Cognitive Paradigms - The Published Capabilities of NeuroJet

NeuroJet is more than a neural network simulation program. One of it's major purposes is to use essential, biological aspects of the brain to reproduce cognition and behavior. NeuroJet accomplishes this using modeling techniques which minimize computational overhead and mathematical complexity as much as possible. NeuroJet is currently geared towards modeling the hippocampal cognition and behavior as proof-of-concept. The program is designed in a way that compartmentalizes biological functionality so that extending its use to other regions will be easier. Researchers already use NeuroJet to replicate a number of learning paradigms. The chart below attempts to concisely summarize its successes. Further details can be found in the published articles which are partially cited below the chart.

 Cognitive, Behavioral, and Cellular Predictions and Explanations by the Model

Spatial Tasks

Configural Tasks

Trace Conditioning

Paradigms

Simple sequence completion (various)

Transverse Patterning (8,11,12)

Trace Conditioning (16,17)

One trial learning (1, 19)

Transverse Non-Patterning (NP1) (13)


Jump ahead recall (2,3)

Transitive Inference (14,15,24)


Circular sequence completion (4,5)



Sequence Disambiguation (6,7,8)



Shortcut finding (6)



Goal finding (6)



Combining appropriate subsequences (9,10)



Demonstrations and Observations

Simple sequence completion (1,18)

1. Memory capacity (9)

2. One trial learning (1)


Circular sequence completion

1. Off-line Compression (2, 4,5)

2. On-line Compression vs.

noise, context length (8)

3. Off-line Spontaneous replay

(3,4,5)


Cell firing ahead of place (16,17, 23, 25, 26; see also on-line compression in 26)



Transverse Patterning

1. Learning paradigms

(Concurrent, staged,

progressive – 20)

2. Learning rates (12)


NP1 learning rates (13)


Transitive Inference population variability (21)



Maximum learnable trace interval

(17, 23)


Number of trials required to learn (17)


Different classes of neurons that bridge the trace interval (17)


Trace prediction is nonlinear in number of training trials (17)





Predictions

T-maze choice point decision (22)

Lack of stimulus encoding neurons during trace interval (17)


Relative neuronal codes (20)

Increasing CS/US longevity increases learnable trace interval (23)

1. Minai and Levy, 1993b; 2. August and Levy, 1996; 3. Prepscius and Levy, 1994; 4. Levy, Sederberg and August, 1998; 5. August and Levy, 1999 appended; 6. Levy, Wu, And Baxter, 1995; 7. Minai, Barrows, and Levy, 1994; 8. Wu, Baxter, and Levy 1996; 9. Levy and Wu, 1996; 10. Wu and Levy, 1996; 11. Levy, Wu, and Tyrcha, 1996; 12. Wu, Tyrcha and Levy, 1998; 13. Wu and Levy, 2002; 14. Wu and Levy, 1998; 15. Wu and Levy, 2001; 16. Levy and Sederberg, 1997; 17. Rodriguez and Levy, 2001, appended; 18. Amarasingham and Levy 1998; 19. Greene et al. 2000; 20. Shon, Wu and Levy 2000; 21. Levy et al. 2003; 22. Monaco and Levy 2003; 23. Wu and Levy 2005a; 24. Smith, Wu and Levy 2000, appended; 25. Mitman et al. 2003; 26. Levy et al. 2005a