Machine learning foundations

Publishing authors

1,378 distinct OpenAlex Author IDs across all 575 matching papers. This is an observed count within the query result; no publication-rate assumption is used.

Full query result: 575 of 575 papers (100.0%). 14.7% of authorship records lack an Author ID; 13.6% of retrieved works lack an abstract. Retrieved 2026-09-26.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $215M.

AI infrastructure/models/research/governance: $143.2B × 0.15%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

ICML 2026 — research topics

Scope and related research

Covers learning guarantees. Training algorithms are listed under optimization.

Organizations

Selected examples

Google Research Research
Learning theory, optimization, reinforcement learning and differential privacy research.
Microsoft Research
Microsoft Research studies theoretical foundations of machine learning.
Amazon / AWS Research
Research establishes generalization guarantees for learned ensemble strategies.

Training optimization

Stochastic and distributed optimizers, learning-rate schedules, training stability, and matrix methods.

Optimizers Training stability

2,045 papers

1.8× vs. 2023

Authors · scenario: 3.2k–6.7k
Investment · scenario: $430M–$1.72B

Sources and methodology

Publication activity

2019: 581
2020: 799
2021: 995
2022: 1,036
2023: 1,166
2024: 1,460
2025: 2,045

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("stochastic gradient descent" OR "neural network optimization" OR "adaptive optimizer" OR "learning rate schedule" OR "sharpness aware minimization" OR "second order optimization") AND (("machine learning" OR "deep learning" OR "neural network" OR "artificial intelligence"))

Retrieved 2026-09-26.

Publishing authors

Base scenario: 3.2k publishing authors. Paper count × 3.29 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 3,188 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 2,045 papers (48.9%). 13.2% of authorship records lack an Author ID; 20.3% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 3.29; 95% bootstrap interval 3.13–3.44. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $859M.

AI infrastructure/models/research/governance: $143.2B × 0.60%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

ICML 2026 — research topics

Scope and related research

Focuses on learning model parameters. Scheduling and routing problems sit under combinatorial optimization.

Organizations

Selected examples

Google Research Research
Learning theory, optimization, reinforcement learning and differential privacy research.
Microsoft Research
Research on optimization methods for machine learning.
Google DeepMind Research
Optax provides gradient transformations and optimization algorithms.

More organizations (1)

Meta Research
Schedule-free optimization studies training methods that do not require learning-rate decay schedules.

Representation learning

Self-supervised, contrastive and semi-supervised learning, clustering, and dimensionality reduction.

Self-supervised learning Contrastive learning

16,868 papers

1.5× vs. 2023

Authors · scenario: 15k–73k
Investment · scenario: $2.11B–$8.46B

Sources and methodology

Publication activity

2019: 3,062
2020: 4,729
2021: 6,676
2022: 8,888
2023: 11,349
2024: 14,113
2025: 16,868

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("self supervised learning" OR "representation learning" OR "contrastive learning" OR "semi supervised learning" OR "deep clustering" OR "unsupervised learning")

Retrieved 2026-09-26.

Publishing authors

Base scenario: 24k publishing authors. Paper count × 4.30 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 4,227 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 16,868 papers (5.9%). 11.8% of authorship records lack an Author ID; 26.0% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 4.30; 95% bootstrap interval 4.14–4.46. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $4.23B.

AI infrastructure/models/research/governance: $143.2B × 1.00%
Pharmaceutical: $10.6B × 15.00%
Biotech: $4.84B × 25.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

ICML 2026 — research topics KDD 2026 — research topics

Scope and related research

Methods used across data types; application areas provide a separate view.

Organizations

Selected examples

Meta Research
DINOv3 provides self-supervised visual representations.
OpenAI Research
CLIP learns visual representations through image-text contrastive training.
Google Research Research
SimCLR research learns visual representations through contrastive training.

More organizations (1)

Apple Machine Learning Research Research
MobileCLIP studies efficient contrastive image-text representation learning.

Transfer and continual learning

Domain adaptation, few-shot and meta-learning, continual learning, and distribution shift.

Continual learning Domain adaptation Few-shot learning

20,365 papers

1.5× vs. 2023

Authors · scenario: 16k–81k
Investment · scenario: $358M–$1.43B

Sources and methodology

Publication activity

2019: 3,576
2020: 6,137
2021: 8,757
2022: 10,794
2023: 13,205
2024: 16,094
2025: 20,365

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("transfer learning" OR "domain adaptation" OR "continual learning" OR "meta learning" OR "few shot learning" OR "domain generalization")

Retrieved 2026-09-26.

Publishing authors

Base scenario: 27k publishing authors. Paper count × 3.97 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 3,928 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 20,365 papers (4.9%). 11.3% of authorship records lack an Author ID; 21.3% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 3.97; 95% bootstrap interval 3.76–4.23. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $716M.

AI infrastructure/models/research/governance: $143.2B × 0.50%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

ICML 2026 — research topics CVPR 2026 — research topics

Scope and related research

Covers transferring and retaining knowledge. Model-specific fine-tuning also appears under post-training.

Organizations

Selected examples

OpenAI Research
CLIP research evaluates zero-shot transfer from image-text pretraining.
Google Research Research
Meta-Dataset studies few-shot learning and generalization to unseen datasets.
Physical Intelligence Research
Generalist robot policies learned across tasks and robot embodiments.

More organizations (1)

Apple Machine Learning Research Research
MobileCLIP evaluates zero-shot classification and retrieval across downstream datasets.

Probabilistic ML and uncertainty

Bayesian inference, graphical models, Monte Carlo, probabilistic programming, calibration, and conformal prediction.

Bayesian ML Conformal prediction

4,617 papers

1.8× vs. 2023

Authors · scenario: 3.9k–19k
Investment · scenario: $1.56B–$6.24B

Sources and methodology

Publication activity

2019: 824
2020: 1,303
2021: 1,713
2022: 2,181
2023: 2,549
2024: 3,202
2025: 4,617

Many matches apply probabilistic methods. This count does not measure the size of the methods research community.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("Bayesian neural network" OR "Gaussian process" OR "probabilistic graphical model" OR "probabilistic programming" OR "uncertainty quantification" OR "conformal prediction") AND (("machine learning" OR "deep learning" OR "neural network" OR "artificial intelligence"))

Retrieved 2026-09-26.

Publishing authors

Base scenario: 6.3k publishing authors. Paper count × 4.11 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 3,926 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 4,617 papers (21.7%). 9.2% of authorship records lack an Author ID; 24.3% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 4.11; 95% bootstrap interval 3.89–4.38. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $3.12B.

AI infrastructure/models/research/governance: $143.2B × 0.20%
Medical and healthcare: $11.8B × 5.00%
Pharmaceutical: $10.6B × 10.00%
Biotech: $4.84B × 15.00%
Energy management: $4.64B × 10.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

ICML 2026 — research topics AAAI 2026 — AI taxonomy

Scope and related research

Causal inference is listed separately. Uncertainty research extends beyond language-model hallucinations.

Organizations

Selected examples

PyMC Labs Tools & infrastructure
PyMC Labs develops Bayesian modeling tools and supports probabilistic inference with PyMC.
Amazon / AWS Research
Published research on probabilistic forecasting and time series.
Microsoft Tools & infrastructure
Infer.NET provides probabilistic programming and Bayesian inference.

More organizations (1)

Google Research Research
Research develops automated structured variational inference for probabilistic programs.

Causal inference

Causal discovery, causal representations, intervention effects, and uplift modeling.

Causal ML Causal discovery

1,661 papers

2.8× vs. 2023

Authors · scenario: 3.5k–6.3k
Investment · scenario: $1.32B–$5.28B

Sources and methodology

Publication activity

2019: 186
2020: 286
2021: 391
2022: 505
2023: 588
2024: 849
2025: 1,661

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("causal inference" OR "causal discovery" OR "causal representation" OR "causal machine learning" OR "treatment effect" OR "uplift modeling") AND (("machine learning" OR "deep learning" OR "neural network" OR "artificial intelligence"))

Retrieved 2026-09-26.

Publishing authors

Base scenario: 3.5k publishing authors. Paper count × 3.79 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 3,470 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 1,661 papers (60.2%). 12.3% of authorship records lack an Author ID; 19.1% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 3.79; 95% bootstrap interval 3.53–4.05. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $2.64B.

AI infrastructure/models/research/governance: $143.2B × 0.15%
Marketing, digital ads: $2.51B × 5.00%
Medical and healthcare: $11.8B × 5.00%
Pharmaceutical: $10.6B × 10.00%
Fintech: $6.52B × 10.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

ICML 2026 — research topics AAAI 2026 — AI taxonomy

Scope and related research

Focuses on cause and effect rather than prediction alone.

Organizations

Selected examples

Microsoft Research
Causal machine learning.
Uber Tools & infrastructure
Uber-origin CausalML provides uplift modeling and treatment-effect estimation.
PyMC Labs Tools & infrastructure
CausalPy supports Bayesian causal analysis of quasi-experiments.

More organizations (1)

IBM Tools & infrastructure
Causal 360 provides methods for causal inference and intervention analysis.

AutoML and experimental design

Model selection, hyperparameter search, Bayesian optimization, active learning, and choosing the next experiment.

AutoML Bayesian optimization Active learning

4,822 papers

1.8× vs. 2023

Authors · scenario: 4.1k–20k
Investment · scenario: $1.79B–$7.17B

Sources and methodology

Publication activity

2019: 751
2020: 1,256
2021: 1,780
2022: 2,026
2023: 2,624
2024: 3,466
2025: 4,822

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("automated machine learning" OR "AutoML" OR "neural architecture search" OR "Bayesian optimization" OR "active learning" OR "optimal experimental design") AND (("machine learning" OR "deep learning" OR "neural network" OR "artificial intelligence"))

Retrieved 2026-09-26.

Publishing authors

Base scenario: 6.8k publishing authors. Paper count × 4.24 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 4,100 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 4,822 papers (20.7%). 9.4% of authorship records lack an Author ID; 26.2% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 4.24; 95% bootstrap interval 4.03–4.49. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $3.59B.

AI infrastructure/models/research/governance: $143.2B × 0.35%
Pharmaceutical: $10.6B × 20.00%
Biotech: $4.84B × 20.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

ICML 2026 — research topics AAAI 2026 — AI taxonomy

Scope and related research

Allocates the search or observation budget rather than optimizing model weights.

Organizations

Selected examples

Preferred Networks Tools & infrastructure
Optuna automates hyperparameter search and experiment selection.
Meta Tools & infrastructure
Ax and BoTorch provide adaptive experimentation and Bayesian optimization.
H2O.ai Product
Automated ML and predictive modeling tools.

More organizations (3)

Citrine Informatics Product
Materials informatics.
DataRobot Product
Predictive AI and model operations.
Secondmind Product
Data-efficient engineering optimization.

Data-centric AI

Data curation, labeling, synthetic data, provenance, deduplication, and leakage prevention.

Synthetic data Data curation

10,400 papers

3.1× vs. 2023

Authors · scenario: 7.6k–38k
Investment · scenario: $9.58B–$38.3B

Sources and methodology

Publication activity

2019: 734
2020: 1,204
2021: 1,685
2022: 2,252
2023: 3,362
2024: 6,120
2025: 10,400

Data and quality are broad terms, so this query may include substantial noise from applied work.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("data curation" OR "synthetic data" OR "data quality" OR "data cleaning" OR "data annotation" OR "training data selection" OR "data contamination") AND (("machine learning" OR "deep learning" OR "neural network" OR "artificial intelligence") OR ("language model" OR "large language model" OR "vision language model"))

Retrieved 2026-09-26.

Publishing authors

Base scenario: 13k publishing authors. Paper count × 3.66 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 3,642 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 10,400 papers (9.6%). 11.2% of authorship records lack an Author ID; 14.4% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 3.66; 95% bootstrap interval 3.46–3.89. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $19.2B.

Data management, processing: $31.6B × 55.00%
Pharmaceutical: $10.6B × 10.00%
Biotech: $4.84B × 15.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.