CN109688076A - The blind checking method of noise chaotic neural network based on discrete more level sluggishnesses - Google Patents
The blind checking method of noise chaotic neural network based on discrete more level sluggishnesses Download PDFInfo
- Publication number
- CN109688076A CN109688076A CN201810351419.XA CN201810351419A CN109688076A CN 109688076 A CN109688076 A CN 109688076A CN 201810351419 A CN201810351419 A CN 201810351419A CN 109688076 A CN109688076 A CN 109688076A
- Authority
- CN
- China
- Prior art keywords
- discrete
- level
- neural network
- chaotic neural
- sluggishnesses
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
- H04L25/03006—Arrangements for removing intersymbol interference
- H04L25/03165—Arrangements for removing intersymbol interference using neural networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/0001—Systems modifying transmission characteristics according to link quality, e.g. power backoff
- H04L1/0036—Systems modifying transmission characteristics according to link quality, e.g. power backoff arrangements specific to the receiver
- H04L1/0038—Blind format detection
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Power Engineering (AREA)
- Quality & Reliability (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
- Complex Calculations (AREA)
Abstract
The invention discloses the blind checking method of the noise chaotic neural network based on discrete more level sluggishnesses, include the following steps: that construction receives data matrix XN；To the reception data matrix XNCarry out singular value decomposition；Weight matrix W is setRI, and structural behavior function；Stepped annelaing function is introduced into chaotic neural network, discrete more level sluggishness chaotic neural networks based on stepped annelaing are constructed；The dynamical equation of new model after the improvement of noise chaotic neural network of the building based on discrete more level sluggishnesses, operation is iterated to the dynamical equation of new model after the improvement, then the result of each iteration is substituted into the energy function E (t) of the noise chaotic neural network based on discrete more level sluggishnesses, when the energy function E (t) reaches minimum value, discrete more level sluggishness chaotic neural networks reach balance, and iteration terminates.The present invention improves the noise Mechanics in Chaotic Neural Networks that activation primitive constructs discrete more level sluggishnesses, preferably neural network is avoided to fall into minimum point.
Description
Technical field
The present invention relates to the blind checking method of the noise chaotic neural network based on discrete more level sluggishnesses, belong to wireless
Signal of communication process field more specifically belongs to Hopfield neural network blind Detecting technical field.
Background technique
With discrete hopfield neural network image restore, the extensive use of associative memory etc., network it is steady
Qualitative is the basis of these applications, and network will be stabilized to a fixed point to the end, document [Gao H.S., Zhang J.,
Stability for Discrete Hopfield Neural Networks with Delay[C].2008Fourth
International Conference on Natural Computation,Jinan,China,October 18-20,
2008,560-563.] certain research has been done to the stability of Discrete Neural Network, but these results of study are only limitted to two
The signal of system or two level carries out image procossing.Document [H.J.Liu, Y.Sun.Blind bilevel image
restoration using Hopfield neural networks[A].Proceedings of IEEE
International Conference on Neural Networks[C],San Francisco, CA,USA,Mar.1993
(3): 1656-1661. it] is also limited to real number neural network simultaneously.Document [Zhang Yun, plural Hopfield neural network fanaticism number
Detect the Nanjing [D]: library, Nanjing Univ. of Posts and Telecommunications, 2012:102-147.] and document [Ruan Xiukai, Zhang Zhiyong be based on connect
QAM signal blind Detecting [J] the electronics and information journal of continuous Hopfield type neural network, 2011 (2011 7): 1-6.] point
The plural and continuous more level plural number Hopfield neural networks of discrete more level are not proposed, but are easily trapped into local pole
Small value point, required data volume length are larger.In order to overcome the problems referred above, the present invention is introducing chaotic neural network (Complex-
valued Transiently Chaotic Hopfield Neural Network Real-Imaginary-type Hard-
Multistate-activation-function CTCNN_RIHM) it avoids after falling into locally optimal solution, Annealing function is improved,
Using stepped annelaing function, accelerate the convergence rate of blind Detecting, introduces random noise, avoid chaotic neural network preferably
Local minimum point proposes sluggish feature introducing chaotic neural network according to response characteristic of the sluggishness in neural network
Middle change activation primitive constructs sluggish activation primitive.Then the noise chaotic neural network based on discrete more level sluggishnesses is proposed
Signal blind checking method (Complex-valued Hysteresis Noisy Transiently Chaotic Hopfield
Neural Network Real-Imaginary-type Hard-Multistate-activation-function
HNCTCNN_RIHM), new activation primitive is constructed, the search efficiency of chaotic neural network is improved.
Summary of the invention
For overcome the deficiencies in the prior art, aiming at the problem that performance of chaotic neural network method, random noise is utilized
With feature of the sluggishness in neural network, the invention proposes the noise chaotic neural networks based on discrete more level sluggishnesses
Blind checking method further increases the performance of chaotic neural network blind checking method.
The present invention adopts the following technical scheme: the blind Detecting of the noise chaotic neural network based on discrete more level sluggishnesses
Method, which comprises the steps of:
Step SS1: construction receives data matrix XN；
Step SS2: to the reception data matrix XNCarry out singular value decomposition；
Step SS3: setting weight matrix WRI, and structural behavior function；
Step SS4: stepped annelaing function is introduced into chaotic neural network, and construction is discrete mostly electric based on stepped annelaing
Flat chaotic neural network；
Step SS5: on the basis of discrete more level chaotic neural networks of the stepped annelaing of the step SS4 introduce with
Machine noise, construction is based on discrete more level noise chaotic neural networks, while improving traditional activation primitive, introduces sluggish sharp
Function living improves the error performance of blind checking method, and noise chaotic neural network of the building based on discrete more level sluggishnesses changes
Into the dynamical equation of rear new model, operation is iterated to the dynamical equation of new model after the improvement, is then changed each
The result in generation substitutes into the energy function E (t) of the noise chaotic neural network based on discrete more level sluggishnesses, when the energy
Function E (t) reaches minimum value, and discrete more level sluggishness chaotic neural networks reach balance, and iteration terminates.
As a kind of preferred embodiment, the construction in the step SS1 receives data matrix XNIt specifically includes:
Receiving end receives single user's transmission signal and obtains the reception equation of discrete-time channel through over-sampling:
XN=S ΓT
In formula, XNIt is to receive data matrix, S is to send signal matrix, and Γ is by channel impulse response hjjThe block of composition
Toeplitz matrix；(·)TThe transposition of representing matrix.
A kind of transmission signal matrix S as preferred embodiment, in the step SS1 are as follows:
S=[sL+M(k),L,sL+M(k+N-1)]T=[sN(k),L,sN(k-M-L)]N×(L+M+1)；
Wherein, M is channel exponent number, and L is balanced device order, and N is required data length； sL+M(k)=[s (k), L, s (k-
L-M)]T；Moment k is natural number；
The channel impulse response h in the step SS1jjAre as follows: hjj=[h0,L,hM]q×(M+1), jj=0,1, L, M；q
It is oversample factor, value is positive integer；
The reception data matrix in the step SS1 are as follows: XN=[xL(k),L,xL(k+N-1)]TIt is N × (L+1) q
Data matrix is received, wherein xL(k)=Γ sL+M(k)。
As a kind of preferred embodiment, the step SS2 is specifically included:
To the reception data matrix XNCarry out singular value decomposition, it may be assumed that
In formula, ()HIt is Hermitian transposition；
U is N × (L+M+1) unitary matrice in singular value decomposition；
0 is (N- (L+M+1)) × (L+1) q null matrix；
V is (L+1) q × (L+1) q unitary matrice；
UcIt is N × (N- (L+M+1)) unitary matrice；
D is (L+M+1) × (L+1) q singular value matrix.
As a kind of preferred embodiment, the step SS3 is specifically included: setting weight matrix WRI=[A-QRI], wherein A
It is N × N-dimensional unit matrix,QRIt is the real part for mending projection operator Q, QIIndicate the void of benefit projection operator Q
Portion,Structural behavior function is as follows accordingly:
Wherein, s is N-dimensional complex vector, and the real part of element is sR, imaginary part sI, real and imaginary parts belong to set B, B
=± 1, and ± 3, L, ± gn|gn=1+2 (n-1) }, g1=1, Δ g=gii+1-gii=2, ii ∈ [1, n-1], 2n are to send letter
Number set level number；K is discrete time；
Indicate the optimal estimating value of signal, argmin () indicates that variate-value when being minimized objective function, d are
Delay factor, d=0, L, M+L.
A kind of discrete more level chaos based on stepped annelaing as preferred embodiment, in the step SS4
The dynamical equation of model after the improvement of neural network are as follows:
si(t)=σ (xi(t))
Wherein, si(t), xiIt (t) is respectively S and XNState of i-th of component in t moment, ωijIt is from j-th of component sj
To i-th of component siBetween weight size, and wii=wji；T is discrete more level chaotic neural networks based on stepped annelaing
The time run in iterative process, continuous time t in discrete more level chaotic neural networks based on stepped annelaing and
Discrete time k is realized by Euler's formula and is converted；
α is coefficient of disturbance, and ε is coupling factor；λ is decay factor, and 0≤λ≤1；
σ(xiIt (t)) is the activation primitive of neuron；
Receive signal s (t)=[s1(t), s2(t), L, sN(t)]T, complex signal are as follows: { sj(t)=sRi(t)+i·sij
(t), sRj(t) ∈ B, sIj(t) ∈ B | j=1,2, L, N }, discrete more level chaotic neural networks based on stepped annelaing reach
When finally balancing, it is confirmed as the s of each neuroni(t)=xi(t), siIt (t) is the transmission signal asked；Dividual simulation is moved back
Fiery function zi(t) the self-feedback connection weights value among the adjustment of self feed back coefficient of connection as i-th of neuron, γ are introduced1,
γ2For variable zi(t) control parameter, γ1, γ2∈ (0,1), zi(0) random to generate.
A kind of noise chaos based on discrete more level sluggishnesses as preferred embodiment, in the step SS5
The dynamical equation of new model after the improvement of neural network are as follows:
si(t)=σ (xi(t))
Wherein, si(t), xiIt (t) is respectively S and XNState of i-th of component in t moment, ωijIt is from j-th of component sj
To i-th of component siBetween weight size, and wji=wji；T is the noise chaotic neural network based on discrete more level sluggishnesses
The time run in iterative process, continuous time t in the noise chaotic neural network based on discrete more level sluggishnesses and
Discrete time k is realized by Euler's formula and is converted；
α is coefficient of disturbance, and ε is coupling factor；λ is decay factor, and 0≤λ≤1；
σ(xiIt (t)) is the activation primitive of neuron；
Receive signal s (t)=[s1(t), s2(t), L, sN(t)]TComplex signal are as follows: { sj(t)=sRj(t)+i·sIj
(t), sRj(t) ∈ B, sIj(t) ∈ B | j=1,2, L, N }, the noise chaotic neural network based on discrete more level sluggishnesses reaches
When finally balancing, it is confirmed as the s of each neuroni(t)=xi(t), siIt (t) is the transmission signal asked；
By dividual simulation Annealing function zi(t) it introduces among the adjustment of self feed back coefficient of connection as i-th neuron
Self-feedback connection weights value, γ1,γ2For variable zi(t) control parameter, γ1,γ2∈ (0,1), zi(0) random to generate；
ηi(t) it indicates random noise function, enters local minimum point to further avoid chaotic neural network, in which:
ηi(t)=ηi(t)/ln(e+γ1(1-ηi(t)))。
As a kind of preferred embodiment, the sluggish activation primitive in the step SS5 is σ (x), is specifically expressed as follows:
σ (x)=σR(x)+i·σI(x), and σR(x)=σI(x):
M indicates R or I,It indicates to be rounded downwards, | t | expression takes absolute value, and t is function argument, mod (, N)
Indicate that a is constant, a ∈ (0,1) to N remainder.
As a kind of preferred embodiment, the energy letter of the noise chaotic neural network of discrete more level sluggishnesses
Number E (t) are as follows:
Under synchronized update mode:
Under asynchronous refresh mode:
Wherein:
N indicates the number of the neuron of the noise chaotic neural network of discrete more level sluggishnesses；
E (k) is the energy function of the noise chaotic neural network of discrete more level sluggishnesses；
For receive signal, b=(Δ g) 2,
sRj(k), sIjIt (k) is signal s respectivelyRIj(k) real part and imaginary.
Advantageous effects of the invention: the present invention applies stepped annelaing function and random noise discrete mostly electric
In flat chaotic neural network MQAM constellation signals, and improve the noise chaos mind that activation primitive constructs discrete more level sluggishnesses
Through network model, preferably neural network is avoided to fall into minimum point.New model method can reduce data volume length simultaneously,
The noise resisting ability of more level blind Detectings is improved, comprehensive various aspects improve the performance of more level blind Detectings.MATLAB emulation
The experiment proves that under equal conditions the present invention avoided compared to traditional discrete more level neural network blind checking methods it is sunken
Enter minimum point, reduce data volume length, improve the noise resisting ability of blind Detecting, in contrast to the mostly electric of stepped annelaing
More level noise chaotic neural networks of flat scattered date neural network and stepped annelaing, constringency performance are more preferable.
Detailed description of the invention
Fig. 1 is the noise Mechanics in Chaotic Neural Networks of the invention based on discrete more level sluggishnesses.
Fig. 2 be CTCNN_RIHM with stepped annelaing CTCNN_RIHM under identical signal-to-noise ratio, when the convergence of Blind Detect Algorithm
Between compare figure.
Fig. 3 is CTCNN_RIHM and stepped annelaing CTCNN_RIHM, stepped annelaing NCTCNN_RIHM, stepped annelaing
The bit error rate of the CHNTCNN_RIHM when data volume length is 300 compares figure.CTCNN_RIHM (Complex-valued in figure
Transiently Chaotic Hopfield Neural Network Real-Imaginary-type Hard-
Multistate-activation-function) method is the signal blind checking method of discrete more level chaotic neural networks,
Stepped annelaing CTCNN_RIHM is the signal blind checking method based on discrete more level dividual simulations annealing chaotic neural networks,
Stepped annelaing CNTCNN_RIHM (Complex-valued Noisy Transiently Chaotic Hopfield Neural
Network Real-Imaginary-type Hard-Multistate-activation-function) it is based on discrete more
The signal blind checking method of the level noise dividual simulation annealing flat neural network of chaos, stepped annelaing CHNTCNN_RIHM
(Complex-valued Hysteresis Noisy Transiently Chaotic Hopfield Neural Network
Real-Imaginary-type Hard-Multistate-activation-function) it is sluggish based on discrete more level
Noisy segmentation simulated annealing chaotic neural network signal blind checking method.
It is 300 that Fig. 4, which is stepped annelaing CHNTCNN_RIHM method of the invention in data volume length, signal-to-noise ratio 30dB
When planisphere convergence graph.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.Following embodiment is only used for clearly illustrating this hair
Bright technical solution, and not intended to limit the protection scope of the present invention.
The present invention proposes the signal blind checking method of the noise chaotic neural network based on discrete more level sluggishnesses, tool
Body implementation process is as follows:
When ignoring noise, the reception equation of discrete-time channel is defined as follows
XN=S ΓT (1)
In formula, XNIt is to receive data matrix, S is to send signal battle array, and Γ is by channel impulse response hjjThe block of composition
Toeplitz matrix；(·)TRepresenting matrix transposition；
Wherein, signal matrix is sent:
S=[sL+M(k),L,sL+M(k+N-1)]T=[sN(k),L,sN(k-M-L)]N×(L+M+1),
M is channel exponent number, and L is balanced device order, and N is required data length； sL+M(k)=[s (k) L, s, k (- L-M]T)；
Moment k is natural number；hjj=[h0,L,hM]q×(M+1), jj=0,1, L, M；Q is oversample factor, and value is positive integer；XN=
[xL(k),L,xL(k+N-1)]TIt is that N × (L+1) q receives data matrix, wherein xL(k)=Γ sL+M(k)；
Centainly have when Γ expires column rank for formula (1)Meet QsN(k-d)=0；UcIt is N × (N- (L+M+
1)) tenth of the twelve Earthly Branches basic matrix, by singular value decompositionIn obtain；
Wherein:
(·)HIt is Hermitian transposition；
U is N × (L+M+1) unitary matrice in singular value decomposition；
0 is (N- (L+M+1)) × (L+1) q null matrix；
V is (L+1) q × (L+1) q unitary matrice；
UcIt is N × (N- (L+M+1)) unitary matrice；
D is (L+M+1) × (L+1) q singular value matrix；
Weight matrix W is setRI=[A-QRI], wherein A is N × N-dimensional unit matrix,QRIt is to mend to throw
The real part of shadow operator Q, QIIndicate the imaginary part of benefit projection operator Q,
Structural behavior function and optimization process are as follows accordingly:
Wherein, s is N-dimensional complex vector, and the real part of element is sR, imaginary part sI, real and imaginary parts belong to set B, B
=± 1, and ± 3, L, ± gn|gn=1+2 (n-1) }, g1=1, Δ g=gii+1-gii=2, ii ∈ [1, n-1], 2n are to send letter
Number set level number.Indicate the estimated value of signal, argmin () indicates variate-value when being minimized objective function, d
For delay factor, d=0, L, M+L.In this way, blind Detecting problem just becomes the globally optimal solution problem of formula (3) optimization problem.
Fig. 1 is the signal blind Detecting mould for the noise chaotic neural network based on discrete more level sluggishnesses that the present invention constructs
Type includes weight matrix, activation primitive, decay factor, coupling factor and self feed back item.The dynamical equation of the system are as follows:
si(t)=σ (xi(t)) (5)
Operation is iterated to the equation, the result of each iteration is then substituted into improved discrete more level chaos minds
In energy function E (t) through network, when the energy function E (t) reaches minimum value, i.e. si(t)=si(t-1) when, this is discrete more
Level chaotic neural network reaches balance, and iteration terminates；
Wherein,
si(t), xiIt (t) is respectively S and XNState of i-th of component in t moment, ωijIt is from jth component sjTo i-th
A component siBetween weight size, and wij=wji；T is the time run in network iterative process, the consecutive hours in the network
Between t and discrete time k pass through Euler's formula realize conversion；
ηi(t) it indicates random noise function, enters local minimum point to further avoid chaotic neural network.Wherein:
ηi(t)=ηi(t)/ln(e+γ1(1-ηi(t)))；
α is coefficient of disturbance, and ε is the coupling factor of the network；λ is decay factor, and 0≤λ≤1；
σ (g) is the activation primitive of neuron；
Receive signal s (t)=[s1(t),s2(t),L,sN(t)]T, complex signal { sj(t)=sRj(t)+i·sIj(t),
sRj(t)∈B,sIj(t) ∈ Bj=1,2, L, N }, when which reaches last balance, it can approximately think the s of each neuroni
(t)=xi(t), siIt (t) is required transmission signal；
By stepped annelaing function zi(t) it introduces among the adjustment of self feed back coefficient of connection as the reflexive of i-th neuron
Present connection weight, γ1,γ2For variable zi(t) control parameter, γ1,γ2∈ (0,1), zi(0) random to generate；
σ (g) is the activation primitive of neuron, σ (g)=σR(g)+i·σI(g), and σR(g)=σI(g):
Wherein, m indicates R or I,It indicates to be rounded downwards, | g | expression takes absolute value, and t is argument of function, mod
(, N) and indicate that a is constant, a ∈ (0,1) to N remainder.
B.) energy function
Under synchronized update mode:
Under asynchronous refresh mode:
Wherein:
N indicates the number of the neuron of chaotic neural network；
E (k) is the energy function of the chaotic neural network；
To receive signal, b=(Δ g)2,
sRj(k), sIjIt (k) is signal s respectivelyRIj(k) real part and imaginary；
To realize signal blind Detecting using improved discrete more level chaotic neural networks, solve formula (2), the signal of (3)
Blind Detecting problem makes the minimum point of energy function correspond to the minimum point of performance function.Continuous time t and discrete time k
Between mutually converted by Euler's formula, when neural network reaches stable, x (t) is denoted as the estimated value of s (t)；Energy
The transmission signal that signal at the solution point of the minimum value of function E (k) detects needed for being.
In conclusion the signal blind checking method based on improved discrete more level chaotic neural networks guarantees that network can
To avoid required data volume length is reduced while local minimum point, noise resisting ability is improved, the balance of network is finally reached.
Fig. 2 CTCNN_RIHM with stepped annelaing CTCNN_RIHM under identical signal-to-noise ratio, the convergence time of Blind Detect Algorithm
Compare figure, by comparing figure it is found that using stepped annelaing function blind Detecting convergence rate faster.
Fig. 3 is stepped annelaing CHNTCNN_RIHM method and CTCNN_RIHM method and stepped annelaing of the invention
CTCNN_RIHM method, the contrast simulation lab diagram of stepped annelaing CNTCNN_RIHM method, simulation result here be
Under the same terms, 100 Monte Claro experiments, the method for the present invention and CTCNN_RIHM method and stepped annelaing have been carried out
CTCNN_RIHM method, the bit error rate of the stepped annelaing CNTCNN_RIHM method when data volume length is 300 compare figure.
If Fig. 4 is the planisphere convergence graph of stepped annelaing CHNTCNN_RIHM method of the invention.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the technical principles of the invention, several improvement and deformations can also be made, these improve and become
Shape also should be regarded as protection scope of the present invention.
Claims (9)
1. the blind checking method of the noise chaotic neural network based on discrete more level sluggishnesses, which is characterized in that including walking as follows
It is rapid:
Step SS1: construction receives data matrix XN；
Step SS2: to the reception data matrix XNCarry out singular value decomposition；
Step SS3: setting weight matrix WRI, and structural behavior function；
Step SS4: stepped annelaing function is introduced into chaotic neural network, constructs discrete more level chaos based on stepped annelaing
Neural network；
Step SS5: it introduces on the basis of discrete more level chaotic neural networks of the stepped annelaing of the step SS4 and makes an uproar at random
Sound, construction is based on discrete more level noise chaotic neural networks, while improving traditional activation primitive, introduces sluggish activation primitive
The error performance of blind checking method is improved, new mould after the improvement of the noise chaotic neural network based on discrete more level sluggishnesses is constructed
The dynamical equation of type is iterated operation to the dynamical equation of new model after the improvement, then in the result generation of each iteration
In the energy function E (t) for entering the noise chaotic neural network based on discrete more level sluggishnesses, when the energy function E (t) reaches
Minimum value, discrete more level sluggishness chaotic neural networks reach balance, and iteration terminates.
2. the blind checking method of the noise chaotic neural network according to claim 1 based on discrete more level sluggishnesses,
It is characterized in that, the construction in the step SS1 receives data matrix XNSpecifically include: receiving end receives single user and sends
Signal obtains the reception equation of discrete-time channel through over-sampling:
XN=S ΓT
In formula, XNIt is to receive data matrix, S is to send signal matrix, and Γ is by channel impulse response hjjThe block of composition
Toeplitz matrix；(·)TThe transposition of representing matrix.
3. the blind checking method of the noise chaotic neural network according to claim 2 based on discrete more level sluggishnesses,
It is characterized in that, the transmission signal matrix S in the step SS1 are as follows:
S=[sL+M(k),L,sL+M(k+N-1)]T=[sN(k),L,sN(k-M-L)]N×(L+M+1)；
Wherein, M is channel exponent number, and L is balanced device order, and N is required data length；sL+M(k)=[s (k), L, s (k-L-M)]T；
Moment k is natural number；
The channel impulse response h in the step SS1jjAre as follows: hjj=[h0,L,hM]q×(M+1), jj=0,1, L, M；Q was
Decimation factor, value are positive integer；
The reception data matrix in the step SS1 are as follows: XN=[xL(k),L,xL(k+N-1)]TIt is that N × (L+1) q receives number
According to matrix, wherein xL(k)=Γ sL+M(k)。
4. the blind checking method of the noise chaotic neural network according to claim 1 based on discrete more level sluggishnesses,
It is characterized in that, the step SS2 is specifically included:
To the reception data matrix XNCarry out singular value decomposition, it may be assumed that
In formula, ()HIt is Hermitian transposition；
U is N × (L+M+1) unitary matrice in singular value decomposition；
0 is (N- (L+M+1)) × (L+1) q null matrix；
V is (L+1) q × (L+1) q unitary matrice；
UcIt is N × (N- (L+M+1)) unitary matrice；
D is (L+M+1) × (L+1) q singular value matrix.
5. the blind checking method of the noise chaotic neural network according to claim 1 based on discrete more level sluggishnesses,
It is characterized in that, the step SS3 is specifically included: setting weight matrix WRI=[A-QRI], wherein A is N × N-dimensional unit matrix,QRIt is the real part for mending projection operator Q, QIIndicate the imaginary part of benefit projection operator Q,According to
This structural behavior function is as follows:
Wherein, s is N-dimensional complex vector, and the real part of element is sR, imaginary part sI, real and imaginary parts belong to set B, B=±
1,±3,L,±gn|gn=1+2 (n-1) }, g1=1, Δ g=gii+1-gii=2, ii ∈ [1, n-1], 2n are to send signal set
Level number；K is discrete time；
Indicate the optimal estimating value of signal, argmin () indicates variate-value when being minimized objective function, d be delay because
Son, d=0, L, M+L.
6. the blind checking method of the noise chaotic neural network according to claim 1 based on discrete more level sluggishnesses,
It is characterized in that, model after the improvement of discrete more level chaotic neural networks based on stepped annelaing in the step SS4
Dynamical equation are as follows:
si(t)=σ (xi(t))
Wherein, si(t), xiIt (t) is respectively S and XNState of i-th of component in t moment, ωijIt is from j-th of component sjTo i-th
A component siBetween weight size, andT is discrete more level chaotic neural network iteration based on stepped annelaing
The time run in the process, continuous time t in discrete more level chaotic neural networks based on stepped annelaing and discrete
Time k is realized by Euler's formula and is converted；
α is coefficient of disturbance, and ε is coupling factor；λ is decay factor, and 0≤λ≤1；
σ(xiIt (t)) is the activation primitive of neuron；
Receive signal s (t)=[s1(t),s2(t),L,sN(t)]T, complex signal are as follows:: sj (t)=sRj (t)+isIj (t),
SRj (t) ∈ B, sIj (t) ∈ B | j=1,2, L, N }；Discrete more level chaotic neural networks based on stepped annelaing reach last
When balance, it is confirmed as the s of each neuroni(t)=xi(t), siIt (t) is the transmission signal asked；By dividual simulation Annealing function
zi(t) the self-feedback connection weights value among the adjustment of self feed back coefficient of connection as i-th of neuron, γ are introduced1,γ2For variable
zi(t) control parameter, γ1,γ2∈ (0,1), zi(0) random to generate.
7. the blind checking method of the noise chaotic neural network according to claim 1 based on discrete more level sluggishnesses,
It is characterized in that, new mould after the improvement of the noise chaotic neural network based on discrete more level sluggishnesses in the step SS5
The dynamical equation of type are as follows:
si(t)=σ (xi(t))；
Wherein, si(t), xiIt (t) is respectively S and XNState of i-th of component in t moment, ωijIt is from j-th of component sjTo i-th
A component siBetween weight size, and wij=wji；T is the noise chaotic neural network iteration mistake based on discrete more level sluggishnesses
The time run in journey, continuous time t in the noise chaotic neural network based on discrete more level sluggishnesses and it is discrete when
Between k pass through Euler's formula realize conversion；
α is coefficient of disturbance, and ε is coupling factor；λ is decay factor, and 0≤λ≤1；
σ(xiIt (t)) is the activation primitive of neuron；
Receive signal s (t)=[s1(t),s2(t),L,sN(t)]T, complex signal are as follows: { sj(t)=sRj(t)+i·sIj(t), sRj
(t) ∈ B | j=1,2, L, N }, when the noise chaotic neural network based on discrete more level sluggishnesses reaches last balance, it is confirmed as
The s of each neuroni(t)=xi(t), siIt (t) is the transmission signal asked；
By dividual simulation Annealing function zi(t) self feed back among the adjustment of self feed back coefficient of connection as i-th of neuron is introduced
Connection weight, γ1,γ2For variable zi(t) control parameter, γ1,γ2∈ (0,1), zi(0) random to generate；
ηi(t) it indicates random noise function, enters local minimum point to further avoid chaotic neural network, in which: ηi(t)
=ηi(t)/ln(e+γ1(1-ηi(t)))。
8. the blind checking method of the noise chaotic neural network according to claim 1 based on discrete more level sluggishnesses,
It is characterized in that, the sluggish activation primitive in the step SS5 is σ (x), is specifically expressed as follows: σ (x)=σR(x)+i·σI(x),
And σR(x)=σI(x):
M indicates R or I,It indicates to be rounded downwards, | t | expression takes absolute value, and t is function argument, mod (, N) and it indicates
To N remainder, a is constant, a ∈ (0,1).
9. the blind checking method of the noise chaotic neural network according to claim 1 based on discrete more level sluggishnesses,
It is characterized in that, the energy function E (t) of the noise chaotic neural network of discrete more level sluggishnesses are as follows:
Under synchronized update mode:
Under asynchronous refresh mode:
Wherein:
N indicates the number of the neuron of the noise chaotic neural network of discrete more level sluggishnesses；
E (k) is the energy function of the noise chaotic neural network of discrete more level sluggishnesses；
To receive signal, b=(Δ g)2,
sRj(k), sIjIt (k) is signal s respectivelyRIj(k) real part and imaginary.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810351419.XA CN109688076B (en) | 2018-04-19 | 2018-04-19 | Blind detection method of noise chaotic neural network based on discrete multi-level hysteresis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810351419.XA CN109688076B (en) | 2018-04-19 | 2018-04-19 | Blind detection method of noise chaotic neural network based on discrete multi-level hysteresis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109688076A true CN109688076A (en) | 2019-04-26 |
| CN109688076B CN109688076B (en) | 2021-06-11 |
Family
ID=66184356
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810351419.XA Active CN109688076B (en) | 2018-04-19 | 2018-04-19 | Blind detection method of noise chaotic neural network based on discrete multi-level hysteresis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109688076B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111314255A (en) * | 2020-02-13 | 2020-06-19 | 南京航空航天大学 | Low-complexity SISO and MIMO receiver generation method |
| CN113219836A (en) * | 2021-05-19 | 2021-08-06 | 安徽大学 | Projection synchronization method of fractional order complex value memristor neural network and application of projection synchronization method |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1606725A1 (en) * | 2003-03-22 | 2005-12-21 | Qinetiq Limited | Monitoring electrical muscular activity |
| CN101719885A (en) * | 2009-11-27 | 2010-06-02 | 南京邮电大学 | Multi-level signal blind detection method based on discrete unity-feedback neutral network |
| US20130082858A1 (en) * | 2010-08-26 | 2013-04-04 | David H. Chambers | Classification of subsurface objects using singular values derived from signal frames |
| CN103152133A (en) * | 2013-02-26 | 2013-06-12 | 南京邮电大学 | Hysteretic all feedback neural network-based signal blind detection method |
| CN103888391A (en) * | 2014-03-07 | 2014-06-25 | 南京邮电大学 | Signal blind detection method based on double Sigmoid chaotic neural network |
| CN104751675A (en) * | 2015-04-01 | 2015-07-01 | 苏州闻捷传感技术有限公司 | Parking space detection method based on limited information rate theory pulse signals |
| CN106059385A (en) * | 2016-07-20 | 2016-10-26 | 南京理工大学 | Piezoelectric ceramic driving power supply with hysteresis compensation function |
| CN106101035A (en) * | 2016-06-01 | 2016-11-09 | 四川东鼎里智信息技术有限责任公司 | The signal blind checking method of unity-feedback neutral network |
| CN106130689A (en) * | 2016-06-13 | 2016-11-16 | 南京邮电大学 | A kind of non-linear self-feedback chaotic neural network signal blind checking method |
| CN106612158A (en) * | 2016-12-06 | 2017-05-03 | 南京邮电大学 | Signal blind detection method based on complex sinusoidal chaotic neural network |
| CN106953820A (en) * | 2017-04-07 | 2017-07-14 | 南京邮电大学 | Signal blind checking method based on the plural continuous neural networks of double Sigmoid |
-
2018
- 2018-04-19 CN CN201810351419.XA patent/CN109688076B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1606725A1 (en) * | 2003-03-22 | 2005-12-21 | Qinetiq Limited | Monitoring electrical muscular activity |
| CN101719885A (en) * | 2009-11-27 | 2010-06-02 | 南京邮电大学 | Multi-level signal blind detection method based on discrete unity-feedback neutral network |
| US20130082858A1 (en) * | 2010-08-26 | 2013-04-04 | David H. Chambers | Classification of subsurface objects using singular values derived from signal frames |
| CN103152133A (en) * | 2013-02-26 | 2013-06-12 | 南京邮电大学 | Hysteretic all feedback neural network-based signal blind detection method |
| CN103888391A (en) * | 2014-03-07 | 2014-06-25 | 南京邮电大学 | Signal blind detection method based on double Sigmoid chaotic neural network |
| CN104751675A (en) * | 2015-04-01 | 2015-07-01 | 苏州闻捷传感技术有限公司 | Parking space detection method based on limited information rate theory pulse signals |
| CN106101035A (en) * | 2016-06-01 | 2016-11-09 | 四川东鼎里智信息技术有限责任公司 | The signal blind checking method of unity-feedback neutral network |
| CN106130689A (en) * | 2016-06-13 | 2016-11-16 | 南京邮电大学 | A kind of non-linear self-feedback chaotic neural network signal blind checking method |
| CN106059385A (en) * | 2016-07-20 | 2016-10-26 | 南京理工大学 | Piezoelectric ceramic driving power supply with hysteresis compensation function |
| CN106612158A (en) * | 2016-12-06 | 2017-05-03 | 南京邮电大学 | Signal blind detection method based on complex sinusoidal chaotic neural network |
| CN106953820A (en) * | 2017-04-07 | 2017-07-14 | 南京邮电大学 | Signal blind checking method based on the plural continuous neural networks of double Sigmoid |
Non-Patent Citations (2)
| Title |
|---|
| SHUJUAN YU: ""A novel blind detection algorithm based on double sigmoid hysteretic chaotic hopfield neural network "", 《JOURNAL OF INFORMATION HIDING AND MULTIMEDIA SIGNAL PROCESSING 》 * |
| 张昀: ""基于复数Hopfield神经网络的盲信号检测"", 《CNKI博士学位论文全文库信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111314255A (en) * | 2020-02-13 | 2020-06-19 | 南京航空航天大学 | Low-complexity SISO and MIMO receiver generation method |
| CN113219836A (en) * | 2021-05-19 | 2021-08-06 | 安徽大学 | Projection synchronization method of fractional order complex value memristor neural network and application of projection synchronization method |
| CN113219836B (en) * | 2021-05-19 | 2022-04-08 | 安徽大学 | Projection synchronization method of fractional order complex value memristor neural network and application of projection synchronization method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109688076B (en) | 2021-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Extended Kalman filter under maximum correntropy criterion | |
| CN109462520B (en) | Network traffic resource situation prediction method based on LSTM model | |
| CN108566257B (en) | Signal recovery method based on back propagation neural network | |
| Zhu et al. | Heterogeneous learning in zero-sum stochastic games with incomplete information | |
| Xu et al. | Adaptive control of local updating and model compression for efficient federated learning | |
| CN111641570B (en) | Joint equipment detection and channel estimation method based on deep learning | |
| CN103888391B (en) | Signal blind detection method based on double Sigmoid chaotic neural network | |
| CN109688076A (en) | The blind checking method of noise chaotic neural network based on discrete more level sluggishnesses | |
| CN106130689A (en) | A kind of non-linear self-feedback chaotic neural network signal blind checking method | |
| CN113676266B (en) | Channel modeling method based on quantum generation countermeasure network | |
| WO2022160528A1 (en) | Method for simulating quantum condition master equation in quantum transport process by cyclic neural network | |
| CN108199990B (en) | non-Gaussian noise 3D-MIMO channel estimation method | |
| Li et al. | Communication-efficient decentralized zeroth-order method on heterogeneous data | |
| CN109618311B (en) | Blind detection algorithm based on M2M communication spectrum sharing and coexistence | |
| CN108574653A (en) | Signal blind checking method based on double Sigmoid sluggishnesses noise chaotic neural networks | |
| CN106953820A (en) | Signal blind checking method based on the plural continuous neural networks of double Sigmoid | |
| CN113743012B (en) | Cloud-edge collaborative mode task unloading optimization method under multi-user scene | |
| Mohamed et al. | Generalized Gaussian mixture model | |
| Kajino | A differentiable point process with its application to spiking neural networks | |
| CN115190633A (en) | Intelligent frequency spectrum on-line anti-interference frequency point allocation method for communication between devices | |
| CN114330743A (en) | Cross-equipment federal learning method for minimum-maximum problem | |
| CN111045861B (en) | Sensor data recovery method based on deep neural network | |
| Rieger et al. | Client adaptation improves federated learning with simulated non-IID clients | |
| CN108494524A (en) | Signal blind checking method based on improved discrete more level chaotic neural networks | |
| Qian et al. | Delay-probability-dependent state estimation for neural networks with hybrid delays |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |