPU support enters MATLAB

    GPU capability expanded the execution options for computations that could benefit from accelerated array processing.

  3. 2010

    scikit-learn makes its first public release

    The first public release made a shared implementation of estimators, splitters, and scorers available to Python practitioners. Evaluation workflows became easier to express as reusable code.

  4. 2010

    scikit-learn's first public release

    The February 2010 release put linear regression, ridge, and lasso behind one fit-and-predict interface, embedding least squares as the baseline estimator of the Python machine-learning stack.

  5. 2010

    statsmodels released after Google Summer of Code 2009

    The scipy.stats models code was corrected, tested, and released as statsmodels, giving Python an inference-focused OLS implementation with the coefficient tables and diagnostics regression teaching relies on.

  6. 2010 Jun

    ImageNet ILSVRC Launch

    The Large Scale Visual Recognition Challenge launches with 1000 object categories and 1.2M training images, creating the benchmark that catalyzed the deep learning revolution. All entries that year used hand-crafted features.

  7. 2010 Jul

    Google publishes overlapping experiment infrastructure

    Google described infrastructure that allowed many concurrent experiments while separating parameter namespaces. The paper treated tooling and education as parts of experimentation at scale.

  8. 2011

    scikit-learn published in JMLR

    The scikit-learn paper described a reusable Python machine learning toolkit. Its preprocessing, extraction, selection, and pipeline tools made feature transformations reproducible in ordinary model workflows.

  9. 2011

    SQL:2011 adds temporal features

    SQL:2011 added temporal database features. The standard gained a common way to express time-aware data and queries.

  10. 2011 Feb

    D3 version 1.0 is released

    D3 made data-driven document manipulation available as a web visualization library. It enabled authors to bind data to web standards while retaining direct control over marks and interaction.

  11. 2011 Feb

    RStudio launches as an open-source IDE for R

    RStudio provides an integrated editor, console, plotting, and project environment. It makes the interactive R workflow more accessible without replacing the underlying language or scripts.

  12. 2011 Feb

    scikit-learn's reference paper is published

    Pedregosa et al. published the scikit-learn paper in the Journal of Machine Learning Research, documenting the open-source Python library that gave the classical algorithms this course teaches — linear models, trees, SVMs, clustering, PCA — one consistent, widely adopted API.

  13. 2012

    AlexNet demonstrates deep convolutional classification at ImageNet scale

    Krizhevsky, Sutskever, and Hinton trained a deep convolutional network for large-scale image classification. The result focused broad attention on GPU-trained deep networks for visual recognition.

  14. 2012

    AlexNet wins ImageNet classification

    AlexNet used a deep convolutional network trained on GPUs to win the ImageNet 2012 challenge. The result made large-scale deep visual models a central practical approach.

  15. 2012

    AlexNet wins the ImageNet classification competition

    A deep convolutional network achieved a top-five error rate of 15.3 percent in ILSVRC-2012. The result demonstrated the value of large labeled datasets, GPUs, and deep neural networks for image classification.

  16. 2012

    MMD receives a comprehensive kernel two-sample treatment

    Gretton and coauthors developed maximum mean discrepancy through reproducing-kernel mean embeddings and analyzed practical test statistics.

  17. 2012 Feb 14

    Julia announced publicly

    The creators publish "Why We Created Julia," stating the goal of a language with the speed of C and the dynamism of Ruby, combining strengths they associated with MATLAB, Lisp, Python, R, and C.

  18. 2012 Jun

    pandas 0.8 improves time-series handling

    pandas 0.8 adopted NumPy datetime64 for time-series representation, improving joins, memory use, and serialization around dated data.

  19. 2012 Nov

    AlexNet

    Krizhevsky, Sutskever, and Hinton train a 60M-parameter CNN on GPUs and slash ImageNet top-5 error from 26.2% to 15.3%, proving deep convolutional networks far surpass hand-crafted features. This result is widely credited with launching the deep learning era in computer vision.

  20. 2012 Nov

    Shiny introduces interactive web applications in R

    Shiny makes it possible to build interactive applications from R code. It broadens the path from an analysis object to an interface that other people can use.

  21. 2013

    HDBSCAN adds hierarchy and stability to density clustering

    Campello, Moulavi, and Sander describe hierarchical density-based clustering and a stability criterion for selecting clusters. It broadened density clustering to varying density levels.

  22. 2013

    Word2Vec makes dense word embeddings practical

    Tomáš Mikolov's Word2Vec architecture trained skip-gram and CBOW models on massive corpora in hours, making semantic vector representations of words widely accessible and demonstrating that arithmetic on vectors captured analogies like "king − man + woman ≈ queen."

  23. 2013 Feb

    CUPED variance reduction is introduced

    Researchers described using pre-experiment data to reduce metric variance in online controlled experiments. The method improved sensitivity without replacing randomized assignment.

  24. 2013 Dec

    DQN learns Atari control from pixels

    The DQN preprint combined a convolutional neural network with a Q-learning variant and learned policies from raw Atari screen input. It connected function approximation to a widely used value-based RL method.

  25. 2014

    Attention mechanism introduced for sequence-to-sequence models

    Bahdanau, Cho, and Bengio showed that letting a decoder attend to all encoder states — rather than compressing a sentence into a single fixed-length vector — dramatically improved machine translation quality, fixing the bottleneck that had capped sequence-to-sequence performance.

  26. 2014

    Project Jupyter begins

    Jupyter grew from IPython to support interactive data science and scientific computing across programming languages.

  27. 2014

    Project Jupyter is born from IPython

    Jupyter broadened the notebook model beyond Python for interactive scientific computing and data exploration.

  28. 2014

    R Markdown formalizes executable reporting

    R Markdown combines prose with R code and generated results in one document. The approach makes the analysis and its presentation travel together.

  29. 2014

    Sequence-to-sequence learning frames neural translation

    An encoder network represented an input sequence and a decoder generated the output sequence. This established an end-to-end neural pattern for variable-length sequence tasks.

  30. 2014 Jun 26

    First JuliaCon held in Chicago

    The Julia community holds its first dedicated conference at the University of Chicago's Gleacher Center, over June 26 and 27.

  31. 2014 Sep

    GoogLeNet and VGGNet

    GoogLeNet introduces the Inception module with parallel convolutions at multiple scales, using 12x fewer parameters than AlexNet. VGGNet shows that stacking simple 3x3 convolutions to 16-19 layers consistently improves accuracy. Both dominate ILSVRC 2014.

  32. 2015

    Deep Feature Synthesis

    Deep Feature Synthesis described automated construction of features from relational data by stacking transformation and aggregation primitives. It put automatic candidate generation alongside manual domain-driven engineering.

  33. 2015

    Hidden Technical Debt names ML system risks

    Sculley and colleagues described data dependencies, feedback loops, and changing external conditions as system-level ML risks. Evaluation expanded beyond a one-time offline score toward ongoing evidence.

  34. 2015

    ResNet introduces residual learning for very deep networks

    Residual connections gave layers a reference path to their inputs and made much deeper visual networks easier to optimize. The architecture became a standard building block for deep vision models.

  35. 2015

    Vega presents a declarative visualization grammar

    Vega defined a JSON-based grammar for describing visual appearance and interactive behavior. It made data, transformations, scales, guides, and marks explicit parts of a specification.

  36. 2015 Feb

    DQN is published in Nature

    The published DQN study evaluated one architecture and algorithm across 49 Atari games. It made end-to-end deep RL a standard benchmark setting.

  37. 2015 Jun

    Faster R-CNN

    Ren, He, Girshick, and Sun introduce Region Proposal Networks that share convolutional features with the detector, replacing slow external proposal methods like Selective Search. Near real-time two-stage detection at 5 fps on GPU becomes the standard detection architecture.

  38. 2015 Sep

    Python 3.5 adds async and await syntax

    Python 3.5 added native coroutines and the async and await syntax. Asynchronous I/O became a first-class language-level pattern.

  39. 2015 Nov

    TensorFlow 1.0 announced

    Google releases TensorFlow as an open-source framework for deep learning, emphasizing production deployment with TF Serving and mobile with TF Lite. Its static computation graph model dominates early deep learning adoption.

  40. 2015 Nov

    TensorFlow is open sourced

    TensorFlow was released under the Apache 2.0 license. The release made Google's machine-learning system and associated tools available outside its internal infrastructure.

  41. 2015 Dec

    ResNet

    He, Zhang, Ren, and Sun introduce skip connections that let gradients flow through networks 152 layers deep — eight times deeper than VGG — winning ILSVRC 2015 with 3.57% top-5 error. The residual learning framework became the backbone for nearly all subsequent vision architectures.

  42. 2016

    ASA issues its statement on statistical significance and p-values

    The American Statistical Association clarified that a p-value measures compatibility of data with a specified model, not the probability that a hypothesis is true or the size or importance of an effect.

  43. 2016

    Live Editor combines narrative, code, and output

    The Live Editor made it possible to keep formatted explanation, equations, MATLAB code, and generated results in one interactive document.

  44. 2016

    SQL:2016 adds JSON support

    SQL:2016 incorporated JSON-related features. Relational databases could standardize some work with semi-structured values without abandoning tables, constraints, and joins.

  46. 2016

    YouTube documents deep candidate generation and ranking

    Google's paper described a deep recommendation system as separate candidate-generation and ranking stages. It made the two-stage architecture a widely cited production reference.

  47. 2016 Jan

    AlphaGo combines deep networks with search

    The AlphaGo paper described policy and value networks trained with supervised and reinforcement learning, combined with search. It showed a system that joined learned models with an explicit decision procedure.

  48. 2016 Feb

    LIME proposes local surrogate explanations

    LIME introduced an approach that learns an interpretable model around an individual prediction. It made model-agnostic local explanation a practical pattern for classifiers.

  49. 2016 Mar

    Keras gains TensorFlow backend

    Keras, originally a high-level API supporting multiple backends, solidifies its integration with TensorFlow as the default backend. This relationship eventually leads to tf.keras becoming TensorFlow's primary high-level API in version 2.0.

  50. 2016 Jun

    OpenAI Gym standardizes experiment interfaces

    The Gym paper introduced a toolkit for developing and comparing reinforcement-learning algorithms. Its environment interface made reset-and-step experiments easier to share and reproduce.

  51. 2016 Jun

    YOLO

    Redmon et al. recast detection as a single regression problem from pixels to bounding boxes, achieving 45 fps with a single forward pass — over 100x faster than two-stage detectors. YOLO made real-time object detection practical on edge devices and spawned a family of successors.

  52. 2016 Aug

    Chen and Guestrin publish the XGBoost paper

    Chen and Guestrin's "XGBoost: A Scalable Tree Boosting System" described an optimized, regularized implementation of gradient boosting that became dominant on tabular data competitions and business problems, the algorithm this course names as the production endpoint for structured-data tasks.

  53. 2016 Sep

    PyTorch released by Facebook AI Research

    PyTorch launches with dynamic computation graphs and a Pythonic API that feels like writing NumPy with gradients. Its define-by-run approach and standard Python debugging tools make it immediately popular with researchers.

  54. 2016 Sep

    R2016b adds implicit expansion for array arithmetic

    Compatible singleton dimensions can expand during supported element-wise operations, changing some former size-mismatch errors into computed arrays.

  55. 2016 Oct

    tsfresh initial release on PyPI

    Brought automated time-series feature extraction to Python with integrated hypothesis-test filtering, making scalable feature engineering practical for production pipelines.

  56. 2016 Nov

    First-year platform expansion

    The project reported distributed-training support and ports to iOS and Raspberry Pi. TensorFlow practice was already extending beyond a single server training environment.

  57. 2017

    Attention Is All You Need introduces the Transformer

    Vaswani et al. at Google removed recurrence entirely, replacing it with self-attention. The resulting architecture processed sequences in parallel, scaled reliably with more data and compute, and became the foundation for every major NLP model that followed.

  58. 2017

    Delta Lake reaches Databricks customers

    Delta Lake began as an ACID table storage layer over cloud object stores, giving workloads a transaction log above Parquet files.

  60. 2017

    Tall and categorical arrays broaden data handling

    Tall arrays and categorical arrays extended MATLAB's array model to larger and labeled data-oriented workflows.

  61. 2017

    Transformer replaces recurrence with attention

    The Transformer used attention mechanisms instead of recurrent layers for sequence transduction. Its parallel training path changed how language architectures were built.

  62. 2017

    Transformer uses attention without recurrence or convolution

    Vaswani and colleagues proposed a sequence-transduction architecture based on attention. The design enabled parallel computation across sequence positions during training.

  63. 2017

    Vega-Lite research paper is published

    Vega-Lite introduced a higher-level grammar for interactive graphics. It showed how encoding rules and composition can express layered and multi-view displays concisely.

  64. 2017 Feb

    TensorFlow 1.0 announced

    TensorFlow 1.0 introduced a stable major release and highlighted XLA as a basis for further performance work. The platform's use included multi-GPU and distributed training.

  65. 2017 May

    SHAP unifies additive feature attribution

    The SHAP paper defined a unified additive feature-attribution framework based on Shapley values. It gave practitioners a common vocabulary for assigning feature contributions to one prediction.

  66. 2017 Jun

    "Attention Is All You Need" published

    The Transformer paper from Google introduces the architecture that underpins BERT, GPT, and virtually all modern large language models. PyTorch becomes the framework of choice for implementing and iterating on Transformer variants, accelerating its research lead.

  67. 2017 Jun

    Integrated Gradients introduces an axiomatic method

    Integrated Gradients proposed a gradient-based attribution method for deep networks. It made a model-specific route available for explaining differentiable models.

  68. 2017 Jun

    Transformer architecture is proposed

    Attention Is All You Need proposed a Transformer architecture based on attention mechanisms rather than recurrence or convolution. The paper reported machine-translation results and a more parallelizable training design.

  69. 2017 Jul

    DARPA launches its Explainable AI program

    DARPA's program organized research around explainable models and human-computer interfaces for explanation. It highlighted the trade-off between learning performance and explanation.

  70. 2017 Aug

    Microsoft documents metric-interpretation pitfalls

    Microsoft's paper collected twelve recurring mistakes in interpreting online controlled-experiment metrics. It shifted attention from running a test to checking whether the chosen metrics support a trustworthy decision.

  71. 2017 Aug

    Neural Collaborative Filtering models user-item interactions

    Neural Collaborative Filtering replaced the usual latent-vector inner product with a learnable neural interaction function. The work focused on implicit-feedback recommendation.

  72. 2017 Aug

    The Dirty Dozen metric pitfalls are published

    Microsoft researchers cataloged recurring interpretation errors observed across thousands of experiments. Metric validity and practical interpretation became explicit parts of trustworthy analysis.

  73. 2017 Sep

    Uber describes Michelangelo

    Uber publicly described Michelangelo's centralized Feature Store for canonical, shared features. The design tied feature engineering to training, deployment, and operational reuse.

  74. 2017 Oct

    AlphaGo Zero demonstrates self-play at scale

    AlphaGo Zero learned Go from self-play without human game data, combining a neural network with search. The result made self-play a prominent training pattern for games with clear rules and win signals.

  75. 2017 Oct

    Mask R-CNN

    He, Gkioxari, Dollar, and Girshick extend Faster R-CNN with a parallel mask prediction branch and RoIAlign for pixel-accurate alignment, solving instance segmentation with a single model. The framework runs at 5 fps and won COCO 2017 on detection, segmentation, and keypoint tasks.

  76. 2017 Nov 28

    Celeste.jl exceeds 1 petaflop on Cori

    A Julia astronomy application (Celeste.jl) sustains 1.54 petaflops of double-precision performance across 1.3 million threads on NERSC's Cori supercomputer, the first Julia application to exceed one petaflop per second.

  77. 2017 Dec

    FRESH algorithm paper published on arXiv

    Described the distributed and parallel feature extraction algorithm behind tsfresh, proving that hundreds of features could be extracted and filtered at industrial scale.

  78. 2018

    BERT and GPT establish the two Transformer paradigms

    Google's BERT (encoder-only, bidirectional, masked language modeling) and OpenAI's GPT (decoder-only, left-to-right, causal language modeling) defined the two dominant approaches: BERT for understanding tasks, GPT for generation. Both achieved state-of-the-art results with task-specific fine-tuning or prompting.

  79. 2018

    CatBoost publishes ordered boosting

    CatBoost presented gradient boosting with methods for categorical features and prediction shift.

  80. 2018

    UMAP is released as a manifold-learning implementation

    McInnes and colleagues publish UMAP for manifold approximation and projection. It became a widely used option for exploratory low-dimensional embeddings.

  81. 2018 Jan

    Soft Actor-Critic introduces maximum-entropy control

    Soft Actor-Critic presented an off-policy actor-critic method that optimizes return and policy entropy. It became a common baseline for continuous-control experiments.

  82. 2018 Mar

    TensorFlow Hub introduced reusable modules

    TensorFlow Hub launched as a library for reusable machine-learning modules. Reusing trained modules became a documented alternative to training every model from scratch.

  83. 2018 Mar

    TensorFlow.js introduced

    TensorFlow.js made it possible to define, train, and run models in JavaScript, including in the browser. It created a distinct deployment path from Python applications.

  84. 2018 Apr

    JAX announced by Google

    Google releases JAX, a library for high-performance numerical computing that combines NumPy's interface with automatic differentiation, JIT compilation via XLA, and composable transforms (grad, jit, vmap, pmap). It targets researchers who need fine control over differentiation and hardware-specific optimization.

  85. 2018 Apr

    Large-scale experimentation platform anatomy is documented

    The architecture separated portal, execution, log processing, and analysis services. This made the end-to-end measurement path a platform design concern rather than only a statistical calculation.

  86. 2018 Jul

    The GDPR becomes applicable

    The General Data Protection Regulation became applicable across the European Union. Its provisions on automated decision-making made transparency and human safeguards an operational concern for affected decisions.

  87. 2018 Aug 8

    Julia 1.0 released

    Julia reaches 1.0 after nearly a decade of work, adding a commitment to language API stability: code written for 1.0 continues to work on later 1.x releases.

  88. 2018 Sep

    Google releases the What-If Tool

    The What-If Tool made visual model inspection and counterfactual-style exploration accessible in TensorBoard. It brought inspection of individual predictions and dataset slices into a developer workflow.

  89. 2018 Sep

    tsfresh paper published in Neurocomputing

    Formalized the FRESH algorithm and its filtering procedure in a peer- reviewed venue, giving the approach academic credibility alongside its industrial adoption.

  90. 2019 Jan

    Feast open-source release

    Google Cloud and GO-JEK announced Feast as an open-source feature store for managing, storing, and discovering features. It made feature-store concepts available as a reusable platform rather than only an internal system.

  91. 2019 Jan

    renv introduces project-local package environments

    renv records a project's package state in a lockfile. This addresses the gap between a script that runs locally and one that can be restored by a collaborator or deployment system.

  92. 2019 Feb

    Delta Lake becomes generally available

    The February platform release recorded Delta Lake general availability, making reliable table behavior a standard platform capability.

  93. 2019 Apr

    BERT4Rec applies bidirectional Transformers to sequences

    BERT4Rec used a Cloze-style task to model item sequences with bidirectional Transformer representations. It targeted dynamic and evolving preferences from historical behavior.

  96. 2019 Jul

    Microsoft publishes a sample-ratio-mismatch taxonomy

    The paper organized causes of observed assignment counts differing from the configured ratio across assignment, execution, log processing, and analysis. It made sample-ratio mismatch a formal diagnostic before product decisions.

  97. 2019 Jul

    SRM diagnosis receives a practitioner taxonomy

    The KDD paper organized sample ratio mismatch causes across assignment, execution, logging, and analysis. SRM became a concrete validity gate with diagnostic rules.

  98. 2019 Jul 23

    Composable multi-threading announced (Julia 1.3)

    Julia 1.3 introduces `Threads.@spawn` and a dynamic task scheduler, letting nested parallel calls share threads without oversubscribing the machine.

  99. 2019 Sep

    AI Explainability 360 documents a toolkit

    IBM researchers documented AI Explainability 360 as an open-source toolkit with multiple explanation algorithms and metrics. The toolkit helped turn a growing research catalog into reusable evaluation components.

  100. 2019 Sep

    sktime unified time-series ML framework launched

    Provided a scikit-learn-compatible interface for forecasting, classification, and feature extraction, reducing the code gap between training and serving for time-series models.

  101. 2019 Sep

    TensorFlow 2.0 released

    TensorFlow 2.0 made eager execution and Keras integration central to the developer experience. SavedModel became the standardized format for several target runtimes.

  102. 2019 Sep

    TensorFlow 2.0 with eager execution as default

    TensorFlow 2.0 makes eager execution the default, integrates Keras as its primary high-level API, and deprecates the session-based workflow. This acknowledges the research community's preference for dynamic graphs that PyTorch demonstrated.

  103. 2019 Oct

    Python 3.8 adds assignment expressions

    Python 3.8 added assignment expressions through the colon-equals operator. Expressions can now bind a name where that form improves a condition or comprehension.

2020s

43 milestones

  1. 2020

    Drift theory separates temporal change from finite-sample evidence

    A formal treatment clarified that drift detection depends on temporal assumptions and that not every distribution change is identifiable from a finite observed stream.

  2. 2020

    GPT-3 demonstrates few-shot in-context learning

    OpenAI's 175-billion-parameter GPT-3 showed that scaling language models produced emergent capabilities: the same model could classify, translate, and answer questions with no fine-tuning, only examples in the prompt. The "foundation model" paradigm was born.

  3. 2020

    pandas 1.0 marks a project milestone

    The 1.0 release documented pandas' maturation into a widely used data-analysis foundation and a stable basis for reproducible EDA work.

  4. 2020

    Retrieval-augmented generation paper published

    Lewis et al. at Facebook AI showed that combining a pretrained sequence-to-sequence model with a neural retrieval component grounded generation in source documents, reducing hallucination and enabling knowledge-intensive tasks without retraining the generator.

  5. 2020

    Vision Transformer applies Transformers to image patches

    Vision Transformer treated image patches as a sequence and applied a pure Transformer to image classification. It extended the attention-based architecture into mainstream visual recognition.

  6. 2020 Jan

    pandas 1.0 establishes its major-version policy

    pandas 1.0 formalized a deprecation and major-release policy, making API evolution more legible for long-lived analysis code.

  7. 2020 Mar

    Catch22 distills HCTSA to twenty-two features

    Reduced over seven thousand HCTSA features to twenty-two with high discriminative power, demonstrating that curated subsets can match full feature sets at a fraction of the computational cost.

  8. 2020 Jul

    TensorFlow 2.3 adds tf.data service capabilities

    TensorFlow 2.3 added mechanisms aimed at input-pipeline bottlenecks and resource utilization. Data input pipelines became a more explicit scaling concern.

  9. 2020 Sep

    TensorFlow Recommenders is introduced

    TensorFlow Recommenders was introduced as an open-source package for candidate models, multi-task objectives, evaluation, and serving. It packaged common modern recommender workflows for TensorFlow users.

  10. 2020 Sep

    XGBoost reaches 1.0

    XGBoost, the gradient boosting library that dominates Kaggle competitions and production tabular ML, reaches its 1.0 release with a stable API. It implements scikit-learn's fit/predict interface while providing GPU training and missing value handling that scikit-learn lacks.

  11. 2020 Oct

    Vision Transformer (ViT)

    Dosovitskiy et al. apply a standard Transformer directly to sequences of image patches with no convolutional inductive bias. When pre-trained at sufficient scale (JFT-300M), ViT matches or exceeds the best CNNs while requiring fewer compute resources to train.

  13. 2021 Jan

    Hugging Face Transformers surpasses 10,000 models

    The Hugging Face model hub crosses 10,000 pretrained models, establishing itself as the de facto registry for NLP, vision, and audio models. Its unified API across PyTorch, TensorFlow, and JAX makes framework selection less of a locked-in decision.

  14. 2021 Mar

    Temporian introduces leakage-prevention by default

    Made future-leakage prevention the default feature computation mode, requiring explicit opt-in for any feature that reads past the cutoff, embodying the principle this course teaches as the availability rule.

  15. 2021 Mar 24

    Julia 1.6 released, and named as the LTS candidate

    Julia 1.6 ships with the compiler and release team flagging it as the likely next Long-Term Support release, a status the project confirmed later that year once the release had been used in the field.

  16. 2021 Apr

    DINO

    Caron et al. show that self-supervised ViT features spontaneously encode semantic segmentation boundaries without any labels, and reach 80.1% top-1 on ImageNet with k-NN classification alone. The method establishes self-supervised vision transformers as powerful feature extractors.

  17. 2021 May

    Unity Catalog is introduced

    Databricks introduced Unity Catalog for fine-grained governance of data and artificial intelligence assets across workspaces.

  18. 2021 Sep

    NIST publishes four principles of explainable AI

    NISTIR 8312 set out Explanation, Meaningful, Explanation Accuracy, and Knowledge Limits. It separated the quality of an explanation from the accuracy of the model decision.

  19. 2021 Oct

    Python 3.10 adds structural pattern matching

    Python 3.10 added match and case statements for structural pattern matching. Code can dispatch on the shape of a value as well as ordinary Boolean conditions.

  20. 2021 Dec

    Databricks SQL reaches general availability

    Databricks SQL became generally available, establishing the SQL warehouse experience as a supported production surface.

  21. 2022

    ChatGPT brings instruction-tuned LLMs to a mass audience

    OpenAI's ChatGPT combined GPT-3.5 with RLHF-based instruction tuning and a conversational interface, reaching 100 million users in two months and demonstrating that aligned, prompted language models could serve as general-purpose assistants.

  22. 2022 May

    TensorFlow 2.9 introduces DTensor

    TensorFlow 2.9 released DTensor for moving between data parallelism and model parallelism. Distribution choices became available through a higher-level API.

  23. 2022 Aug

    Unity Catalog reaches general availability

    Unity Catalog became generally available, making its governed namespace and access model a stable platform foundation.

  24. 2022 Oct

    Python 3.11 improves tracebacks

    Python 3.11 added fine-grained error locations in tracebacks and delivered interpreter speed improvements. Diagnosing an expression failure gained more precise source context.

  25. 2022 Nov

    PyTorch 2.0 introduces torch.compile

    PyTorch 2.0 ships torch.compile, a JIT compiler that accelerates PyTorch code by tracing Python operations and fusing them into optimized kernels via TorchDynamo and Triton. This closes the performance gap with TensorFlow's XLA compilation while preserving PyTorch's dynamic execution model.

  26. 2022 Nov 30

    ChatGPT launches as a research preview

    OpenAI introduced ChatGPT as a conversational model in a public research preview. The release brought a large-language-model interface to a broad public audience.

  27. 2022 Dec

    tsfresh 0.20.0 supports Polars and modern pandas

    Updated the extraction engine to support Polars DataFrames and pandas 2.0, acknowledging that feature extraction speed is a production constraint as much as extraction quality.

  28. 2023

    ISO publishes the 2023 SQL framework edition

    ISO/IEC 9075-1:2023 published the sixth framework edition and references SQL property graph queries as part of the standard family. SQL continues to grow through coordinated parts rather than one vendor's dialect alone.

  29. 2023

    Open-weight models reach competitive quality

    Meta's LLaMA and its derivatives, along with Mistral and other open-weight releases, showed that model quality was no longer gated by proprietary access: teams could fine-tune, deploy, and audit models on their own infrastructure without compromising accuracy.

  30. 2023 Apr

    DINOv2

    Oquab et al. train a 1B-parameter ViT with curated self-supervised learning on 142M images, producing frozen features that rival weakly-supervised CLIP models on classification, segmentation, and depth estimation. The model establishes self-supervised pretraining as a viable path to general-purpose vision backbones.

  31. 2023 Apr

    SAM (Segment Anything Model)

    Meta releases a promptable segmentation foundation model trained on 1.1 billion masks from 11M images. SAM transfers zero-shot to new image distributions via point, box, or text prompts, often matching fully supervised specialist models without any fine-tuning.

  32. 2023 May 9

    Julia 1.9 adds package extensions and native code caching

    Julia 1.9 introduces package extensions (conditional code that loads only when another package is present) and caches native code across sessions, substantially cutting "time to first plot" compilation latency.

  33. 2023 Jul

    ONNX Runtime gains hardware-specific optimization levels

    ONNX Runtime matures into the standard cross-framework inference engine, supporting execution providers for CUDA, TensorRT, OpenVINO, and Core ML. It becomes the practical bridge between training in one framework and deploying on specialized hardware.

  34. 2023 Nov

    functime brings Polars-native feature extraction

    Implemented over one hundred time-series feature extractors as Polars lazy queries, delivering orders-of-magnitude speed-ups over tsfresh and validating the shift toward vectorized feature pipelines.

  35. 2023 Dec 27

    Julia 1.10 ships a new parser

    Julia 1.10 replaces the C-based parser with JuliaSyntax.jl, a parser written in Julia itself, improving parse performance and giving more precise syntax-error locations.

  36. 2024 Jan

    pytimetk unifies time-series feature engineering APIs

    Provided a single function-level API for calendar features, rolling windows, and wavelet transforms across pandas and Polars engines, reducing the framework-switching cost of temporal feature work.

  37. 2024 Jan

    PyTorch Foundation moves to Linux Foundation

    PyTorch transitions from Meta's direct governance to the PyTorch Foundation under the Linux Foundation, ensuring the framework that dominates research remains vendor-neutral. This mirrors the governance model Kubernetes established for infrastructure software.

  39. 2024 Jun

    NumPy 2.0 is released

    NumPy 2.0 marked a major release of the numerical layer beneath many Python analysis workflows.

  40. 2024 Aug

    The EU AI Act enters into force

    Regulation (EU) 2024/1689 entered into force and established a risk-based framework for AI systems. Its transparency and information requirements made explainability relevant to high-risk-system operations.

  41. 2025

    scikit-learn documents version 1.7

    The release history records the continuing maintenance of the library's evaluation and model-selection surface. Practitioners can treat splitters and metrics as maintained infrastructure rather than one-off utility code.

  42. 2025 May

    Keras 3 ships multi-backend support

    Keras 3 enables the same model code to run on TensorFlow, PyTorch, or JAX backends. This decouples model definition from execution backend, making the framework choice in this course's selection guide a runtime decision rather than a code commitment.

  43. 2025 Oct

    PyTorch surpasses 90 percent research adoption

    Framework usage surveys show PyTorch in over 90 percent of deep learning papers at major AI conferences, while TensorFlow's market share stabilizes around 37 percent in production deployments. The research-production gap narrows but does not close.