E-cell (equivalent cell) and the basic circuit modules of e-circuits: e-cell pair totem, the basic memory circuit and association extension

• US 6,397,201 B1
• Filed: 11/30/1998
• Issued: 05/28/2002
• Est. Priority Date: 12/02/1997
• Status: Expired due to Fees

• Current Assignee: David W. Arathorn
• Original Assignee: David W. Arathorn
• Inventors: Arathorn, David W.
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): STARKS, WILBERT L
• Application Number:
  US09/201,395
• Time in Patent Office: 1,275 Days
• Field of Search: 706/26,33,40,41 257/14 708/801 382/257
• US Class Current: 706/33
• CPC Class Codes: G06N 3/065 Analogue means

Abstract

The e-cell is a computational element for the construction of circuits capable of performing transform invariant pattern recognition, scene segmentation, and regularity extraction in a variety of sensory modes including vision, hearing, olefaction, touch, etc. The e-cell consists of a cell body, an output terminal, and set of tree-connected input regions each of which have one or more input terminals and corresponding stored weights. The e-cell is capable of learning and subsequently discriminating multidimensional (in the mathematical sense) activation patterns applied to its inputs. The strictness with which the e-cell requires the presence in the target of each dimension of a learned pattern is dynamically adjustable, thus making it suitable to recognition of learned patterns under occlusion and various kinds of degradation. The strictness of discrimination across the region dimension is independent of the strictness of recognition within each dimension. The e-cell can also be configured to perform non-memory roles such as contrast enhancement, pattern comparison, signal routing, and event timing.

The e-cell pair totem provides a circuit construction module of cortically inspired circuits for transform invariant pattern recognition and scene segmentation in a variety of sensory modes including vision, hearing, olefaction, touch, etc. The e-cell pair totem implements the forward and backward paths, feedback connections and consequent oscillatory dynamics characteristic of e-circuits. The basic totem circuit is specialized to create a basic memory circuit. The basic memory circuit is extended to allow contextual and associative information to influence recognition by the memory stage of the basic memory circuit.

Claims

1. An e-cell pair, comprising:

a forward e-cell having a forward e-cell input region, a forward e-cell cell body, a forward e-cell inhibitory synapse, and a forward e-cell output;

a backward e-cell having a backward e-cell input region, a backward e-cell cell body, and a backward e-cell output; and

an inter e-cell having an inter e-cell excitatory synapse and an inter e-cell output;

wherein the inter e-cell output is coupled to the forward e-cell inhibitory synapse, and the backward e-cell output is coupled to the inter e-cell excitatory synapse.

2. An e-cell pair as in claim 1,

wherein the backwarde-cell further has a backward e-cell excitatory synapse, and the forward e-cell output is coupled to the backward e-cell excitatory synapse.

3. An e-cell pair totem, comprising:

a first e-cell pair having a first forward e-cell input region, a first forward e-cell output, a first backward e-cell input region, and a first backward e-cell output; and

a second e-cell pair having a second forward e-cell input region, a second forward e-cell output, a second backward e-cell input region, and a second backward e-cell output;

wherein the first forward e-cell output is coupled to a forward synapse within the second forward e-cell input region, and the second backward e-cell output is coupled to a backward synapse within the first backward e-cell input region.

4. An e-cell pair totem as in claim 3,

wherein the first e-cell pair further has a first control input, and the second e-cell pair further has a second control input.

5. An e-cell circuit, comprising:

a first plurality of first e-cell pairs each having a first forward e-cell input region, a first forward e-cell output, a first backward e-cell input dendritic tree structure, and a first backward e-cell output;

a second plurality of second e-cell pairs each having a second forward e-cell input dendritic tree structure, a second forward e-cell output, a second backward e-cell input region, and a second backward e-cell output; and

wherein the second forward e-cell input dendritic tree structures each include one or more forward synapses coupled to one of the first forward e-cell outputs, and the first backward e-cell input dendritic tree structures each include one or more backward synapses coupled to one of the second backward e-cell outputs.

6. A e-cell circuit as in claim 5,

wherein each of the second forward e-cell input dendritic tree structures and the first backward e-cell input dendritic tree structures includes at least one dendritic combining node.

7. A memory circuit, comprising:

a stage b stage having k_b stage-b e-cell pairs, each stage-b e-cell pair having a stage b forward input, a stage b forward output, and k_m stage-b backward input synapses; and

a stage-m stage having k_m stage-m e-cell pairs, each stage-m e-cell pair having k_b stage-m forward input synapses and a stage-m backward output;

wherein each of the k_b stage-b forward outputs is coupled to a corresponding stage-m one of the k_b stage-m forward input synapses, and each of the k_m stage-m outputs is coupled to a corresponding stage-b one of the k_m stage-b backward input synapses.

8. A memory circuit as in claim 7,

wherein the stage-b stage comprises p_b clusters of q_b e-cell pairs; and

wherein each stage-m e-cell pair has a stage-m dendritic input tree having p_b regions each having a corresponding q_b of the k_b stage-m forward input synapses.

9. A memory circuit as in claim 7,

wherein each of the k_m stage-in forward e-cell pairs further has a stage-m learning synapse, and each of the k_b stage-b backward e-cell pairs further has a stage b learning synapse.

10. A memory circuit as in claim 9,

wherein each of the k_m stage-m learning synapses belongs to a region of a stage-m dendritic input tree, and wherein each of the k_b stage-b learning synapses belongs to a region of a stage b dendritic input tree.