Challenge¶
Depending on classification, many no of actions are possible.
$$ 2^n \text{percepts} \rightarrow 2^m \text{actions} $$Why $2^n$? Given a set of percepts, there would be $2^n$ combinations possible. Each percept is either selected or not, so this results in a binomial theorem situation.
Suppose we have $n=3$ percepts, then there are $2^3$ combinations possible. Each combination's digit is indication if a percept is selected or not.
Let X indicate random variable if a percept is selected(head) or not(tails), so $X = H,T$. Let $n$ indicate no of percepts (at each selection layer, we select one percept or not select). Then the tree could be as below. The digit position could indicate the percept number.
from graphviz import Digraph
from coinflipviz import draw_graph, get_combinations
n_flips = 3 # we have 3 percepts
g = Digraph()
g = draw_graph(g, n_flips)
g
combi_df = get_combinations(n_flips)
combi_df
So 8 combinations in each of which, a percept might be selected or not. For eg, $THH$ indicates 1st percept not selected and rest two selected.
Imagine in above hierarchy, once we establish Eagle and Bluebird as birds, then a characterisitic like Size could help us differentiate between Eagle and Bluebird, so size could be a new node downwards. In below example, if all are yes in Eager, Bluebird, Penguin, then base class Bird has that charactersitic as Yes.
May not work all situations
Types of Concepts¶
Axiomatic Concepts¶
Formal set of necessary and sufficient conditions. Easy to program. Easy to communicate.
Eg: Circle; all points in a plane that are equidistant from a single point.
Prototype Concepts¶
Base concepts defined by typical example with overridable propoerties
Eg: Chair. Note base class has default values, and children overrides some of the notions..
Qulia¶
Much lesser formal than Exemplar - raw sensation from senses (eg, bitterness)
