UNSUPERVISED, SUPERVISED, AND REINFORCED LEARNING VIA SPIKING COMPUTATION

• US 20130073493A1
• Filed: 09/16/2011
• Published: 03/21/2013
• Est. Priority Date: 09/16/2011
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Modha, Dharmendra S.
• Granted Patent: US 8,874,498 B2
• Time in Patent Office: Days
• US Class Current: 706/25
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
• CPC Class Codes: G06N 3/063 using electronic means
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/088 Non-supervised learning, e.g. competitive learning

Abstract

The present invention relates to unsupervised, supervised and reinforced learning via spiking computation. The neural network comprises a plurality of neural modules. Each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module. An interconnection network comprising a plurality of edges interconnects the plurality of neural modules. Each edge interconnects a first neural module to a second neural module, and each edge comprises a weighted synaptic connection between every neuron in the first neural module and a corresponding neuron in the second neural module.

Claims

1. A neural network, comprising:

a plurality of neural modules, wherein each neural module comprises multiple digital neurons such that each neuron in a neural module has a corresponding neuron in another neural module; and

an interconnection network comprising a plurality of edges that interconnect the plurality of neural modules, wherein each edge interconnects a first neural module to a second neural module and each edge comprises a weighted synaptic connection between every neuron in the first neural module and a corresponding neuron in the second neural module.

2. The neural network of claim 1, wherein each weighted synaptic connection is directed for information flow through the weighted synaptic connection.

3. The neural network of claim 2, wherein:

for a first set of neurons in the first neural module, information flows from each of said neuron in the first set to a corresponding neuron in the second neural module; and

for a second set of neurons in the second neural module, information flows from each of said neuron in the second set to a corresponding neuron in the first neural module.

4. The neural network of claim 1, wherein:

each neural module comprises the same number of neurons as other neural modules; and

all weighted synaptic connections between a pair of neural modules have the same weight.

5. The neural network of claim 1, wherein:

each neuron has an operational state, and based on its operational state, each neuron determines whether to generate a firing event in response to firing events received as input signals by said neuron via weighted synaptic connections from neurons in other neural modules, wherein each said input signal is weighted by the weighted synaptic connection communicating said input signal to said neuron.

6. The neural network of claim 5, further comprising:

a control module generating signals, which define a set of time steps for operation of the neurons;

wherein, at each time step, each neuron in each neural module updates its operational state and determines whether to generate a firing event that is communicated via the interconnection network.

7. The neural network of claim 5, wherein:

the weights of all weighted synaptic connections in the interconnection network are adaptive as a function of the firing events of the interconnected neurons.

8. The neural network of claim 7, further comprising:

an interface module that updates learning rules associated with each weighted synaptic connection in the interconnection network.

9. The neural network of claim 8, wherein:

a learning rule for learning false negatives is applied to a weighted synaptic connection in the interconnection network when a neuron interconnected via the weighted synaptic connection generates a firing event.

10. The neural network of claim 8, wherein:

a learning rule for unlearning false positives is applied to a weighted synaptic connection in the interconnection network when a neuron interconnected via the weighted synaptic connection generates a firing event.

11. The neural network of claim 8, wherein each weighted synaptic connection is reconfigurable for learning rules comprising one or more of:

Hebbian, anti-Hebbian, spike-timing-dependent-plasticity (STDP), and anti-STDP.

12. The neural network of claim 1, wherein:

the neural modules are organized into multiple populations of neural modules, such that a set of neurons within at least one population of neural modules is configured to receive input events from an input interface to the neural network.

13. The neural network of claim 12, wherein a set of neurons within at least one population of the neural modules is configured to generate output events to an output interface of the neural network.

14. The neural network of claim 13, further comprising:

an evaluation module that, based on the type of the output events, feeds the output events to another set of neurons within other populations of neural modules;

wherein a set of neurons within at least one population of the neural modules are configured to generate output events to the evaluation module.

15. The neural network of claim 14, wherein each neural module comprises four neurons.

16. The neural network of claim 15, wherein each neural module comprises:

a neuron in a learning, bottom-up pathway;

a neuron in a learning, top-down pathway;

a neuron in an unlearning, bottom-up pathway; and

a neuron in an unlearning, top-down pathway.

17. The neural network of claim 8, wherein the interface module updates the learning rules such that the neural network operates as an auto-associative system.

18. The neural network of claim 8, wherein the interface module updates the learning rules such that the neural network operates as a hetero-associative system.

19. The neural network of claim 8, wherein the interface module updates the learning rules such that the neural network operates as a reinforcement learning system.

20-29. -29. (canceled)

30. A computer program product on a computer-readable medium for spiking computation in a neural network comprising a plurality of neural modules interconnected via weighted synaptic connections in an interconnection network, wherein each neural module comprises multiple digital neurons such that every neuron in a first neural module is connected to a corresponding neuron in a second neural module via a weighted synaptic connection, the computer program product comprising instructions which when executed on a computer cause the computer to perform operations including:

generating signals which define a set of time steps for operation of the neurons;

each neuron updating its operational state at each time step, and based on that operational state, said neuron determining whether to generate a firing event in response to firing events received as input signals from corresponding neurons in other neural modules, such that each said input signal is weighted by the weighted synaptic connection communicating said input signal to said neuron;

adapting a synaptic weight associated with each weighted synaptic connection as a function of the firing events of the interconnected neurons;

applying a first learning rule for learning false negatives to a first weighted synaptic connection when a neuron interconnected via the first weighted synaptic connection generates a firing event;

applying a second learning rule for unlearning false positives to a second weighted synaptic connection when a neuron interconnected via the second weighted synaptic connection generates a firing event;

a set of neurons generating output events to an evaluation module of the neural network, wherein, based on the type of the output events generated, the output events are fed to another set of neurons; and

updating the learning rules such that the neural network operates as one or more of;

an auto-associative system, a hetero-associative system, and a reinforcement learning system.