Objects and their component parts

We all agree the world is compositional. For example a dog is composed of four legs, one tail, one head, one trunk. Additionally, these parts have a specific relationship. The tail is at one end of the trunk, the head is at the other end, the fur legs are attached to the bottom of the trunk, and so on. How are objects and their parts represented in HTMs?

I can’t speak for anyone else, but I think of it like this.

Objects are stored by mapping locations in space to sensory features. Instead of a location mapping to a sensory feature, it could map to another object. Both objects and features are represented as associated collections of SDRs.

Cortical Columns perceive the world with different perspectives, because they are attached to different sensory input. They all learn objects as sensory perceptions at locations in space. Each one has unique representations of its library of objects it has perceived.

As you scan over an object with your senses to learn it, you are learning both the object and all its component parts simultaneously with the same cortical columns. In order to do the things we humans to daily, we must have several simultaneous representations of complex things and their immediate associations in memory. It follows that as you learn a complex object, cortical columns are learning all the component parts, their associations to the parent object, and the parent object simultaneously.