A Signal-Flow-Graph Approach to On-line Gradient Calculation

@article{Campolucci2000ASA,
  title={A Signal-Flow-Graph Approach to On-line Gradient Calculation},
  author={Paolo Campolucci and Aurelio Uncini and Francesco Piazza},
  journal={Neural Computation},
  year={2000},
  volume={12},
  pages={1901-1927}
}
A large class of nonlinear dynamic adaptive systems such as dynamic recurrent neural networks can be effectively represented by signal flow graphs (SFGs). By this method, complex systems are described as a general connection of many simple components, each of them implementing a simple one-input, one-output transformation, as in an electrical circuit. Even if graph representations are popular in the neural network community, they are often used for qualitative description rather than for… 
A general approach to gradient based learning in multirate systems and neural networks
  • F. Rosati, P. Campolucci, F. Piazza
  • Computer Science
    Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks. IJCNN 2000. Neural Computing: New Challenges and Perspectives for the New Millennium
  • 2000
TLDR
A new algorithm to estimate the derivative of the output with respect to an internal parameter have been proposed in the literature for discrete-time systems and can be employed for gradient-based learning of general multirate circuits, such as the new "multirate" neural networks.
A novel signal flow graph based solution for calculation of first and second derivatives of dynamical nonlinear systems
TLDR
An attempt of solution by means of signal flow graph (SFG) techniques is proposed and the ability of the proposed method to determine all Hessian matrix entries in a complete automatic way is highlighted.
Learning in linear and nonlinear multirate digital systems by signal flow graphs
TLDR
The proposed method is based on the concept of adjoint graph, which allows one to estimate the derivative of the output with respect to an internal parameter at different rates (time-instant), and can be employed for the gradient-based adaptation of general multirate circuits.
Higher‐order differentiation of network functions using signal flow graphs
  • S. Osowski
  • Mathematics
    Int. J. Circuit Theory Appl.
  • 2012
TLDR
The application of signal flow graphs (SFG) in the calculation of higher-order derivatives (sensitivities) of the linear circuit functions is presented, showing that the SFG approach to the sensitivity calculation allows to reduce greatly the complexity of calculations.
Calculation of non-mixed second derivatives in multirate systems through signal flow graph techniques
TLDR
The overall algorithm represents a useful tool for determination of Jacobean and Hessian based information in learning systems, as was already done in other related but less general contributions in the literature.
Processing of audio signals by neural networks
TLDR
In this tutorial low structural-computational complexity neural networks, suitable for real-time DSP applications, are presented and several audio real-world applications are discussed.
A Neural Abstract Machine
TLDR
This work defines a parameterized Neural Abstract Machine in such a way that the major neural networks in the literature can be described as natural exten- sions or refinements of the NAM.
Audio signal processing by neural networks
Modelling in the Neural Abstract Machine Framework
TLDR
This work proposes to use Abstract State Machines to define a Neural Abstract Machine, a machine able to manipulate the basic objects and functions which constitute the essence of neural computation.
Audio signal processing byneural networks
TLDR
A review of architectures suitable for nonlinear real-time audio signal processing is presented and applications in the -elds of audio signal recovery, speech quality enhancement, and learning based pseudo-phy sical sound synthesis are presented and discussed.
...
...

References

SHOWING 1-10 OF 36 REFERENCES
Signal-flow-graph derivation of on-line gradient learning algorithms
TLDR
A new general method for backward gradient computation of a system output or cost function with respect to past (or present) system parameters is derived, using the signal-flow-graph representation and its known properties.
A general approach to gradient based learning in multirate systems and neural networks
  • F. Rosati, P. Campolucci, F. Piazza
  • Computer Science
    Proceedings of the IEEE-INNS-ENNS International Joint Conference on Neural Networks. IJCNN 2000. Neural Computing: New Challenges and Perspectives for the New Millennium
  • 2000
TLDR
A new algorithm to estimate the derivative of the output with respect to an internal parameter have been proposed in the literature for discrete-time systems and can be employed for gradient-based learning of general multirate circuits, such as the new "multirate" neural networks.
Dynamical systems learning by a circuit theoretic approach
  • P. Campolucci, A. Uncini, F. Piazza
  • Computer Science, Mathematics
    ISCAS '98. Proceedings of the 1998 IEEE International Symposium on Circuits and Systems (Cat. No.98CH36187)
  • 1998
TLDR
A new general method is derived for both on-line and off-line backward gradient computation of a system output, or cost function, with respect to system parameters, using a circuit theoretic approach.
Learning in linear and nonlinear multirate digital systems by signal flow graphs
TLDR
The proposed method is based on the concept of adjoint graph, which allows one to estimate the derivative of the output with respect to an internal parameter at different rates (time-instant), and can be employed for the gradient-based adaptation of general multirate circuits.
A learning algorithm for analog, fully recurrent neural networks
  • M. Gherrity
  • Computer Science
    International 1989 Joint Conference on Neural Networks
  • 1989
A learning algorithm for recurrent neural networks is derived. This algorithm allows a network to learn specified trajectories in state space in response to various input sequences. The network
Signal flow graphs-Computer-aided system analysis and sensitivity calculations
TLDR
The introduction of two matrices, the summing matrix S and the branching matrix B, which completely describe the topology of a signal flow graph leads to a formulation of system equations in terms of submatrices of the S- and B-matrices suitable for digital-computer programming.
An Efficient Gradient-Based Algorithm for On-Line Training of Recurrent Network Trajectories
A novel variant of the familiar backpropagation-through-time approach to training recurrent networks is described. This algorithm is intended to be used on arbitrary recurrent networks that run
Backpropagation Through Time: What It Does and How to Do It
TLDR
This paper first reviews basic backpropagation, a simple method which is now being widely used in areas like pattern recognition and fault diagnosis, and describes further extensions of this method, to deal with systems other than neural networks, systems involving simultaneous equations or true recurrent networks, and other practical issues which arise with this method.
Back propagation through adjoints for the identification of nonlinear dynamic systems using recurrent neural models
TLDR
Back propagation is reinvestigated for an efficient evaluation of the gradient in arbitrary interconnections of recurrent subsystems and it is shown that it is sufficient to back propagate as many time steps as the order of the system for convergence.
Gradient methods for the optimization of dynamical systems containing neural networks
An extension of the backpropagation method, termed dynamic backpropagation, which can be applied in a straightforward manner for the optimization of the weights (parameters) of multilayer neural
...
...