Economic Silo Guide

What is the Economic Silo?

The Economic Silo is the resource allocation controller in the Liquid Conglomerate architecture. It manages computation budgets, energy economics, trade between individuals, and creates efficiency pressure so that networks must be cost-effective - not just fit.

Think of the Economic Silo as a central bank and marketplace combined. It allocates computational budgets, tracks energy consumption, facilitates trade between individuals, collects taxes, and redistributes wealth. Without it, all individuals receive equal resources regardless of promise, leading to massive waste.

The Economic Silo solves two fundamental problems in neuroevolution:

1. Cost Awareness: Training should respect real-world budget constraints
2. Efficiency Pressure: Better fitness/cost ratio should be rewarded

Architecture Overview

The Economic Silo operates as a three-level hierarchical controller:

Key Principle: Efficiency Matters

The Economic Silo operates on the principle that fitness alone isn't enough:

• Cost-effective solutions are better than expensive ones
• Budget pressure drives innovation in efficiency
• Trade enables specialization (networks can share capabilities)
• Redistribution prevents wealth concentration

How It Works

Sensors (Inputs)

The Economic Silo observes 12 sensors describing economic dynamics:

Actuators (Outputs)

The Economic Silo controls 10 parameters governing economic behavior:

Budget Flow & Economic Cycle

The economic cycle operates as follows:

1. Budget Allocation

%% Each individual receives budget based on strategy
{ok, Budget} = economic_silo:allocate_budget(EconPid, IndividualId),
%% Strategy 0.0 = equal allocation
%% Strategy 1.0 = fitness-proportional allocation

2. Energy Consumption

%% Charge energy during evaluation
ok = economic_silo:charge_energy(EconPid, IndividualId, EvaluationCost),
%% Tracks cost_incurred for efficiency calculation

3. Trade Execution

%% Individuals can trade resources
ok = economic_silo:execute_trade(EconPid, BuyerId, SellerId, Amount),
%% Enables specialization and comparative advantage

4. Taxation

%% Collect taxes on energy income
ok = economic_silo:collect_taxes(EconPid),
%% Funds redistribution pool

5. Redistribution

%% Redistribute from wealthy to poor
ok = economic_silo:redistribute_wealth(EconPid),
%% Reduces inequality, prevents runaway concentration

The Control Loop

1. Per Operation: Allocate budgets, charge energy, execute trades
2. Per Generation: Collect taxes, redistribute wealth, update prices
3. Per Update Cycle: Collect sensors, adjust actuators via TWEANN
4. Emit Events: Publish economic events to event bus

Integration with the Neuroevolution Engine

Wiring Diagram

The Economic Silo integrates with evaluation and other silos:

Data Sources:

• evaluation_engine - Fitness per cost, resource usage
• morphological_silo - Network complexity costs
• temporal_silo - Evaluation time (affects cost)

Data Consumers:

• evaluation_engine - Budget allocations, efficiency requirements
• temporal_silo - Budget constraints on evaluation time
• morphological_silo - Efficiency requirements for network size
• neuroevolution_events - Event bus for monitoring

Cross-Silo Interactions

The Economic Silo exchanges signals with other silos:

Signals Sent: | Signal | To | Description | |--------|-----|-------------| | economic_pressure | Task | High pressure = need simpler solutions | | budget_available | Temporal | Budget constrains evaluation time | | efficiency_requirement | Morphological | Efficiency targets for network size |

Signals Received: | Signal | From | Effect | |--------|------|--------| | resource_level | Ecological | Resources affect available budget | | complexity_cost | Morphological | Network size affects energy cost | | episode_efficiency | Temporal | Efficiency information for pricing |

Engine Integration Points

%% Start Economic Silo
{ok, _} = economic_silo:start_link(#economic_config{
    enabled = true,
    initial_energy = 100.0,
    initial_budget = 1.0,
    enable_trade = true,
    enable_taxation = true,
    emit_events = true
}),

%% Before evaluation: allocate budget
{ok, Budget} = economic_silo:allocate_budget(EconPid, IndividualId),

%% During evaluation: track costs
economic_silo:charge_energy(EconPid, IndividualId, StepCost),

%% End of generation: economic cycle
economic_silo:collect_taxes(EconPid),
economic_silo:redistribute_wealth(EconPid).

Training Velocity Impact

The Economic Silo provides significant cost optimization by creating selection pressure for efficient solutions.

Practical Examples

Example 1: Budget-Proportional Allocation

%% Scenario: Some individuals are more promising
%% allocation_strategy = 0.8 (fitness-proportional)

%% High-fitness individual gets more budget:
%% fitness=0.9 -> budget=1.8x average

%% Low-fitness individual gets less:
%% fitness=0.1 -> budget=0.2x average

%% Result: Promising individuals get more evaluation time

Example 2: Trade Enables Specialization

%% Scenario: Agent A is good at exploration, Agent B at exploitation
%% They can trade capabilities

ok = economic_silo:execute_trade(EconPid, AgentA, AgentB, 50.0),

%% Trade event emitted:
{trade_completed, #{
    buyer_id => AgentA,
    seller_id => AgentB,
    amount => 50.0,
    price => 500.0
}}

%% Result: Both benefit from comparative advantage

Example 3: Wealth Redistribution Reduces Inequality

%% Scenario: High Gini coefficient (0.8 = very unequal)
%% wealth_redistribution_rate = 0.1

%% Before: Agent A wealth=1000, Agent B wealth=10
%% After: Agent A wealth=901, Agent B wealth=109

%% Event:
{wealth_redistributed, #{
    rate => 0.1,
    total_transferred => 99.0
}}

%% Result: Poor agents can still compete

Tuning Guide

Key Parameters

Common Pitfalls

1. Budget too low: Good individuals don't have time to show potential

   • Symptom: All fitnesses converge to same low value
   • Fix: Increase budget_per_individual to 2.0+

2. No trade: Missing specialization benefits

   • Symptom: All individuals evolve same strategies
   • Fix: Increase trade_incentive to 0.3+

3. High inequality: Poor individuals can't compete

   • Symptom: Same few individuals always win
   • Fix: Increase wealth_redistribution_rate to 0.2+

4. No cost pressure: Bloated, inefficient networks

   • Symptom: Networks grow without bound
   • Fix: Lower budget_per_individual, increase debt_penalty

Debugging Tips

%% Get individual account
Account = economic_silo:get_account(EconPid, IndividualId),
io:format("Energy: ~.2f~n", [Account#individual_account.energy]),
io:format("Debt: ~.2f~n", [Account#individual_account.debt]),
io:format("Efficiency: ~.3f~n", [
    Account#individual_account.fitness_earned /
    max(1.0, Account#individual_account.cost_incurred)
]),

%% Get overall economic state
State = economic_silo:get_state(EconPid),
io:format("Market price: ~.2f~n", [State#economic_state.market_price]),
io:format("Total tax collected: ~.2f~n", [State#economic_state.total_tax_collected]),
io:format("Bankruptcies: ~p~n", [State#economic_state.bankruptcies]).

Events Reference

The Economic Silo emits events on significant actions:

Example Event Payload:

{trade_completed, #{
    silo => economic,
    timestamp => 1703318400000,
    generation => 42,
    payload => #{
        trade_id => <<"trade_12345">>,
        buyer_id => <<"agent_a">>,
        seller_id => <<"agent_b">>,
        quantity => 50.0,
        price => 500.0
    }
}}

L0 Hyperparameters (L1-Tuned)

L1 Hyperparameters (L2-Tuned)

Configuration Examples

Cost-Constrained Cloud Training

#economic_config{
    enabled = true,
    initial_energy = 50.0,         % Limited
    initial_budget = 0.5,          % Tight
    enable_trade = true,
    enable_taxation = true,
    emit_events = true
}.

%% With actuators:
%% budget_per_individual = 0.3
%% debt_penalty = 0.15 (aggressive)
%% efficiency emphasized

Research/Exploration Mode

#economic_config{
    enabled = true,
    initial_energy = 1000.0,       % Abundant
    initial_budget = 5.0,          % Generous
    enable_trade = false,          % Independent evolution
    enable_taxation = false,       % No redistribution
    emit_events = true
}.

%% Let evolution explore freely without cost pressure

Multi-Agent Marketplace

#economic_config{
    enabled = true,
    initial_energy = 200.0,
    initial_budget = 2.0,
    enable_trade = true,            % Critical
    enable_taxation = true,
    emit_events = true
}.

%% With actuators:
%% trade_incentive = 0.4 (high)
%% wealth_redistribution_rate = 0.15
%% Enable specialization through trade

Source Code Reference

Further Reading

• Liquid Conglomerate Overview - Full LC architecture
• Temporal Silo - Time and episode management
• Task Silo - Hyperparameter adaptation
• Morphological Silo - Network structure optimization
• Meta-Controller Guide - L2 strategic layer

References

Evolutionary Economics

• Potts, J. (2000). "The New Evolutionary Microeconomics: Complexity, Competence and Adaptive Behaviour." Edward Elgar Publishing.

Resource Allocation in Distributed Systems

• Buyya, R., et al. (2002). "Economic Models for Resource Management and Scheduling in Grid Computing." Concurrency and Computation: Practice and Experience.

Multi-Agent Economies

• Varian, H.R. (2014). "Intermediate Microeconomics: A Modern Approach." W.W. Norton.
• Arthur, W.B. (1994). "Increasing Returns and Path Dependence in the Economy." University of Michigan Press.