IMPALA
E428323
deep reinforcement learning architecture
distributed reinforcement learning system
scalable RL architecture
IMPALA is a scalable deep reinforcement learning architecture designed for efficient distributed training of agents across many tasks and environments.
All labels observed (1)
| Label | Occurrences |
|---|---|
| IMPALA canonical | 2 |
How this entity was disambiguated
This entity first appeared as the object of triple T4293699 — 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: IMPALA Context triple: [A3C, inspiredAlgorithms, IMPALA]
-
A.
Apache Hive
Apache Hive is a data warehouse and SQL-like query system built on top of Hadoop for managing and analyzing large datasets stored in distributed storage.
-
B.
Presto
Presto is an open-source, distributed SQL query engine designed for fast, interactive analytics on large-scale data from multiple sources.
-
C.
Presto
Presto is a discontinued proprietary browser engine developed by Opera Software that powered older versions of the Opera web browser.
-
D.
Apache Pig
Apache Pig is a high-level platform for creating MapReduce programs used to analyze large data sets in the Hadoop ecosystem.
-
E.
Apache Spark
Apache Spark is an open-source, distributed data processing engine designed for large-scale data analytics, machine learning, and stream processing.
- 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: IMPALA Target entity description: IMPALA is a scalable deep reinforcement learning architecture designed for efficient distributed training of agents across many tasks and environments.
-
A.
Apache Hive
Apache Hive is a data warehouse and SQL-like query system built on top of Hadoop for managing and analyzing large datasets stored in distributed storage.
-
B.
Presto
Presto is an open-source, distributed SQL query engine designed for fast, interactive analytics on large-scale data from multiple sources.
-
C.
Presto
Presto is a discontinued proprietary browser engine developed by Opera Software that powered older versions of the Opera web browser.
-
D.
Apache Pig
Apache Pig is a high-level platform for creating MapReduce programs used to analyze large data sets in the Hadoop ecosystem.
-
E.
Apache Spark
Apache Spark is an open-source, distributed data processing engine designed for large-scale data analytics, machine learning, and stream processing.
- F. None of above. chosen
Statements (47)
| Predicate | Object |
|---|---|
| instanceOf |
deep reinforcement learning architecture
ⓘ
distributed reinforcement learning system ⓘ scalable RL architecture ⓘ |
| affiliation |
DeepMind Technologies
ⓘ
linked to:
DeepMind
|
| architectureType | actor-critic ⓘ |
| citationVenue |
ICML 2018
ⓘ
linked to:
ICML
|
| comparedWith |
A2C
ⓘ
A3C ⓘ |
| contribution | demonstrated scalable distributed deep RL with stable learning ⓘ |
| designedFor |
distributed training of agents
ⓘ
multi-task reinforcement learning ⓘ scalable deep reinforcement learning ⓘ |
| developedBy | DeepMind ⓘ |
| enables | training with thousands of actors ⓘ |
| evaluationDomain |
Atari
ⓘ
linked to:
Atari, Inc.
DeepMind Lab ⓘ multi-task environments ⓘ |
| field |
artificial intelligence
ⓘ
deep learning ⓘ reinforcement learning ⓘ |
| fullName | Importance Weighted Actor-Learner Architectures ⓘ |
| handles |
large-scale distributed training
ⓘ
off-policy data ⓘ policy lag between actors and learner ⓘ |
| hasAlgorithm | V-trace ⓘ |
| improves |
data efficiency
ⓘ
scalability ⓘ throughput ⓘ |
| keyIdea |
decouple acting from learning via distributed actors and a central learner
ⓘ
use importance weighting to correct for policy lag ⓘ |
| language | implemented primarily in TensorFlow in the original work ⓘ |
| notableComponent | V-trace off-policy correction algorithm ⓘ |
| optimizationMethod |
policy gradient
ⓘ
value-based learning ⓘ |
| outperforms |
A2C on large-scale multi-task benchmarks
ⓘ
A3C on large-scale multi-task benchmarks ⓘ
linked to:
Asynchronous Advantage Actor-Critic
|
| paperTitle | IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures ⓘ |
| publishedIn |
International Conference on Machine Learning
ⓘ
linked to:
ICML
|
| supports |
large-scale experiments
ⓘ
many tasks and environments ⓘ multi-task learning ⓘ |
| uses |
V-trace
ⓘ
actor-learner architecture ⓘ centralized learner ⓘ distributed actors ⓘ off-policy correction ⓘ |
| yearIntroduced | 2018 ⓘ |
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: IMPALA Description of subject: IMPALA is a scalable deep reinforcement learning architecture designed for efficient distributed training of agents across many tasks and environments.
Referenced by (2)
Full triples — surface form annotated when it differs from this entity's canonical label.