tensor programs framework
E102299
The tensor programs framework is a theoretical approach developed by Greg Yang that rigorously analyzes and characterizes the behavior and scaling limits of large neural networks using tools from probability and random matrix theory.
All labels observed (1)
| Label | Occurrences |
|---|---|
| tensor programs framework canonical | 1 |
How this entity was disambiguated
This entity first appeared as the object of triple T871414 — resolving that mention is where its identity was fixed. The disambiguator weighed these candidate entities and picked the highlighted one (or “None”, minting a new entity). This is how homonymy is resolved: the same surface form can point to different entities.
NED1
Entity disambiguation (via context triple)
gpt-5-mini-2025-08-07
Target entity: tensor programs framework Context triple: [Greg Yang, notableConcept, tensor programs framework]
-
A.
Swift for TensorFlow
Swift for TensorFlow is an experimental machine learning platform that integrates TensorFlow directly into the Swift programming language to enable differentiable programming and high-performance model development.
-
B.
PlaidML
PlaidML is an open-source, hardware-agnostic deep learning engine designed to accelerate neural network computation on a wide range of GPUs and other devices.
-
C.
PyTorch
PyTorch is an open-source deep learning framework widely used for building and training neural networks, known for its dynamic computation graph and strong support for research and production in Python.
-
D.
Theano
Theano is an open-source numerical computation library for Python that allows efficient definition, optimization, and evaluation of mathematical expressions, particularly those involving multi-dimensional arrays, and was widely used as a backend for deep learning frameworks.
-
E.
TensorFlow
TensorFlow is an open-source, end-to-end machine learning and deep learning framework widely used for building, training, and deploying neural network models at scale.
- F. None of above. chosen
- G. Unsure - the case is ambiguous/there is not enough information to decide.
NED2
Entity disambiguation (via description)
gpt-5-mini-2025-08-07
Target entity: tensor programs framework Target entity description: The tensor programs framework is a theoretical approach developed by Greg Yang that rigorously analyzes and characterizes the behavior and scaling limits of large neural networks using tools from probability and random matrix theory.
-
A.
Swift for TensorFlow
Swift for TensorFlow is an experimental machine learning platform that integrates TensorFlow directly into the Swift programming language to enable differentiable programming and high-performance model development.
-
B.
PlaidML
PlaidML is an open-source, hardware-agnostic deep learning engine designed to accelerate neural network computation on a wide range of GPUs and other devices.
-
C.
PyTorch
PyTorch is an open-source deep learning framework widely used for building and training neural networks, known for its dynamic computation graph and strong support for research and production in Python.
-
D.
Theano
Theano is an open-source numerical computation library for Python that allows efficient definition, optimization, and evaluation of mathematical expressions, particularly those involving multi-dimensional arrays, and was widely used as a backend for deep learning frameworks.
-
E.
TensorFlow
TensorFlow is an open-source, end-to-end machine learning and deep learning framework widely used for building, training, and deploying neural network models at scale.
- F. None of above. chosen
Statements (42)
| Predicate | Object |
|---|---|
| instanceOf |
neural network theory framework
ⓘ
theoretical framework ⓘ |
| analyzes |
behavior of wide neural networks
ⓘ
scaling behavior of deep neural networks ⓘ |
| appliesTo |
convolutional neural networks
ⓘ
fully connected neural networks ⓘ transformer-style architectures ⓘ |
| associatedWith |
Greg Yang's research on deep learning limits
ⓘ
Microsoft Research ⓘ
linked to:
Microsoft
|
| assumes |
large-width asymptotics
ⓘ
random initialization of network parameters ⓘ |
| basedOn |
Gaussian process limits
ⓘ
probabilistic limit theorems ⓘ random matrix theory techniques ⓘ |
| characterizes |
distributional behavior of activations at initialization
ⓘ
gradient behavior in wide networks ⓘ scaling of parameters with width and depth ⓘ |
| developer | Greg Yang ⓘ |
| enables |
rigorous proofs of convergence of network statistics
ⓘ
systematic study of architectural variations at infinite width ⓘ |
| field |
deep learning theory
ⓘ
machine learning theory ⓘ probability theory ⓘ random matrix theory ⓘ |
| hasConcept |
master theorem for tensor programs
ⓘ
program limit ⓘ tensor program ⓘ |
| influencedBy |
Gaussian process theory
ⓘ
classical random matrix theory ⓘ |
| influences |
design of scalable neural network architectures
ⓘ
theoretical understanding of large-scale deep learning ⓘ |
| provides |
a language for describing tensor computations in neural nets
ⓘ
rigorous conditions for infinite-width limits ⓘ tools for analyzing signal propagation in deep networks ⓘ |
| purpose |
characterization of scaling limits of neural networks
ⓘ
rigorous analysis of large neural networks ⓘ |
| relatedTo |
infinite-width neural networks
ⓘ
neural tangent kernel ⓘ scaling laws in deep learning ⓘ |
| usedFor |
deriving kernel limits of neural networks
ⓘ
designing scaling rules for neural network hyperparameters ⓘ understanding training dynamics in the infinite-width limit ⓘ |
How these facts were elicited
The pipeline generated the facts above by prompting gpt-5.1 with this entity's name + description and the instruction below.
Instruction
You are a knowledge base construction expert. Given a subject entity and a description of it, return factual statements that you know for the subject as a JSON list of dictionaries(triples), where keys must be "subject", "predicate" and "object". The number of facts may be very high, between 25 to 50 or more, for very popular subjects. For less popular subjects, the number of facts can be very low, like 5 or 10. # Requirements - If you don't know the subject at all, return an empty list. - If the subject is not a named entity, return an empty list. - Include at least one triple where predicate is "instanceOf". - Do not get too wordy. - Separate several objects into multiple triples with one object.
Input
Subject: tensor programs framework Description of subject: The tensor programs framework is a theoretical approach developed by Greg Yang that rigorously analyzes and characterizes the behavior and scaling limits of large neural networks using tools from probability and random matrix theory.
Referenced by (1)
Full triples — surface form annotated when it differs from this entity's canonical label.