Information Technology Reference
In-Depth Information
Sophisticated Intelligence Traders (SIT). Several kinds of SIT can evolve in ATOM:
i) Cognitive Agents generally have a full artificial intelligence, although it can be
designed to be rather minimal (usual features to develop such agents are memory, infor-
mation analysis processes, expectations, strategies and learning capacities). For exam-
ple, an agent buying at a specific price and sending immediately a ”stop order” to short
her position if the price drops under θ % times the current price, will fall in this category.
Agents using strategic order splitting (see for example [20]) or exploiting sophisticated
strategies (for instance, [21]) can also be considered as Cognitive Agents.
ii) Evolutionary Agents are the ultimate form of SIT; they outperform Cognitive
Agents in terms of complexity since they are able to evolve with their environment.
These agents can also generate new rules or strategies (this can require genetic algo-
rithms for example).
iii) Risk Averse Agents set up portfolio models in which the individual chooses a set
of assets in order to maximise some function of wealth. For each investment possibility
I from the set of alternatives F , the agent will undertake one I opt , that will maximize
resulting wealth W ( I ) or E [ W ( I opt )] = max I∈F E [ W ( I )] . Utility function provides
relative measure of investor's preferences for wealth and the amount of risk they are
willing to undertake in order to maximize their wealth. In ATOM, the agents have a
choice between different utility functions: Constant Absolute Risk Aversion (CARA),
Constant Relative Risk Aversion (CRRA), logarithmic and quadratic.
iv) Mean-variance Agents are investors trading over several order topics, hence refer
to portfolio optimisation aspects. When an investor wants to reoptimize her portfolio,
she chooses and ”ideal” portfolio from a mean-variance efficient frontier [22], that is
based on analysis of internal and external information. The choice of portfolio depends
on the trader's risk aversion. These agents send buy and sell orders in order to get closer
to ”ideal” portfolio. Such population of the agents is heterogeneous due to their initial
cash available, reoptimization (trading) frequency and risk aversion.
We introduce environment-based interactions, where market restricts agents behaviour
and at the same time it evolves in response to agents activities. Thus, the environment
has its own state and rules of changes [23]. Traders submit orders depending on the state
of the order topic or best quotes (environment). These orders may result in a change in
the best prices. The state of the market changes over time. A feedback loop is formed: a
trader submits orders which affect the state of the market which affects the decisions of
the trader on what order to submit. This aspect relies to Interaction Movement Compu-
tation ( MIC ) [24], where the environment defines actions sets of autonomous agents
to achieve their goals. Agents interact with one another in order to achieve either a com-
mon or individual objectives through environment. The agents can also interact through
the common variable of the past price history, but they are not directly affected by the
actions of others. In order to keep agents equality and to avoid the biases in the internal
information access, agents should be informed about topic order changes simultane-
ously. Notification method is realized in ATOM in accordance with Influence Reaction
Model for Simulation (IRM4S)[25], [26]: all orders, as influences, are collected in the
Search WWH ::




Custom Search