An Improved Multi-Dimensional Approach to Wave Digital Filters with Topology-Related Delay-Free Loops using Automatic Differentiation

Kolonko, Lech; Velten, Jorg; Kummert, Anton

doi:10.1109/mwscas.2019.8885341

Cited by 6 publications

(2 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the early nineties, despite the limited computational resources available at the time, Feldmann et al [13] advocated for the use of AD to evaluate the Jacobian matrices needed for circuit simulations. In [14], AD was adopted for optimizing artificial port resistances in multi-dimensional Wave Digital Filter (WD) structures with topology-related delayfree-loops; the same approach was then extended in [15] for the case of WD structures containing multiple nonlinearities. Recently, Shintani et al [16], building upon the results of [17], showed that AD-based parameter extraction of a power metaloxide-semiconductor field-effect transistor (MOSFET) can be up to 3.5 times faster than a corresponding numerical differentiation method.…”

Section: Introductionmentioning

confidence: 99%

Data-Driven Parameter Estimation of Lumped-Element Models via Automatic Differentiation

Mezza,

Giampiccolo,

Bernardini

2023

IEEE Access

View full text Add to dashboard Cite

Lumped-element models (LEMs) provide a compact characterization of numerous real-world physical systems, including electrical, acoustic, and mechanical systems. However, even when the target topology is known, deriving model parameters that approximate a possibly distributed system often requires educated guesses or dedicated optimization routines. This article presents a general framework for the data-driven estimation of lumped parameters using automatic differentiation. Inspired by recent work on physical neural networks, we propose to explicitly embed a differentiable LEM in the forward pass of a learning algorithm and discover its parameters via backpropagation. The same approach could also be applied to blindly parameterize an approximating model that shares no isomorphism with the target system, for which it would be thus challenging to exploit prior knowledge of the underlying physics. We evaluate our framework on various linear and nonlinear systems, including time-and frequency-domain learning objectives, and consider real-and complex-valued differentiation strategies. In all our experiments, we were able to achieve a near-perfect match of the system state measurements and retrieve the true model parameters whenever possible. Besides its practical interest, the present approach provides a fully interpretable inputoutput mapping by exposing the topological structure of the underlying physical model, and it may therefore constitute an explainable ad-hoc alternative to otherwise black-box methods.

show abstract

Section: Introductionmentioning

confidence: 99%

Data-Driven Parameter Estimation of Lumped-Element Models via Automatic Differentiation

Mezza,

Giampiccolo,

Bernardini

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…For instance, it provides run-time efficient algorithms well interpretable in terms of electrical circuits due to their direct correspondence with the reference circuit. Moreover, multidimensional wave digital algorithms [18][19][20][21][22] can be used, which can be very efficient for emulating larger circuit setups, cf. 19 The wave digital concept is hence well suited for this work.…”

Section: Introductionmentioning

confidence: 99%

Wave digital model of calcium‐imaging‐based neuronal activity of mice

Jenderny

Ochs

2022

Int J Numerical Modelling

View full text Add to dashboard Cite

The modeling of real neuronal networks by close‐to‐biology circuits can contribute to a better understanding of these networks from a circuit theoretical point of view. This can potentially lead to novel design principles for electrical circuits. This modeling process involves the mimicking of real measured neuronal activity, which is often only available as calcium imaging recordings. In this work, we develop a wave digital model as a circuit emulator capable of mimicking the calcium‐imaging‐based neuronal activity of mouse neurons. For this purpose, we design an electrical reference circuit of a mouse neuron based on an extended Morris–Lecar model in combination with an RC circuit. The circuit parameters are determined by a parameter scaling without requiring any parameter optimization. The input signal of the resulting wave digital model is designed based on estimating spike times with the spike inference software MLspike. Wave digital emulations show the successful mimicking of calcium‐based neuronal activity for mouse neurons located in the visual, motor and somatosensory cortex as well the possibility to predict ion currents occurring during this activity. This proves the suitability of the reference circuit as a model for the considered types of mouse neurons. It also implies that the circuit can act as an observer for electrical signals of mouse neurons. This is achieved despite the lack of electrophysiological measurements in most cases. Hence, the proposed wave digital model can be seen as an important step toward being able to verify close‐to‐biology electrical circuits as a model of real neuronal networks without electrophysiology.

show abstract

Parallelized Hardware Acceleration of Automatic Differentiating Wave Digital Filters

Kolonko,

Velten,

Kummert

2024

2024 IEEE International Symposium on Circuits and Systems (ISCAS)

View full text Add to dashboard Cite

An Improved Multi-Dimensional Approach to Wave Digital Filters with Topology-Related Delay-Free Loops using Automatic Differentiation

Cited by 6 publications

References 8 publications

Data-Driven Parameter Estimation of Lumped-Element Models via Automatic Differentiation

Data-Driven Parameter Estimation of Lumped-Element Models via Automatic Differentiation

Wave digital model of calcium‐imaging‐based neuronal activity of mice

Parallelized Hardware Acceleration of Automatic Differentiating Wave Digital Filters

Contact Info

Product

Resources

About