###### tags: `Prof Thomas` `Brown`

# Framework for intelligence and cortical function based on grid cells in the network - 2019

## Some interesting points

1. Vernon Mountscale proposed that all the regions of the neocortex are fundamentally the same. What distinguishes one region from another, he argued, is mostly determined by the inputs to a region and not by differences in intrinsic circuitary and function. He further proposed that a small volume of cortex, a cortical column, is the unit of replication.
2. Compositional structure is present in almost all object in the world, therefore cortical columns must have a neural mechanism that represents a new object as an arrangment of previously-learned objects. For example, a logo on a coffee cup. The task of representing the log on the cup can be achieved by creating a "displacement" vector that converts any point in cup space to the equivalent point in logo space.
3. Cell activity in multiple displacement cell modules represents a unique displacement in much the same way as the cell activity in the multiple grid cell modules represents a unique location. Hence, a single displacement vector can represent the logo on the coffee cup at a specific relative position. A displacement vector not only represents the relative position of two objects, it also is unique to the two objects.
4. Grid cells and displacements cells perform complementary operations
**Grid cells : (Location 1 + Displacement => Location2)**
**Displacement cells : (Location2 - Location1 => Displacement)**
5. If the location are in the same psace, then grid cells and displacement cells are useful for navigation. If the two locations are in different spaces (that is the same physical location relative to two different objects), then grid cells and displacement cells are useful for representing the relative position of two objects.
6. Object behaviors - A stapler is considered. The top and and bottom parts of the stapler can move relative to each other, but the opened and closed stapler are not reffered to as different object. Thus, the movement of a part of the object relative to other parts of the object is a "behavior" of the object. The relative position of the top and the bottom parts is represented by a displacement vector in the same way as the relative position of the logo and the coffee cup. However, unlike the relative position of the logo and the coffee cup, the two halves of the stapler can move relative to each other. As the stapler top rotates upwards, the displacement of the stapler top to bottom changes. Thus, rotation of the stapler is represented by a sequence of displacement vectors. Opening and closing of the stapler are different behaviors yet they are composed of the same displacement elements, just in reverse order. These are sometimes referred to as **"high order sequences"**.
7. Regions of the neocortex are organized in a hierarchy. It is commonly believed that when sensory input enters the neocortex, the first region detects simple features. The ouput of this region is passed to a second region that combines simple features into more complex features. This process is repeated until, several levels up in the hierarchy, cells respond to complete objects.
8. The Thousand Brains Theory of Intelligence - One of the classic questions about perception is how does the neocortex fuse different sensory inputs into a unified model of a perceived object. We propose that the neocortex implements a decentralized model of sensor fusion. For example, there is no single model of a coffee cup that includes what a cup feels like and looks like. Instead there are 100s of models of a cup. Each model can infer the cup on it own by ovserving input over movements of its associated sensors. The main idea is that neocortex has 100s, likely 1000s of models of each object in the world. The integration of observed features does not just occur at the top of the hierarchy, it occurs in every column at all levels of the hierarchy. This is know as "The Thousand Brains Theory of Intelligence".
9. Grid cells in the neocortex suggests that all knowledge is learned and stored in the context of locations and location spaces and that "thinking" is movement through those location spaces.

## Some important concepts/terms to know
1. Grid Cells - A grid cell is a type of neuron within a entorhinal cortex that fires at regular intervals as an animal navigates an open area, allowing it to understand its position in space by storing and integrating information about the location, distance and direction. A grid cell is a neuron that becomes active at multiple locations in an environmen, typically in a grid-like, or tiled, traingular lattice. A "grid cell module" is a set of grid cells that activate with the same lattice spacing and orientation but at shifted locations within the environment. As the animcal moves, the active grid cells in a grid cell module change to reflect the animal's updated location. This change occurs even if the animal is in the dark, telling us that grid cells are updated using an internal, or "efference", copy of motor commands. This process, called "path integration", has the desirable property that regardless of the path of movement, when the animal returns to the same physical location, then the same grid cells in a module will be active.

2. Place Cells - A place cell is a kind of pyramidal neuron within the hippocampus that becomes active when an animal enters a particular place in its environment, which is known as the place field. Place cells are thoughts, collectively, to act as a cognitive representation of a specific location in space, known as cognitive map.
3. Hippocampus - Hippocampus is a complex brain structure embedded deep into temporal lobe. It has a major role in learning and memory. It is a plastic and vulnerable structure that gets damaged by a variety of stimuli. Studies have shown that it also gets affected in a variety of neurological and psychiatric disorders.
4. Cortical column - A cortical column, also called hypercolumn, macrocolumn, functional column or sometimes cortical module, is a group of neurons in the cortex of the brain that can be successively penetrated by a probe inserted perpendiculary to the cortical surface, and which have nearly identical receptive fields. ***In this particular context, the cortical column is being thought of as being about one square millimeter of cortical area (although this size is not critical and could vary by species and region).***
5. Entorhinal cortex - The entorhinal cortex is an area of the brain's allocortex, located in the medial temporal lobe, whose function includes being a widespread network hub for memory, navigation and the perception of time. **The EC is the main interface between the hippocampus and the neocortex.**
6. Allocortex - It a a much smaller area of the cortex taking just about 10%, the neocortex takes up the rest 90%. It is characterized by having just threee or four cortical layers, in contrast with six layers of the neocortex.
7. Displacement Cells - Displacement cells determine what displacement is required to reach a new location from the current location. If the two locations are in the same space, then grid cells and displacements cells are useful for navigation. In this case, grid cells predict a new location based on a starting location and a given movement.
8. Allocentric - Allocentrism is a collectivistic personality attribute whereby people center their attention and actions on other people rather than themselves. (think of the use in the paper) [link Here's the comment on allocentric versus egocentric reference frames (in the context of spatial cognition) that I wanted to make regarding this entry and the next in your glossary definitions: Within an egocentric reference frame, spatial information is encoded in terms of relations to the physical body of a navigator, whereas the allocentric reference frame defines relations of objects among each other, that is independent of the physical body of an "observer" and thus in a more absolute way, which takes metrical conditions and general alignments like cardinal directions into account. This suggests, that route knowledge, which is supported by direct navigation, is more likely to be encoded within an egocentric reference frame and survey knowledge, which is supported by map learning, to be more likely to be encoded within an allocentric reference frame in turn.](https://en.wikipedia.org/wiki/Spatial_cognition)
9. Egocentric - Thinking only of oneself, without regard for the feelings of desires of others. (think of the use in the paper)
10. Head direction cells - Head direction cells are neurons found in several brain areas that discharge in relation to the animal's bidirectional heading with respect to the environment in the horizontal (yaw) plane. These cells represent the allocentric location of an animal relative to its environment. Inferring where you are via sensation, predicting what you will sense after moving, and determining how to move to get to a new location all require knowing your current orientation relative to your environment.

## Questions
1. Under the topic `Grid Cells and Displacement Cells Perform Complementary Operations`, locations are defined to be in the `same space`, or in `different spaces`. The term `different spaces`, means that two different objects have the same physical location. Isnt the nomenclature confusing, or am I missing