# Black Holes and Revelations
### Using software group

#### How do we create a raster of uncertainty from the probabilities of interactions and occurrence?

* The probability of interaction around 0.5 if the most uncerain - we should somehow point users to go sample in the locations with this probability.
* We are trying to find the locations with the highest probability of co-occurrence and the with the probability of interaction closer to 0.5.
* Similar environments should result in the same probability of interactions.
* We want to exclude locations with super high or super low probability of interactions before we proceed.
> *:warning: Our uncertainty about an interaction in a given location is influenced by the probability of interaction in the neighbour cells.*

* How do we measure the distance to the most uncertain interaction and most likely co-occurrence
* Euclidean distance? But this gives weird situations depending on which side the distance comes from
* Weights would be a better alternative than the Euclidean distance
* If we want to use the full probability distribution for the interaction in the metaweb, we will need to use a double integral
* How do we optimize this distance
* Minimize something like $W_1 * (P(A \rightarrow B) - 0.5)^2 + W_2 * (P(A)P(B) - 0.5)^2$

* we've got a _curiosity score_ which shows us where we should go sampling when its value is high.

## Code notes
* Network table has less lines than expected because forbidden links were excluded

* coocurrence vector is:
* each column is a site
* each row is a pair of species
#### TODO
* Add 0s to diagonal of cooccurrence layer

## GH notes
* cooccurrence branch - devs finalized cooccurrence matrix, excluding forbidden links

## Glossary
Please add terms / edit definitions as you see fit

---WIP

**INTERACTION**: species A eats species B (consumer-resource interaction)
**NETWORK**: Structured set of interactions; set of vertices $V$, function $f:(V * Y) -> {0,1}$ / web showing all possible species interactions present.
**GLOBAL METAWEB**: Network of all potential interactions within total species pool.
**LOCAL METAWEB**: subset of potential species interactions from the global metaweb for species present at a specific site.
**PROBABILITY**: Degree of belief; measurement of the frequency of observations in a 🌟 universe of possibilities 🌟.
**SAMPLING**: point/area in which to collect data on species occurrence and interactions.
**UNCERTAINTY**: Degree of belief that two species interact and co-occur at a spatial location
**ENTROPY**: Amount of information provided by _new_ data.
**SPECIES DISTRIBUTION MODEL (SDM)**: uses environmental data to statistically infer the spatial distribution of a species population
**PROBABILITY OF CO-OCCURANCE** $P(A)P(B)$: likelihood that species A and species B are present at a sampling location
**PROBABILITY OF INTERACTION** $P(A→B)$ : likelihood that species A eats species B
**CURIOSITY**: value given to a sampling location, based on the degree of proposed sampling effort and the rarity of interactions being searched for
**API**: Any function you call but didn't write (Application Programatic Interface).
**Potential Interaction**: Any interaction with a probability of interaction >0.
**Realised Interaction**: Obeserved empirical interaction.
**Bias**: Systematically misestimating.