Skip to main content

Cortical Areas

Cortical areas are the fundamental building blocks of your genome. Each cortical area is a 3D volume of neurons that processes information, stores patterns, or interfaces with the external world.

What is a Cortical Area?​

A cortical area is a structured group of neurons organized in a 3D grid (measured in voxels). Each voxel can contain one or more neurons. Cortical areas:

  • Process incoming signals from connected areas
  • Generate outputs based on their internal state and connections
  • Learn and adapt through synaptic changes
  • Specialize in different types of processing

Types of Cortical Areas​

Brain Visualizer supports several types of cortical areas, each with specific purposes:

Input Processing Unit (IPU)​

IPU Icon

Purpose: Receive data from external sources (sensors, cameras, etc.)

Characteristics:

  • Color: Dark Gray
  • Interface: Connected to embodiment devices
  • Configuration: Based on device templates (vision, audio, text, etc.)
  • Dimensions: Determined by device type and count

Common Uses:

  • Visual input from cameras
  • Audio input from microphones
  • Sensor data (temperature, distance, etc.)
  • Text or symbolic input
  • Controller input (keyboard, gamepad)

Output Processing Unit (OPU)​

OPU Icon

Purpose: Send data to external actuators (motors, displays, etc.)

Characteristics:

  • Color: Orange
  • Interface: Connected to embodiment devices
  • Configuration: Based on device templates (motor, display, etc.)
  • Dimensions: Determined by device type and count

Common Uses:

  • Motor control (movement, rotation)
  • Display output (text, graphics)
  • Audio output (speech, sounds)
  • Control signals (on/off, analog values)

Memory​

Memory Icon

Purpose: Store and recall patterns for learning and reference

Characteristics:

  • Color: Dark Red
  • Function: Pattern recognition and recall
  • Behavior: Learns associations between inputs
  • Dimensions: User-configurable

Common Uses:

  • Short-term memory
  • Long-term pattern storage
  • Associative recall
  • Context maintenance

Custom (Interconnect)​

Custom Icon

Purpose: Internal processing and transformation

Characteristics:

  • Color: Blue
  • Function: General-purpose neural processing
  • Behavior: Transforms inputs to outputs
  • Dimensions: User-configurable

Common Uses:

  • Feature extraction
  • Pattern transformation
  • Decision making
  • Intermediate processing layers
  • Custom neural algorithms

Core​

Purpose: System-level processing (advanced)

Characteristics:

  • Color: Dark Blue
  • Function: Specialized system operations
  • Behavior: FEAGI internal processing
  • Dimensions: System-defined

Note: Core areas are typically created by FEAGI itself and rarely created manually.

Creating Cortical Areas​

Method 1: Quick Access (IPU/OPU)​

For input and output areas:

  1. Click Inputs or Outputs in the top toolbar
  2. Click the + button
  3. Select a template (e.g., Vision, Motor)
  4. Configure:
    • Device Count: How many instances (e.g., 2 cameras)
    • Unit ID: Unique identifier for the device
    • Location: 3D position in genome
    • Data Type: How data is encoded
  5. Click Add

The new area appears in both Circuit Builder and Brain Monitor.

Method 2: Create from Circuit Builder​

For all types:

  1. Right-click empty space in Circuit Builder
  2. Select Create Cortical Area
  3. Choose type:
    • Click Input for IPU
    • Click Output for OPU
    • Click Interconnect for Custom
    • Click Memory for Memory
  4. Configure properties
  5. Click Add

Method 3: Clone Existing Area​

Duplicate an area with similar settings:

  1. Right-click existing cortical area
  2. Select Clone
  3. Modify name and properties as needed
  4. Click Clone

Configuring IPU/OPU Areas​

Templates​

IPU and OPU areas use templates that define their structure:

Common IPU Templates:

  • Vision: Camera input (2D image data)
  • Audio: Microphone input (frequency bands)
  • Text Input: Character or word data
  • Generic Sensor: Numerical sensor data
  • Controller: Button/joystick input

Common OPU Templates:

  • Motor: Servo or motor control
  • Display: Text or graphics output
  • Audio Output: Sound generation
  • Generic Actuator: Numerical control signals

Device Count​

Specifies how many instances of the device exist:

  • 1 camera = single vision IPU
  • 2 cameras = stereo vision (two separate IPUs or one multi-unit IPU)
  • 4 motors = four separate motor OPUs

Each count creates the appropriate cortical structure.

Unit ID​

Unique identifier connecting the cortical area to physical/virtual hardware:

  • Must match the ID used by your embodiment controller
  • Allows FEAGI to route data to/from correct devices
  • Required for IPU/OPU areas

Data Type Configuration​

Determines how data is encoded:

For IPU (Input):

  • Signed Percentage: Normalized -100% to +100%
  • Unsigned Percentage: Normalized 0% to 100%
  • Absolute Values: Raw numeric values
  • Binary: On/off states

For OPU (Output):

  • Similar encoding options
  • Must match what the embodiment expects

Frame Handling​

For vision IPU areas:

  • Single Frame: Process one frame at a time
  • Frame Stack: Stack multiple frames for temporal awareness
  • Differential: Process frame-to-frame changes

Positioning​

How multi-unit data is arranged spatially:

  • Stack XYZ: Stack along specific axis
  • Grid: Arrange in 2D grid
  • Linear: Arrange in line

Configuring Custom/Memory Areas​

Dimensions​

Set the 3D size of the cortical area:

  • X, Y, Z: Dimensions in voxels
  • Total neurons: X × Y × Z × neurons per voxel
  • Larger areas = more neurons = more processing capacity

Example:

  • 10 × 10 × 10 = 1,000 voxels
  • If 1 neuron per voxel = 1,000 neurons
  • If 10 neurons per voxel = 10,000 neurons

Position​

3D location in the genome:

  • X, Y, Z: Coordinates in 3D space
  • Position is organizational (doesn't affect processing)
  • Group related areas nearby for clarity

Neurons Per Voxel​

How many neurons exist in each voxel:

  • 1: Single neuron per voxel (typical)
  • Higher values: Multiple neurons per voxel (advanced)
  • Affects total neuron count and processing

Connectivity Rule​

Default connectivity rule for connections:

  • Pattern: Defines connection structure
  • Parameters: Shape and density
  • Can be overridden per connection

See [Connectivity Rules](connectivity rules.md) for details.

Viewing and Editing Properties​

Cortical Area Details Window​

Right-click area → Details opens the comprehensive Cortical Area Details window.

This window provides complete control over:

  • Basic properties (name, dimensions, position)
  • Neuron firing parameters (threshold, leak, refractory period)
  • Memory parameters (lifespan, consolidation)
  • Post-synaptic potential settings (connection strength)
  • Neuron coding (for IPU/OPU areas)
  • Monitoring and visualization settings
  • Connection management (afferents, efferents, recursive)
  • Delete and reset operations

For complete documentation of all features and parameters, see:

Quick Access:

  • Right-click cortical area → Details
  • Double-click cortical area node
  • Quick Menu → Details

Organizing Cortical Areas​

Naming Conventions​

Use clear, descriptive names:

Good:

  • "Vision_Left_Camera"
  • "Motor_Front_Left_Wheel"
  • "Memory_Visual_Patterns"
  • "Custom_Edge_Detection"

Avoid:

  • "CA_001"
  • "Untitled"
  • "Test"

Grouping with Regions​

Organize related areas into brain circuits:

  1. Select cortical areas to group
  2. Right-click → Create Region
  3. Name the region descriptively
  4. Areas move into the new region

See Brain Circuits for more details.

Spatial Organization​

In Brain Monitor (3D), position areas logically:

  • Inputs: One side or top
  • Processing: Middle layers
  • Memory: Central or dedicated zone
  • Outputs: Opposite side from inputs

Organized layout aids understanding and debugging.

Connecting Cortical Areas​

Cortical areas become functional when connected:

Creating Connections​

  1. In Circuit Builder: Drag from output port to input port
  2. Quick Connect: Right-click → Quick Connect → choose destination
  3. Mapping Editor: Specify connectivity rule and parameters

See Mapping Connections for complete guide.

Connection Types​

  • Feedforward: Input → Processing → Output (typical)
  • Feedback: Higher layer → Lower layer (modulation)
  • Lateral: Same-level areas (integration)
  • Recursive: Area to itself (temporal)

Best Practices​

  1. Start Simple: Connect inputs to outputs with one processing layer
  2. Test Incrementally: Add connections and test behavior
  3. Avoid Over-Connection: Not everything needs to connect to everything
  4. Use Appropriate Connectivity Rules: Match connection patterns to function
  5. Document: Name connections and regions to clarify intent

Common Operations​

Renaming​

  1. Right-click area → Details
  2. Edit the name field
  3. Press Enter or click away to save

Moving (2D Position)​

In Circuit Builder:

  • Drag the node to new position
  • Position saves automatically after brief delay

Via Menu:

  • Right-click → Relocate 2D
  • Enter exact X, Y coordinates

Moving (3D Position)​

Via Menu:

  • Right-click → Move 3D
  • Drag colored arrows in 3D view
  • X=Red, Y=Green, Z=Blue

Via Properties:

  • Right-click → Details
  • Edit position X, Y, Z values
  • Click Apply

Resizing​

For Custom and Memory areas:

Via 3D Gizmo:

  • Right-click → Resize 3D
  • Drag corner/edge handles in 3D view

Via Properties:

  • Right-click → Details
  • Edit dimensions X, Y, Z
  • Click Apply

Note: IPU/OPU dimensions are determined by templates and cannot be directly resized.

Cloning​

Create a copy with similar settings:

  1. Right-click → Clone
  2. Modify name (required)
  3. Adjust position/dimensions if needed
  4. Click Clone

Connections are NOT cloned (area starts unconnected).

Resetting​

Clear all neural state (neuron values, learning):

  1. Right-click → Reset
  2. Confirm the reset
  3. Area returns to initial state

Useful for:

  • Starting fresh after testing
  • Clearing corrupted state
  • Beginning new training

Deleting​

Remove a cortical area permanently:

  1. Right-click → Delete
  2. Review confirmation (shows affected connections)
  3. Confirm deletion

Warning: This removes all connections to/from the area. Cannot be undone.

Monitoring Cortical Area Activity​

In Brain Monitor​

Active cortical areas light up:

  • Bright spots: Highly active neurons
  • Patterns: Spatial activity distribution
  • Changes: Real-time updates

Hover over area to see its connections.

Activity Indicators​

  • Color Intensity: Firing rate
  • Spatial Patterns: Which voxels are active
  • Temporal Patterns: How activity changes over time

Debugging​

If area isn't showing expected activity:

  1. Check Connections: Verify inputs are connected
  2. Check Input Activity: Ensure upstream areas are active
  3. Check Mappings: Verify connectivity rules are correct
  4. Check Data Flow: Trace from inputs through processing
  5. Check Configuration: Verify area settings are correct

Performance Considerations​

Neuron Count Limits​

Your genome has a maximum neuron count:

  • Check current vs max in top toolbar
  • Creating large areas consumes budget
  • Balance size vs. quantity

Planning Capacity​

Before creating areas:

  1. Estimate neurons needed per area
  2. Calculate total neurons
  3. Ensure within genome limits
  4. Adjust dimensions if needed

Optimization Tips​

  • Start small: Create minimal areas, expand if needed
  • Use templates efficiently: IPU/OPU sizes match data dimensions
  • Memory areas: Size based on pattern storage needs
  • Custom areas: Optimize for actual processing requirements

Troubleshooting​

"Can't create IPU/OPU"

  • Ensure template is selected
  • Verify unit ID is unique
  • Check neuron count limit isn't exceeded

"Area not visible"

  • Use Fit All in Circuit Builder
  • Focus using top toolbar dropdowns
  • Check area is in expected region

"No activity showing"

  • Verify area has input connections
  • Check upstream areas are active
  • Ensure burst rate > 0 Hz
  • Verify FEAGI is processing

"Can't resize area"

  • IPU/OPU dimensions are template-defined
  • Check if area type allows resizing
  • Use Properties window for precise control

"Neuron count exceeded"

  • Reduce area dimensions
  • Delete unused areas
  • Increase genome neuron limit (if possible)

Back to Overview