RECORD
Maze Learning using a Hyperdimensional Predictive Processing Cognitive Architecture
Artificial Intelligence · Submitted 31 Mar 2022
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim We test CogNGen on four maze-learning tasks, including those that test memory and planning, and find that CogNGen matches performance of deep reinforcement learning models and exceeds on a task designed to test memory.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
We present the COGnitive Neural GENerative system (CogNGen), a cognitive architecture that combines two neurobiologically-plausible, computational models: predictive processing and hyperdimensional/vector-symbolic models. We draw inspiration from architectures such as ACT-R and Spaun/Nengo. CogNGen is in broad agreement with these, providing a level of detail between ACT-R's high-level symbolic description of human cognition and Spaun's low-level neurobiological description, furthermore creating the groundwork for designing agents that learn continually from diverse tasks and model human performance at larger scales than what is possible with current systems. We test CogNGen on four maze-learning tasks, including those that test memory and planning, and find that CogNGen matches performance of deep reinforcement learning models and exceeds on a task designed to test memory.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | We present the COGnitive Neural GENerative system (CogNGen), a cognitive architecture that combines two neurobiologically-plausible, computational models: predictive processing and hyperdimensional/vector-symbolic models. | contribution 0.91 | 0.83 | 0.07 | asserted 1.00 |
| s1 | We draw inspiration from architectures such as ACT-R and Spaun/Nengo. | background 0.96 | 0.15 | 0.02 | asserted 0.98 |
| s2 | CogNGen is in broad agreement with these, providing a level of detail between ACT-R's high-level symbolic description of human cognition and Spaun's low-level neurobiological description, furthermore creating the groundwork for designing agents that learn continually from diverse tasks and model human performance at larger scales than what is possible with current systems. | contribution 0.39 | 0.30 | 0.11 | asserted 0.89 |
| s3 | We test CogNGen on four maze-learning tasks, including those that test memory and planning, and find that CogNGen matches performance of deep reinforcement learning models and exceeds on a task designed to test memory. | result 1.00 | 0.97 | 0.74 | asserted 1.00 |
What was examined, and what was not
- arXiv listing abstract v2
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
We present the COGnitive Neural GENerative system (CogNGen), a cognitive architecture that combines two neurobiologically-plausible, computational models: predictive processing and hyperdimensional/vector-symbolic models. We draw inspiration from architectures such as ACT-R and Spaun/Nengo. CogNGen is in broad agreement with these, providing a level of detail between ACT-R's high-level symbolic description of human cognition and Spaun's low-level neurobiological description, furthermore creating the groundwork for designing agents that learn continually from diverse tasks and model human performance at larger scales than what is possible with current systems. We test CogNGen on four maze-learning tasks, including those that test memory and planning, and find that CogNGen matches performance of deep reinforcement learning models and exceeds on a task designed to test memory.
Human review
How the literature uses it
1 arXiv paper cite this in 1 sentence
What it cites · 5 sentences without a confident label
Builds on 1 paper
-
Building upon the notion of planning-as-inference [ref] , as in [ref] , we generalize NGC to the case of action-driven tasks, which we call active neural generative coding (ANGC).
Backprop-Free Reinforcement Learning with Active Neural Generative Coding · Neural Generative Motor Control · 0.98
Uses the method or tool of 3 papers
-
To generate the value of the epistemic reward [ref] ), the dynamics model first settles to a prediction $ẑ_t+1$ using the equations above given the value of CogNGen's next latent state $𝐳_t+1$.
Backprop-Free Reinforcement Learning with Active Neural Generative Coding · Neural Generative Procedural Memory · 0.90
-
The BeBold DQN-CNT utilized global state visitation counts to compute its intrinsic reward bonus (meaning that we implemented and tuned the “episodic restriction on intrinsic reward”, or ERIR, model in [ref] ).
BeBold: Exploration Beyond the Boundary of Explored Regions · Baseline Models · 0.99
-
All NGC circuit synaptic efficacies were adjusted by computing their updates as described in the earlier in Section [ref] (the learning step of CogNGen) and using the Adam adaptive learning rate [ref] , with a step size $η= 0.0005$, to ultimately apply the synaptic parameter adjustments.
Adam: A Method for Stochastic Optimization · CogNGen Model Instantiation · 1.00
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.