Training Deep Nets with Sublinear Memory Cost
E899038
"Training Deep Nets with Sublinear Memory Cost" is a research paper that introduces techniques to drastically reduce the memory required for training deep neural networks, enabling the training of larger models or using limited hardware resources more efficiently.
All labels observed (1)
| Label | Occurrences |
|---|---|
| Training Deep Nets with Sublinear Memory Cost canonical | 2 |
How this entity was disambiguated
This entity first appeared as the object of triple T11003402 — 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: Training Deep Nets with Sublinear Memory Cost Context triple: [Łukasz Kaiser, coAuthorOf, Training Deep Nets with Sublinear Memory Cost]
-
A.
Large-Scale Distributed Deep Networks
Large-Scale Distributed Deep Networks is a seminal research work that introduced methods for training deep neural networks efficiently across large-scale distributed computing infrastructure, enabling breakthroughs in modern large-scale AI systems.
-
B.
“A fast learning algorithm for deep belief nets”
“A fast learning algorithm for deep belief nets” is a seminal 2006 paper by Geoffrey Hinton that introduced an efficient unsupervised pretraining method for deep neural networks using stacked restricted Boltzmann machines.
-
C.
“Large-Scale Machine Learning with Stochastic Gradient Descent”
“Large-Scale Machine Learning with Stochastic Gradient Descent” is a widely cited work by Léon Bottou that analyzes and advocates stochastic gradient descent as an efficient optimization method for large-scale machine learning problems.
-
D.
“Stochastic Gradient Descent Tricks”
“Stochastic Gradient Descent Tricks” is a well-known paper by Léon Bottou that surveys practical techniques and heuristics for effectively applying stochastic gradient descent in machine learning.
-
E.
Adam: A Method for Stochastic Optimization
"Adam: A Method for Stochastic Optimization" is a highly influential machine learning paper that introduces the Adam optimizer, a widely used adaptive gradient-based optimization algorithm for training deep neural networks.
- 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: Training Deep Nets with Sublinear Memory Cost Target entity description: "Training Deep Nets with Sublinear Memory Cost" is a research paper that introduces techniques to drastically reduce the memory required for training deep neural networks, enabling the training of larger models or using limited hardware resources more efficiently.
-
A.
Large-Scale Distributed Deep Networks
Large-Scale Distributed Deep Networks is a seminal research work that introduced methods for training deep neural networks efficiently across large-scale distributed computing infrastructure, enabling breakthroughs in modern large-scale AI systems.
-
B.
“A fast learning algorithm for deep belief nets”
“A fast learning algorithm for deep belief nets” is a seminal 2006 paper by Geoffrey Hinton that introduced an efficient unsupervised pretraining method for deep neural networks using stacked restricted Boltzmann machines.
-
C.
“Large-Scale Machine Learning with Stochastic Gradient Descent”
“Large-Scale Machine Learning with Stochastic Gradient Descent” is a widely cited work by Léon Bottou that analyzes and advocates stochastic gradient descent as an efficient optimization method for large-scale machine learning problems.
-
D.
“Stochastic Gradient Descent Tricks”
“Stochastic Gradient Descent Tricks” is a well-known paper by Léon Bottou that surveys practical techniques and heuristics for effectively applying stochastic gradient descent in machine learning.
-
E.
Adam: A Method for Stochastic Optimization
"Adam: A Method for Stochastic Optimization" is a highly influential machine learning paper that introduces the Adam optimizer, a widely used adaptive gradient-based optimization algorithm for training deep neural networks.
- F. None of above. chosen
Statements (40)
| Predicate | Object |
|---|---|
| instanceOf |
research paper
ⓘ
scientific publication ⓘ |
| addresses |
memory bottleneck in deep neural network training
ⓘ
trade-off between memory usage and computation in backpropagation ⓘ |
| aimsTo |
achieve sublinear memory cost with respect to network depth
ⓘ
improve scalability of deep network training ⓘ |
| applicableTo |
backpropagation-based training algorithms
ⓘ
feedforward neural networks ⓘ very deep neural networks ⓘ |
| assumes | standard deep learning training frameworks ⓘ |
| compatibleWith |
GPU-based training
ⓘ
large-scale deep learning workloads ⓘ |
| contributesTo |
efficient deep learning
ⓘ
resource-constrained neural network training ⓘ |
| demonstrates | substantial memory savings compared to standard backpropagation ⓘ |
| enables |
training larger neural network models
ⓘ
training on hardware with limited memory resources ⓘ |
| evaluatedBy | empirical experiments on deep networks ⓘ |
| field |
artificial intelligence
ⓘ
deep learning ⓘ machine learning ⓘ |
| focusesOn |
memory-efficient training of deep neural networks
ⓘ
reducing memory usage during backpropagation ⓘ training deep nets with sublinear memory cost in network depth ⓘ |
| impact |
enables experimentation with deeper architectures on the same hardware
ⓘ
reduces hardware requirements for training large models ⓘ |
| improves | memory footprint during training ⓘ |
| language | English ⓘ |
| mayIncrease | computational overhead due to recomputation ⓘ |
| motivatedBy |
limitations of GPU memory capacity
ⓘ
need to train deeper and larger models ⓘ |
| proposes | techniques to drastically reduce memory required for training deep neural networks ⓘ |
| relatedTo |
efficient backpropagation techniques
ⓘ
gradient checkpointing ⓘ memory-computation trade-offs in neural networks ⓘ |
| title | Training Deep Nets with Sublinear Memory Cost ⓘ |
| typeOfContribution |
algorithmic optimization for training
ⓘ
memory optimization method ⓘ |
| uses |
checkpointing strategies for intermediate activations
ⓘ
recomputation of activations during backpropagation ⓘ |
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: Training Deep Nets with Sublinear Memory Cost Description of subject: "Training Deep Nets with Sublinear Memory Cost" is a research paper that introduces techniques to drastically reduce the memory required for training deep neural networks, enabling the training of larger models or using limited hardware resources more efficiently.
Referenced by (2)
Full triples — surface form annotated when it differs from this entity's canonical label.
subject linked to:
Lukasz Kaiser
Training Deep Nets with Sublinear Memory Cost
→
title
→
Training Deep Nets with Sublinear Memory Cost
ⓘ