Welcome to the Creatures Wiki! Log in and join the community.

Concept lobe

From Creatures Wiki
Jump to: navigation, search

The concept lobe or concept space is the largest lobe in the brain and stores memories of events that occur over a creature's lifetime. This recognition of situations allows the concept lobe to provide processed sensory information to the decision lobe. In Nova Subterras, it also provides this information to the attention lobe. It serves a similar role to the combination lobe of the Creatures Evolution Engine games but has a more limited capacity to represent information.

Sensory Processing[edit]

The concept lobe's neurons represent the situations a creature is able to recognize while interpreting the raw sensory inputs. Each neuron activates in the presence of the unique situation it is currently assigned to detect. For example, a neuron could represent "I am hungry, and I am looking at food". This recognition is achieved by having some inputs (between one and three of them) from neurons in the perception lobe. The cell fires when its input neurons are all firing. Since each perception lobe neuron represents some feature extracted from the environment (including "I am hungry" and "I am looking at food"), a neuron fires whenever some combination of things is true about the creature's immediate circumstances.

Other possible meanings include "I am hungry and I see food" or "There is a creature and it is a Grendel and it is approaching".

These connections do not have to make sense - they just have to be true. [1] In this way, creatures can learn maladaptive behaviours as well as beneficial ones.

Decision Support[edit]

Main article: Decision lobe

Each neuron outputs to zero or more of the decision lobe neurons. Each output has a weight, which can be positive or negative and determines whether the concept lobe neuron's firing will increase or decrease that decision's score. Semantically, the weight is how good an idea that decision is when this situation is happening.

Learning[edit]

The concept lobe’s inputs and outputs are where almost all learning takes place. Since a brain can only contain a few hundred concepts at once, the reward and punishment mechanism is designed to make each neuron try assigning itself to random situations until it happens upon one that helps the creature seek reward or avoid punishment. This is implemented technically by strengthening recently-active inputs when reward or punishment happens and weakening them whenever the creature tries to do something but isn't rewarded or punished (called "disappointment”). If a neuron’s input connections get too weak, they are reassigned at random to a new set of neurons in the perception lobe.

Notes from the .gno file[edit]

Chemical receptors CHEM0 responds to Reinforcement chemical and is used to increase den strengths CHEM1 responds to ConASH (Concept nei Atrophy Suppressing Hormone), which suppresses further dendrite atrophy when enough loose dendrites exist to provide for future concepts.

Reinforcement / Atrophy Cells get stronger if they fire when reinforcement is present - this is not really good enough - they should be made SUSCEPTIBLE by firing and stronger when susceptible and reinforcement is present, otherwise, they'll not be strengthened if reward comes after the situation has changed!!!!! NEEDS CODE CHANGE.

Cells get weaker when ConASH chemical is NOT present, i.e. when there are no loose cells available for forming new concepts. Given low initial strengths, large numbers will come loose simultaneously, thus providing a long-term stock without weakening valuable memories unduly. However, given a lack of weak neus, good memories will become weakened until a stock of loose cells is available again. Thus the brain never runs out of learning power.


See also[edit]