Skip to main content

Mapping Connections

Mappings are the neural connections between cortical areas. They define how information flows through your genome, using connectivity rules to specify the exact wiring patterns.

What is a Mapping?​

A mapping is a connection from one cortical area (source) to another (destination) that:

  • Uses a connectivity rule to define connection structure
  • Creates synapses between neurons
  • Allows information flow during processing
  • Can learn and adapt based on activity

Without mappings, cortical areas are isolated - mappings make them into an integrated neural network.

Mapping Components​

Each mapping consists of:

Source and Destination​

  • Source (Afferent): The cortical area sending signals
  • Destination (Efferent): The cortical area receiving signals
  • Direction matters (A→B is different from B→A)

Connectivity Rule​

  • Defines which neurons connect to which
  • Determines connection density and pattern
  • Can be shared across multiple mappings

See Connectivity Rules for details.

Parameters​

  • PSP: The mapping's PSP multiplier; becomes the initial weight of each synapse
  • Inhibitory: Synapses subtract instead of add
  • Delay: Synaptic delay in whole bursts (minimum 1)
  • Plasticity mode: Off, STDP or R-STDP
  • Window / Constant / LTP x / LTD x: Learning parameters (see below)
  • Eligibility Decay: Trace decay in bursts (R-STDP only)

How learning and PSP interact​

  • A synapse delivers weight x PSP x sign; PSP is the source area's value, stamped on its outgoing synapses at creation
  • LTP/LTD change weight only, never PSP (weight is floored at 0; punishment never raises it)
  • PSP scales every learned change: a synapse is silent if either weight or PSP is 0
  • Changing an area's PSP affects only synapses created afterwards
  • The weight cap (max_weight), learning rate (eta) and freeze are set via API/MCP, not this editor

Creating Mappings​

Method 1: Drag in Circuit Builder​

Create Mapping Icon

The most intuitive method:

  1. Open Circuit Builder
  2. Click and hold on output port (right side) of source area
  3. Drag to input port (left side) of destination area
  4. Release to open Mapping Editor
  5. Configure connectivity rule and parameters
  6. Click Create or Apply

Method 2: Quick Connect​

Quick Connect Icon

From context menu:

  1. Right-click source cortical area
  2. Select Quick Connect
  3. Click the destination in the current brain monitor, or click Edit to open Cortical Area Explorer and pick destinations from any circuit
  4. Choose a connectivity rule. An interconnect area mapped into a memory area has one qualified rule, Episodic Memory, so Quick Connect shows that rule only and skips the rule picker.
  5. Optionally enable Advanced, above Establish, to set PSP multiplier, inhibitory, synaptic delay, and STDP plasticity. Episodic Memory and classifier destinations do not show Advanced. Reward-modulated plasticity stays in the mapping editor.
  6. Establish the connection

Method 3: Create Multiple​

Connect one area to many:

  1. Right-click source area
  2. Select Quick Connect
  3. Click Edit on the destination step and check multiple destinations in Cortical Area Explorer (or Ctrl-click visible areas)
  4. Set common connectivity rule
  5. Create all mappings at once. If Advanced is enabled, those parameters apply to every destination.

Method 4: Bidirectional​

Create two-way connection:

  1. Create mapping A → B
  2. Create mapping B → A
  3. Or use bidirectional option if available
  4. Each direction can use different connectivity rule

Opening a Mapping from Brain Monitor​

  1. Hover a cortical area so its incoming and outgoing mapping lines appear
  2. Hold Shift. The lines stay up after the pointer leaves the area
  3. Click the button in the middle of a line. The button shows how many mappings connect those two areas
  4. The mapping editor opens for that pair, the same window Circuit Builder opens from its connection button

Release Shift and the lines hide, unless the pointer is still over the area.

Mapping Editor​

The Mapping Editor is your detailed interface for configuring connections.

Editor Sections​

Header:

  • Source area name and type
  • Destination area name and type
  • Direction indicator (→)

Connectivity Rule Selection:

  • Dropdown of existing connectivity rules
  • Preview of selected connectivity rule
  • Button to create new connectivity rule

Parameters:

  • Postsynaptic current
  • Plasticity settings
  • Learning parameters
  • Weight configuration

Statistics:

  • Estimated synapse count
  • Connection density
  • Memory usage estimate

Actions:

  • Apply/Create button
  • Cancel button
  • Advanced options

Selecting Connectivity Rule​

Use Existing:

  1. Click connectivity rule dropdown
  2. Browse available connectivity rules
  3. Select appropriate pattern
  4. Preview updates

Create New:

  1. Click + New Connectivity Rule
  2. Configure connectivity rule type and parameters
  3. Name the connectivity rule
  4. It becomes available in the list

Configuring Parameters​

PSP (multiplier):

  • Becomes the initial weight of each synapse (not the PSP itself)
  • Must be 0 or higher; use the Inhibitory toggle for negative effect

Inhibitory / Delay:

  • Inhibitory: drive is multiplied by -1
  • Delay: bursts between source firing and delivery (minimum 1)

Plasticity mode:

  • Off: Weights stay fixed
  • STDP: Eligibility trace is rebuilt each burst; weight changes by eta x trace
  • R-STDP: Trace persists and decays; weight changes by eta x (reward - punishment) x trace

Window / Constant / LTP x / LTD x:

  • If source and destination both fired within the last Window bursts: trace += Constant x LTP x
  • If only one fired: trace -= Constant x LTD x (skipped on associative-memory mappings)
  • Each step (Constant x LTP x or Constant x LTD x) is clamped to 0-255; LTP x = 0 disables potentiation

Eligibility Decay:

  • R-STDP only: trace is multiplied by exp(-1/Decay) each burst

Creating the Mapping​

Once configured:

  1. Review parameters
  2. Check synapse count estimate
  3. Click Create or Apply
  4. Mapping is created in FEAGI
  5. Connection line appears in Circuit Builder
  6. Connection is active in Brain Monitor

Viewing Existing Mappings​

In Circuit Builder​

Mappings appear as lines between cortical areas:

  • Green: excitatory
  • Red: inhibitory
  • Yellow: episodic (episodic_memory and episodic_scan)
  • Dashed: plastic
  • Arrows: Show direction of information flow

Click a line to view/edit its properties.

In Brain Monitor​

Outside a text field, press the backtick key (`) to show every mapping at once. Press it again, press Escape, or use Stop Inspector to hide them. The same control is under Inspectors on the main top bar and the Brain Monitor tab bar.

Hover over a cortical area to see:

  • Outgoing connections (from this area)
  • Incoming connections (to this area)
  • Active data flow (if neuron activity present)

Connection Lines:

  • Highlight when hovering
  • Show direction with arrows or flow
  • Intensity indicates activity level

Via Context Menu​

Right-click cortical area:

  1. Select Details
  2. Navigate to Connections tab
  3. View lists:
    • Afferent (incoming): Areas sending to this one
    • Efferent (outgoing): Areas this one sends to
    • Recursive: Connections to itself
  4. Click connection to edit

Editing Mappings​

Modifying Parameters​

  1. Click connection line in Circuit Builder OR
  2. Right-click area → Details → Connections tab
  3. Click mapping to edit
  4. Mapping Editor opens
  5. Modify parameters
  6. Click Apply

Changes take effect immediately.

Changing Connectivity Rule​

  1. Open mapping in Mapping Editor
  2. Select different connectivity rule from dropdown
  3. Connection pattern updates
  4. Synapse count may change
  5. Click Apply

Caution: Major connectivity rule changes may dramatically alter behavior.

Copying Mappings​

To replicate a mapping configuration:

  1. Note the connectivity rule and parameters
  2. Create new mapping
  3. Apply same settings
  4. Or use batch creation for efficiency

Deleting Mappings​

Single Mapping​

  1. Click connection line in Circuit Builder
  2. Press Delete key OR
  3. Right-click line → Delete
  4. Confirm deletion

OR

  1. Right-click cortical area → Details
  2. Go to Connections tab
  3. Find mapping in list
  4. Click Delete button
  5. Confirm

Multiple Mappings​

To delete all connections:

  1. Select cortical area
  2. Right-click → Reset (clears state but keeps connections)
  3. OR manually delete each mapping

On Area Deletion​

When deleting a cortical area:

  • All mappings to/from it are automatically deleted
  • You'll see a confirmation showing affected mappings
  • Ctrl+Z (Cmd+Z on macOS) puts the area and its mappings back, but not learned synapse weights

Connection Patterns​

Feedforward (Input → Processing → Output)​

Typical flow:

IPU (Vision) → Custom (Edge Detection) → Custom (Object Recognition) → OPU (Action)

Best Practices:

  • Use appropriate connectivity rules for each stage
  • All-to-All for mixing and integration
  • One-to-One for spatial preservation

Feedback (Higher → Lower)​

Top-down modulation:

Memory (Context) → Custom (Visual Processing)

Use Cases:

  • Attention and gating
  • Expectations and predictions
  • Context-dependent processing

Lateral (Within Layer)​

Same-level integration:

Custom (Left Visual Field) ↔ Custom (Right Visual Field)

Use Cases:

  • Information sharing
  • Lateral inhibition
  • Continuous representations

Recursive (Area to Itself)​

Temporal processing:

Memory (Short Term) → Memory (Short Term)

Use Cases:

  • Holding state over time
  • Temporal integration
  • Working memory

Multiple Mappings​

You can create multiple mappings between the same two areas:

Why Multiple Mappings?

  • Different connectivity rules for different purposes
  • Excitatory and inhibitory connections
  • Different learning rates
  • Parallel pathways

Managing Multiple:

  • Each mapping is independent
  • Synaptic effects combine
  • Can have different parameters
  • Visualized as separate lines (or combined)

Mapping Direction​

Direction is critical:

A → B:

  • Neurons in A send to neurons in B
  • Activity in A influences B
  • Information flows A to B

B → A:

  • Separate mapping (does NOT exist automatically)
  • Must be created explicitly
  • Can have different connectivity rule and parameters

Bidirectional:

  • Create both A → B and B → A
  • Allows information to flow both ways
  • Each direction is independent

Connection Debugging​

No Activity Flowing​

If expected activity doesn't appear:

  1. Verify Connection Exists: Check Circuit Builder for line
  2. Check Source Activity: Ensure source area is active
  3. Check Connectivity Rule: Verify appropriate pattern
  4. Check PSC: Ensure postsynaptic current is not too weak
  5. Check Path: Trace from inputs through all connections

Unexpected Behavior​

If area behaves unexpectedly:

  1. Review Connections: Check all afferent mappings
  2. Check Connectivity Rules: Verify connection patterns
  3. Check Parameters: Look for unusual PSC or learning rates
  4. Check for Loops: Recursive or feedback loops can cause issues
  5. Isolate: Temporarily remove connections to identify culprit

Performance Issues​

If too many connections slow the system:

  1. Reduce Density: Use sparser connectivity rules
  2. Limit Connections: Set max connections per neuron
  3. Remove Unused: Delete unnecessary mappings
  4. Optimize Connectivity Rules: Use efficient patterns

Advanced Mapping Techniques​

Gating​

Use one area to control another:

Context Area (modulation) → Processing Area (modulated)

Low PSC or conditional connections act as gates.

Ensembles​

Connect multiple areas to one target for voting/integration:

Evidence A → Decision
Evidence B → Decision
Evidence C → Decision

Hierarchical​

Build processing hierarchies:

Low-Level Features → Mid-Level Features → High-Level Features

Attentional​

Use feedback to enhance relevant processing:

Task Area (what to attend) → Sensory Area (what's attended)

Synapse Budget​

Mappings consume synapses:

  • Check current vs max synapse count in top toolbar
  • All-to-All mappings use the most
  • Sparse patterns conserve synapses
  • Plan ahead for large genomes

Estimating Synapse Count:

  • Small areas, All-to-All: (M × N) synapses
  • Large areas, sparse: Much fewer
  • Mapping Editor shows estimate

Best Practices​

Start Simple​

  1. Create minimal connections first
  2. Test functionality
  3. Add complexity incrementally
  4. Verify each addition

Use Appropriate Connectivity Rules​

  • One-to-One: Spatial tasks
  • All-to-All: Integration and mixing
  • Lateral: Context and smoothing
  • Sparse: Efficiency

Name and Document​

  • Use meaningful cortical area names
  • Document connection purposes
  • Note unusual parameter choices
  • Explain design decisions

Test Incrementally​

  • Create one mapping at a time
  • Verify expected behavior
  • Check activity flow
  • Then add next connection

Monitor Performance​

  • Watch synapse count
  • Check processing speed
  • Optimize as needed
  • Remove unnecessary connections

Common Workflows​

Creating a Simple Circuit​

  1. Create Input (IPU) cortical area
  2. Create Processing (Custom) area
  3. Create Output (OPU) area
  4. Connect IPU → Custom with All-to-All
  5. Connect Custom → OPU with All-to-All
  6. Test with data

Building a Vision Pipeline​

  1. Vision IPU (camera input)
  2. Edge Detection Custom (lateral connectivity rule)
  3. Feature Extraction Custom (All-to-All)
  4. Object Recognition Custom (All-to-All)
  5. Motor Control OPU (One-to-One or All-to-All)

Adding Memory​

  1. Existing circuit: Input → Processing → Output
  2. Create Memory area
  3. Connect Processing → Memory (storage)
  4. Connect Memory → Processing (recall)
  5. Adjust parameters for learning

Troubleshooting​

"Can't create mapping"

  • Verify both areas exist
  • Check you're dragging to input port
  • Ensure areas are in same FEAGI instance
  • Try Quick Connect instead

"Synapse count exceeds limit"

  • Use sparser connectivity rule
  • Reduce cortical area dimensions
  • Limit connections per neuron
  • Increase genome synapse limit (if possible)

"Activity not propagating"

  • Check connection exists and direction is correct
  • Verify source area is active
  • Check PSC is not too small
  • Look for blocking connections (inhibitory)

"Mapping line disappeared"

  • May be hidden (check visualization settings)
  • Might have been deleted accidentally
  • Verify in cortical area Details → Connections

"Too many connections are confusing"

  • Use Split View to focus on sub-circuits
  • Hover to highlight specific connections
  • Hide global connections toggle
  • Work on one region at a time

Back to Overview