Information Technology Reference
In-Depth Information
5.
comp
i
(
t
k
) is the computational resources (memory and processing power) avail-
able currently to agent
i
.
The action set of the agent depends on its identity (
id
i
). If it is a buyer, it has
A
i
=
<
Re
+
and
silent
is no bid and an action that does not impact
on the market. Correspondingly, if it is a seller, its action set is
bid
i
,silent>
where
bid
i
∈
A
i
=
<ask
i
,silent >
where
ask
i
Re
+
. It should be noted that in the CDA,
SA
i
will only be singletons
(i.e. an agent can only take a single action at a time). The state transfer function
T
CDA
is the rules for acceptance and rejection of bids and asks as well as the clearing rules
(see below). The standard CDA is not influenced by external signals (i.e. the transfer
function
T
CDA
has no
ext
1
,...,ext
n
arguments
7
) and the market changes each time
an agent submits a bid or an ask and thus simultaneous bidding does not occur. Thus
p
M
(
t
k
+1
)=
T
CDA
(
p
M
(
t
k
)
,H
(
p
M
(
t
k−
1
))
,SA
i
) whereby
T
(
.
) is defined by the fol-
lowing rules:
∈
-
if
SA
i
=
bid
i
,then
•
if
bid
i
<bid
(
t
k
) then
bid
i
is rejected and
p
M
(
t
k
+1
)=
p
M
(
t
k
).
•
if
bid
(
t
)
<bid
i
<ask
(
t
k
) then
bid
(
t
k
+1
)=
bid
i
and all other market vari-
ables remain unchanged.
if
ask
(
t
)
<bid
i
,then
price
(
t
k
+1
)=
cr
(
ask
(
t
k
)+
bid
i
) (where
cr
(
.
) is a
clearing rule stating the transaction price at which the clearing should occur)
8
,
bid
(
t
k
+1
)=0and
ask
(
t
k
+1
)=
max
ask
(where
max
ask
is the maximum ask
an agent can submit in the CDA)
-
if
SA
i
=
ask
i
, it follows the same intuition as above.
-
if
SA
i
=
silent
•
t
k
> inactivity
limit
or
t
k
+1
=
deadline
,
then the auction ends.
inactivity
limit
is a pre-defined period of inactivity whereby
no bid or ask is submitted, and
deadline
, the preset time when the market closes.
∀
i
∈I
and
t
k
+1
−
Furthermore, an agent's state will also change, conditional on whether its bid or ask
is accepted in the market. If an agent's bid
bid
i
results in a transaction,
n
i
(
t
k
+1
)=
n
i
(
t
k
)
.
If an agent's bid is unsuccessful, then the MS relays this private information to the
agent. The agent's visibility is restricted to only bids and asks being submitted in the
market (with the agent that submitted a bid or an ask, not disclosed) and successful
transactions. This information is publicly available in the MS. Based on the information
that describes the market conditions, the agent strategises to submit a competitive offer
to buy or sell. Given this background, we now analyse a selection of the most popular
strategies for the CDA, from the perspective of the IKB model. We provide a summary
of the analysis in table 1.
−
1,
budget
i
(
t
k
+1
)=
budget
i
t
k
−
price
(
t
k
+1
) and
v
i
=
{
v
2
,i
,...,v
n
i
(
t
k
)
,i
}
- The Zero-Intelligence (ZI) Strategy [10]:
The ZI has a random behaviour: it is
non-predictive and does not use the history of market information. It effectively
ignores the market state (MS) and considers only its limit price,
v
n
i
(
t
k
)
,i
(its private
information state in the IL) when submitting a bid or an ask. The KL does not
compute any knowledge and simply forwards
v
n
i
(
t
k
)
,i
from the IL to the BL.
7
Thus, a CDA strategy does not consider external information.
8
This varies according to the CDA; examples include the midway value or
ask
(
t
k
).
