Patent EP0360674A2 successfully refreshed from Google Patents.
EP0360674A2
Expired / Lapsed
EP
⏱️ 0.0 years remaining
Signal processing system and learning processing system
Assignee: Sony Corp
•
Filing: 1989-09-15
•
Granted: —
•
Expires: 2009-09-15
Abstract
Abstract The present invention is concerned with a signal processing system having a learning function pursuant to the back-propagation learning rule by the neural network, in which the learning rate is dynamically changed as a function of input values to effect high-speed stable learning. The signal processing system of the present invention is so arranged that, by executing signal processing for the input signals by the recurrent network formed by units (u I1 to u Ix , u H1 to u Hy , u O1 to u Oz ) each corresponding to a neuron, the features of the sequential time series pattern such as voice signals fluctuating on the time axis can be extracted through learning the coupling state of the recurrent network. The present inventicn is also concerned with a learning processing system adapted to cause the signal processing section (10) formed by a neural network (L I , L H , L O ) to undergo signal processing pursuant to the back-propagation learning rule wherein the local minimum state in the course of the learning processing may be avoided by learning the coefficient of coupling strength while simultaneously increasing the number of the unit (u H1 to u Hy ) of the intermediate layer (L H ).
Inventors
- 👤 Atsunobu Hiraiwa
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-11-05 19:00
CPC Classifications
G; G06; G06N; G06N3/00; G06N3/02; G06N3/08; G06N3/082; G06N3/084
Claims Analysis
Total Claims: 10
Claim 1
Independent Claim
237 words
1 A signal processing system comprising a signal processing section including at least an input layer, an intermediate layer and an output layer, each of these layers being made up of a plurality of units each performing signal processing, and a learning processing section for repeatedly and sequentially computing, from the output layer towards the input layer, a coefficient W ji of coupling strength between the units, on the basis of an error data ji between the output value of said output layer for input signal patterns entered into said input layer and and a desired output value afforded as a teacher signal, and performing learning processing of said coefficient W ji of coupling strength, wherein said learning processing section computes a variant ΔW ji of said coefficient W ji of coupling strength on the basis of η·β ( δ pj ·O pi ) where η stands for the learning constant and β the learning variable and supplies the coefficient W ji of coupling strength W ji = W ji + ΔW ji to each unit of said signal processing section, said learning processing section performing the learning processing of the coefficient W ji of coupling strength at such a learning rate in which the learning constant η is normalized by the learning variable β which is indicated as a reciprocal of the square sum of input values O pi in each unit added to by 1.
Claim 2
Dependent Claim
237 words
1 A signal processing system comprising a signal processing section including at least an input layer, an intermediate layer and an output layer, each of these layers being made up of a plurality of units each performing signal processing, and a learning processing section for repeatedly and sequentially computing, from the output layer towards the input layer, a coefficient W ji of coupling strength between the units, on the basis of an error data ji between the output value of said output layer for input signal patterns entered into said input layer and and a desired output value afforded as a teacher signal, and performing learning processing of said coefficient W ji of coupling strength, wherein said learning processing section computes a variant ΔW ji of said coefficient W ji of coupling strength on the basis of η·β ( δ pj ·O pi ) where η stands for the learning constant and β the learning variable and supplies the coefficient W ji of coupling strength W ji = W ji + ΔW ji to each unit of said signal processing section, said learning processing section performing the learning processing of the coefficient W ji of coupling strength at such a learning rate in which the learning constant η is normalized by the learning variable β which is indicated as a reciprocal of the square sum of input values O pi in each unit added to by 1.
Claim 3
Dependent Claim
45 words
2 The signal processing system according to claim 1 wherein delay means are provided in each unit of said intermediate and output layers of said signal processing section and werein said signal processing section is constituted by a recurrent network including loop and feedback circuitry.
Claim 4
Dependent Claim
144 words
3 A signal processing system comprising a signal processing section including at least an input layer, an intermediate layer and an output layer, each of these layers being made up of a plurality of units each performing signal processing, and a learning processing section for repeatedly and sequentially computing, from the output layer towards the input layer, a coefficient of coupling strength between the units, on the basis of an error data between the output value of said output layer for input signal patterns entered into said input layer and a desired output value afforded as a teacher signal, and performing learning processing of said coefficient of coupling strength, wherein each of the units of said intermediate layer and the output layer is provided with delay means and wherein said signal processing section is constituted by a recurrent network including loop and feedback circuitry.
Claim 5
Dependent Claim
144 words
3 A signal processing system comprising a signal processing section including at least an input layer, an intermediate layer and an output layer, each of these layers being made up of a plurality of units each performing signal processing, and a learning processing section for repeatedly and sequentially computing, from the output layer towards the input layer, a coefficient of coupling strength between the units, on the basis of an error data between the output value of said output layer for input signal patterns entered into said input layer and a desired output value afforded as a teacher signal, and performing learning processing of said coefficient of coupling strength, wherein each of the units of said intermediate layer and the output layer is provided with delay means and wherein said signal processing section is constituted by a recurrent network including loop and feedback circuitry.
Claim 6
Dependent Claim
59 words
4 The signal processing system according to any one of claims 1 to 3 wherein said signal processing section is a neural network of a multi-layer structure including an input layer, an intermediate layer and an output layer, each of these layers being constituted by a plurality of units each performing signal processing corresponding to that of a neuron.
Claim 7
Dependent Claim
68 words
5 The signal processing system according to any one of claims 1 to 3 wherein control means are provided in said learning processing section for increasing the number of the units of said intermediate layer, and wherein said learning processing section performs learning processing of the coefficient of coupling strength as said learning processing section causes the number of the units of the intermediate layer to be increased.
Claim 8
Dependent Claim
167 words
6 A learning processing system comprising a signal processing section including at least an input layer, an intermediate layer and an output layer, each of these layers being made up of a plurality of units each performing signal processing, and a learning processing section for repeatedly and sequentially computing, from the output layer towards the input layer, a coefficient W ji of coupling strength between the units, on the basis of an error data δ ji between the output value of said output layer for input signal patterns entered into said input layer and a desired output value afforded as a teacher signal, and performing learning processing of said coefficient of coupling strength, wherein control means are provided in said learning processing section for increasing the number of the units of said intermediate layer, and wherein said learning processing section performs learning processing of the coefficient of coupling strength as said learning processing section causes the number of the units of the intermediate layer to be increased.
Claim 9
Dependent Claim
167 words
6 A learning processing system comprising a signal processing section including at least an input layer, an intermediate layer and an output layer, each of these layers being made up of a plurality of units each performing signal processing, and a learning processing section for repeatedly and sequentially computing, from the output layer towards the input layer, a coefficient W ji of coupling strength between the units, on the basis of an error data δ ji between the output value of said output layer for input signal patterns entered into said input layer and a desired output value afforded as a teacher signal, and performing learning processing of said coefficient of coupling strength, wherein control means are provided in said learning processing section for increasing the number of the units of said intermediate layer, and wherein said learning processing section performs learning processing of the coefficient of coupling strength as said learning processing section causes the number of the units of the intermediate layer to be increased.
Claim 10
Dependent Claim
54 words
7 The learning processing system according to claim 6 wherein said signal processing section is a neural network of a multi-layer structure including an input layer, an intermediate layer and an output layer, each of these layers being constituted by a plurality of units each performing signal processing corresponding to that of a neuron.
References Cited (Backward Citations)
No citation records indexed.
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.