beta¶
Specifies the degree to which previous experiences (v-values) are considered in the decision function,
in other words, how much weight is placed on the v-values in the decision function.
A low value for beta
means high exploration.
The value of the parameter beta
is used in the decision function. See also mu.
The parameter beta
can be either a single value, in which case the same weight is placed on all v-values, or
specified per element-behavior pair. Wildcards (*) may be used to set several values at once.
Syntax¶
beta = e1->b1: v1, e1->b2: v2, ..., en->bn: vn, default: d
beta = v1
beta = *->*:v1 # Same as above
beta = e1->*:v1
beta = *->b1:v1
where v1,v2,...,vn
and d
are scalar expressions.
Description¶
beta = e1->b1: v1, e1->b2: v2, ..., en->bn: vn, default: d
sets the individual weight on
v(e1->b1) to v1,
v(e1->b2) to v2, …,
v(en->bn) to v2,
and the weight for all other v-values to d.
The specification is independent of the list order:
beta = e1->b1:v1, e1->b2:v2, default:d
is the same as
beta = e1->b2:v2, default:d, e1->b1:v1
.default
need not be specified if all possible combinationselement->behavior
are present in the list. For example,stimulus_elements = e1, e2 behaviors = b1, b2 beta = e1->b1:v11, e1->b2:v12, e2->b1:v21, e2->b2:v22
beta = v1
sets the weight for all v-values to v1.
beta = v1
is the same asbeta = default: v1
.
Dependencies¶
The properties
stimulus_elements
andbehaviors
must be specified beforebeta
.Each stimulus element used in the specification of
beta
must be present in the parameterstimulus_elements
.Each behavior used in the specification of
beta
must be present in the parameterbehaviors
.
Examples¶
@variables x = 1
beta = element1->behavior1: x, element1->behavior2: x+1, default:x+2
sets the weight for v(element1->behavior1) to 1 and for v(element1->behavior1) to 2, and for the remaining possible element-behavior pairs to 3.
beta = element1->behavior1: 0.5, default:0
sets the weight for v(element1->behavior1) to 0.5, and for the remaining possible element-behavior pairs to 3.
beta = 0.1
sets the weight for all v-values to 0.1.