Their main goal is behavior reproduction, not just making lots of neurons:
Although impressive scaling has been achieved [in other projects], no previous large-scale spiking neuron models have demonstrated how such simulations connect to a variety of specific observable behaviors… In contrast, we present here a spiking neuron model of 2.5 million neurons that is centrally directed to bridging the brain-behavior gap. Our model embodies neuroanatomical and neurophysiological constraints, making it directly comparable to neural data at many levels of analysis. Critically, the model can perform a wide variety of behaviorally relevant functions. We show results on eight different tasks that are performed by the same model, without modification.
The task:
All inputs to the model are 28 by 28 images of handwritten or typed characters. All outputs are the movements of a physically modeled arm that has mass, length, inertia, etc...Many of the tasks we have chosen are the subject of extensive modeling in their own right, e.g. image recognition, serial working memory and reinforcement learning and others demonstrate abilities that are rare for neural network research and have not yet been demonstrated in spiking networks (e.g., counting, question answering, rapid variable creation, and fluid reasoning)...
The eight tasks (termed “A0” to “A7”) that Spaun performs are: (A0) Copy drawing. Given a randomly chosen handwritten digit, Spaun should produce the same digit written in the same style as the handwriting. (A1) Image recognition. Given a randomly chosen handwritten digit, Spaun should produce the same digit written in its default writing. (A2) Reinforcement Learning Spaun should perform a three-armed bandit task, in which it must determine which of three possible choices generates the greatest stochastically generated reward. Reward contingencies can change from trial to trial. (A3) Serial Working Memory. Given a list of any length, Spaun should reproduce it. (A4) Counting. Given a starting value and a count value, Spaun should write the final value (that is, the sum of the two values) (movie S5). (A5) Question answering. Given a list of numbers, Spaun should answer either one of two possible questions: (i) what is in a given position in the list?… or (ii) given a kind of number, at what position is this number in the list? (A6) Rapid variable creation. Given example syntactic input/output patterns (e.g., 0 0 7 4 → 7 4; 0 0 2 4 → 2 4; etc.), Spaun should complete a novel pattern given only the input (e.g., 0 0 1 4 → ?) . (A7) Fluid reasoning. Spaun should perform a syntactic or semantic reasoning task that is isomorphic to the induction problems from the Raven’s Progressive Matrices (RPM) test for fluid intelligence.
How the model works:
The network implementing the Spaun model consists of three compression hierarchies, an action-selection mechanism, and five subsystems. Components of the model communicate using spiking neurons that implement neural representations that we call “semantic pointers,” using various firing patterns. Semantic pointers can be understood as being elements of a compressed neural vector space… Compression is a natural way to understand much of neural processing. For instance, the number of cells in the visual hierarchy gradually decreases from the primary visual cortex to the inferior temporal cortex, meaning that the information has been compressed from a higher-dimensional (image-based) space into a lower-dimensional (feature) space. This same kind of operation maps well to the motor hierarchy, where lower-dimensional firing patterns are successively decompressed (for example, when a lower-dimensional motor representation in Euclidean space moves down the motor hierarchy to higher-dimensional muscle space).
The five subsystems...: (i) map the visual hierarchy firing pattern to a conceptual firing pattern as needed (information encoding), (ii) extract relations between input elements (transformation calculation), (iii) evaluate the reward associated with the input (reward evaluation), (iv) decompress firing patterns from memory to conceptual firing pattern (information decoding), and (v) map conceptual firing patterns to motor firing patterns and control motor timing (motor processing)… It is critical to note that the elements of Spaun are not task-specific.
The performance of the model:
[On Raven’s Progressive Matrices]..Human participants average 89% correct (chance is 13%) on the matrices that include only an induction rule (5 of 36 matrices). Spaun performs similarly, achieving a match-adjusted success rate of 88%…
[On Serial Working Memory]...As with human data, Spaun produces distinct recency (items at the end are recalled with greater accuracy) and primacy (items at the beginning are recalled with greater accuracy) effects. A good match to human data from a rapid serial-memory task using digits and short presentation times is also evident, with 17 of 22 human mean values within the 95% confidence interval of 40 instances of the model.
[Image Recognition]...for which the model achieves 94% accuracy on untrained data from the MNIST handwriting database (human accuracy is ~98%).
[Reinforcement Learning]...for which the model is able to learn reward-dependent actions in a variable environment using known neural mechanisms
[Counting]...for which the model reproduces human reaction times and scaling of variability
[Question Answering]… for which the model generates a novel behavioral prediction
[Rapid Variable Creation]...for which the model instantiates the first neural architecture able to solve this challenging task
Conclusions:
However, the central purpose of this work is not to explain any one of these tasks, but to propose a unified set of neural mechanisms able to perform them all...Although Spaun’s main contribution lies in its breadth, it also embodies new hypotheses regarding how specific tasks are solved....However, Spaun has little to say about how that complex, dynamical system develops from birth. Furthermore, Spaun has many other limitations that distinguish it from developed brains. For one, Spaun is not as adaptive as a real brain, as the model is unable to learn completely new tasks
Interesting quotes from the article.
Their main goal is behavior reproduction, not just making lots of neurons:
The task:
How the model works:
The performance of the model:
Conclusions: