Pith. sign in

REVIEW 4 major objections 4 minor 41 references

iVAMS 3.0: Hierarchical-Machine-Learning-Metamodel-Integrated Intelligent Verilog-AMS for Ultra-Fast, Accurate Mixed-Signal Design Optimization

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Kriging-seeded neural nets make analog chip optimization 24x faster

desk verdict A re-packaging of the authors' own conference papers whose central speedup claim is directly contradicted by its own Table 4. read the letter →

arxiv 2506.01045 v1 pith:RK57NXI4 submitted 2025-06-01 cs.ET

Saraju P. Mohanty , Elias Kougianos This is my paper · ORCID
classification cs.ET
keywords metamodelingKrigingartificialneuralnetworkprocessvariationmixed-signaldesignparticleswarmoptimizationphase-lockedloopnano-CMOS
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes a design flow that replaces slow SPICE-based Monte Carlo analysis of analog/mixed-signal circuits with a metamodel trained on a hybrid of Kriging and an artificial neural network. The central idea is to use Kriging to generate bootstrapped training points that carry information about correlated process variation, then train a fast ANN on those points. The paper claims this Kriging-bootstrapped ANN metamodel is roughly 25 times faster to run Monte Carlo analysis on than a pure Kriging metamodel, while preserving most of Kriging's accuracy. If correct, this would let designers optimize variability-aware nano-CMOS circuits like a 21-parameter phase-locked loop in minutes rather than days.

What carries the argument

The load-bearing object is the Kriging bootstrapped ANN metamodel: a feedforward neural network whose training set is replaced by Kriging-generated point estimates rather than raw LHS samples. Kriging, a geostatistical interpolation method, predicts each response as a weighted combination of basis functions plus a zero-mean stochastic process whose covariance encodes autocorrelation between parameters; the ANN then learns this correlation-augmented surface. The second mechanism is a particle swarm optimizer that searches over the metamodel to minimize $\mu_{\text{pwr}} + 3\sigma_{\text{pwr}}$ subject to a locking-time constraint, with Monte Carlo runs on the metamodel estimating the statistical moments at each candidate design.

What would settle it

Run a fully held-out Monte Carlo set through the parameterized SPICE netlist—say another 1000 points not used in Kriging bootstrapping or ANN training—and compare the mean and standard deviation of power, frequency, locking time, and jitter against the Kriging-bootstrapped ANN predictions. If the frequency standard deviation error remains near 283% or jitter error near 241%, while a plain ANN trained on the same raw samples shows comparable or better agreement, the claimed process-variation awareness from bootstrapping would be refuted.

Watch

Extended reading notes

Core claim

The central claim is that a Kriging bootstrapped artificial neural network metamodel can serve as an accurate, scalable surrogate for SPICE-level statistical simulation of an analog/mixed-signal circuit. The paper demonstrates the claim on a 180 nm CMOS phase-locked loop with 21 design and process parameters, generating four metamodels (power, frequency, locking time, jitter) from a parameterized parasitic RCLK netlist. N Kriging-bootstrapped data points are produced by estimating each Nth point from the other N-1 points, infusing the Kriging correlation structure into the ANN training set. Monte Carlo analysis on the resulting metamodel takes 19 seconds for 1000 runs, versus 468 seconds for pure Kriging and roughly five days for SPICE, at reported root-mean-square errors as low as $10^{-19}$ for jitter. The paper further claims that particle swarm optimization over these metamodels reduces the mean and standard deviation of PLL power while satisfying a locking-time constraint.

Load-bearing premise

The load-bearing premise is that Kriging-generated points, made from the other N-1 sample points, carry enough true SPICE-level information about correlated process variation that training the ANN on them reproduces the SPICE Monte Carlo distribution; the paper offers low RMSE on an unspecified test set as evidence, while Table 3 shows large standard-deviation errors on frequency and jitter.

Editorial extensions

If this is right

  • A 1000-sample Monte Carlo variability analysis of a 21-parameter PLL can be completed in 19 seconds on the Kriging-bootstrapped ANN metamodel, a 24.63x speedup over a pure Kriging metamodel, versus about five days of SPICE simulation.
  • The optimized design flow reduces the mean PLL power from 2.48 mW to 2.35 mW and the power standard deviation from 0.42 mW to 0.39 mW while keeping locking time under constraint.
  • The Kriging-bootstrapped ANN reproduces the standard deviation of power, locking time, and jitter closer to SPICE than a plain ANN does, though frequency and jitter standard deviations still deviate by 282.92% and 240.91% respectively.
  • The bootstrapping step introduces Kriging's correlation awareness into the ANN, so the approach scales to large design spaces where pure Kriging's per-point matrix inversions become expensive.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the bootstrapped ANN's standard-deviation errors against SPICE were confirmed on a held-out test set, the method's practical value would hinge on whether the 19-second runtime holds when the number of process parameters grows well beyond 21; Kriging matrix size grows with sample count, not parameter count, but ANN training cost does grow with input dimension.
  • The 24.63x speedup is measured relative to pure Kriging, not to plain ANN (both take 19s here); the real advantage of bootstrapping over plain ANN must be demonstrated by comparing accuracy on identical held-out Monte Carlo set, which the paper does not report.
  • A natural extension would be to apply the same bootstrapping idea to deep architectures or to hierarchical sub-block metamodels where each PLL subcircuit gets its own Kriging-seeded ANN, then compose the moments; the paper mentions deep learning as future work.
  • The optimization objective $\mu+3\sigma$ is a yield-oriented cost; an immediate testable extension is to verify that the optimized design's actual SPICE yield (within a frequency and locking-time spec window) improves, rather than only the metamodel-predicted moments.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes iVAMS 3.0, a mixed-signal design flow that combines a Kriging-bootstrapped artificial neural network metamodel with particle swarm optimization for variability-aware optimization of a 180 nm CMOS PLL. The central claims are that the Kriging-bootstrapped ANN is 24x faster than a simple ANN metamodel, more accurately captures correlated process variations, and enables effective statistical optimization with reduced mean and standard deviation of figures of merit. The manuscript presents RMSE results, Monte Carlo comparisons against SPICE for Kriging-ANN, Kriging, and ANN metamodels, and a PSO optimization case study.

Significance. If the claims were established, the proposed flow would be a useful contribution to analog/mixed-signal design for cost: it addresses a real bottleneck (slow variability-aware optimization of parasitic-extracted AMS circuits) and the case study is nontrivial (21 design parameters, 1000-run Monte Carlo). The paper has some strengths: it builds the metamodel from a fully extracted RCLK parasitic netlist, it compares against a SPICE Monte Carlo baseline (a non-circular check), and it documents the tool flow in enough detail to be reproduced in principle. However, the headline speed and accuracy advantages over simple ANN metamodeling are directly contradicted or unsupported by the paper's own tables, and the optimization result is not verified in SPICE. As presented, the manuscript does not demonstrate a net benefit over the existing ANN metamodel it is intended to improve.

major comments (4)
  1. [Abstract / Section 2 / Table 4] The headline claim that 'bootstrapped Kriging metamodeling is 24X faster than simple ANN metamodeling' is contradicted by Table 4, which reports 19 s for both Kriging-ANN and ANN and 468 s for Kriging. The 24.63x speedup is relative to Kriging, not to a simple ANN. Because the stated motivation of the hybrid is to combine Kriging accuracy with ANN speed, the absence of any runtime advantage over the plain ANN removes the speed component of the central trade-off claim.
  2. [Table 3 / Section 5] The Kriging-ANN metamodel has larger mean errors than the plain ANN for all four figures of merit (power 3.22% vs 0.81%; frequency 5.64% vs 5.38%; locking time 7.26% vs 5.63%; jitter 10.25% vs 6.61%). Its standard-deviation errors are smaller than ANN's, but they remain very large for frequency (282.92%) and jitter (240.91%) when compared against SPICE. Thus the claim in Section 5 that the resulting models are 'more process aware accurate than the bare ANN models' is not supported by the accuracy data in Table 3.
  3. [Section 5.3 / Table 2] The RMSE values in Table 2 are described as the accuracy of 'Kriging Generated Points', and no independent test set is specified. The bootstrapping procedure generates N points by leave-one-out Kriging on the same N SPICE samples used for ANN training, so low RMSE can reflect smoothing or interpolation of the training data rather than predictive accuracy on unseen SPICE responses. A held-out SPICE test set (or a separate validation set) is needed to support the accuracy claim.
  4. [Table 5 / Section 10.2] The 'After Optimization' column in Table 5 consists of Kriging-ANN metamodel predictions compared with the pre-optimization SPICE baseline; no SPICE Monte Carlo verification of the optimized parameter set is reported. Given the large standard-deviation errors for frequency and jitter in Table 3, the claimed reductions in mean power and in the standard deviations of the figures of merit are unverified and could be artifacts of metamodel prediction bias.
minor comments (4)
  1. [Table 2 / Table 3] Table 2's footer is inconsistent with the text: the caption refers to 'Kriging Generated Points' while Eq. (5) and the surrounding text say RMSE is computed against SPICE results; please clarify what the true responses are in Table 2.
  2. [Section 10 / Reference [36]] The URL in reference [36] contains the typo 'sourcefourge.net' instead of 'sourceforge.net'.
  3. [Section 5.3 / Section 7] The experimental setup does not specify the ANN hidden-layer architecture or the PSO swarm size, iteration count, and inertia/acceleration coefficients; these hyperparameters are needed for reproducibility and for interpreting the reported runtimes.
  4. [Section 9] The phrase 'a Gaussian distribution with 10% standard deviation' is ambiguous; please state whether 10% is the standard deviation relative to the nominal value of each parameter.

Circularity Check

1 steps flagged · score 6.0 of 10

Table 2 validates the Kriging-ANN metamodel on the same Kriging-generated points used for ANN training, making the central accuracy claim partially circular.

  1. fitted input called prediction [Section 5.3 (Kriging Bootstrapped ANN Metamodeling) and Section 10.2 / Table 2]
    "We generate N Kriging bootstrapped data points by using N − 1 points and the Kriging method to estimate the Nth point. N iterations of this process will generate N Kriging bootstrapped data points which are then used for the ANN training. ... Table 2: Statistical Accuracy of Kriging Generated Points [15]. ... The low RMSE values thus demonstrate that the created metamodels are sufficiently accurate and can be used for design exploration."

    The RMSE table is captioned as the accuracy of 'Kriging Generated Points,' and Section 5.3 states that exactly these Kriging-generated points are the targets used for ANN training. Equation 5 defines RMSE with SPICE as the true response, but no independent holdout SPICE set is identified. If the RMSE is evaluated on the Kriging bootstrapped points, then the ANN is being scored on its own training targets, so the low RMSE is forced by the training fit and does not establish accuracy against SPICE. The bootstrapped points themselves are leave-one-out Kriging predictions from the same N SPICE samples, so they add no new circuit-level information. Thus Table 2's support for the central 'accurate metamodel' claim reduces to an in-sample fit.

full rationale

The paper's independent checkpoint is Table 3, which compares Kriging-ANN, Kriging, and ANN Monte Carlo distributions against SPICE Monte Carlo; that comparison is non-circular, and it is the right kind of evidence. However, Table 3 does not favor the hybrid: the Kriging-ANN mean errors are larger than plain ANN for power, frequency, and jitter, and the standard-deviation errors for frequency (282.92%) and jitter (240.91%) are very large. The abstract and Section 2 claim that 'bootstrapped Kriging metamodeling is 24X faster than simple ANN metamodeling,' but Table 4 reports Kriging-ANN and ANN both at 19 s, with the 24.63x speedup measured against Kriging, not against ANN; this is an internal inconsistency rather than a circular reduction. The Kriging implementation is attributed to the authors' prior work, but the method is stated and standard, so self-citation is not load-bearing. Overall, one central accuracy exhibit (Table 2) appears to validate the metamodel on its own Kriging-generated training points, making the accuracy claim partially circular; the remaining SPICE comparison is independent but does not rescue the stated advantage.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests on several unreported experimental choices: the LHS sample count, ANN architecture, PSO coefficients, and Kriging correlation model are all free parameters. The domain assumptions are that the extracted RCLK netlist represents silicon, that random LHS covers the 21-dimensional space, and that a 10% Gaussian sigma on L, W, Vdd, Tox captures the dominant process variation. The most fragile premise is that Kriging-generated points can replace SPICE data in ANN training.

free parameters (6)
  • Number of LHS sample points N = not reported
    Section 5.3 describes generating N sample points and N Kriging bootstrapped points, but N is never given; model accuracy and runtime depend on it.
  • ANN hidden layer architecture = not reported
    Section 5.2 only says a multilayer network with tanh hidden layer and linear output is trained via the MATLAB Neural Network toolbox; the hidden neuron count is omitted.
  • PSO swarm size and iteration count = not reported
    Algorithm 1 defines N and Maxiteration as inputs, but the experiments never state their values.
  • PSO inertia and acceleration coefficients = not reported
    Algorithm 1 line 11 uses w, rho_p, and rho_g; no values are provided.
  • Process variation sigma and parameter ranges = 10% Gaussian sigma; ranges not reported
    Section 9 says a Gaussian distribution with 10% standard deviation is used for L, W, Vdd, Tox; exact ranges and correlations are not specified.
  • Kriging correlation function = not reported
    Section 5.1 presents generic Kriging with covariance r(s,t); the specific correlation model used by mGstat is not stated.
assumptions (5)
  • domain assumption Kriging interpolation provides a valid stochastic predictor of SPICE responses over the design domain.
    Section 5.1 introduces Kriging as a metamodel for circuit responses; the entire bootstrapping step presupposes this.
  • domain assumption The extracted RCLK parasitic netlist is an accurate silicon-level model of the 180nm PLL.
    Sections 4 and 5 state that the fully extracted parasitic netlist ensures silicon-level accuracy; no silicon measurements validate this.
  • domain assumption Random Latin hypercube sampling adequately covers the 21-dimensional design space.
    Section 5.3 states that Random LHS is used; no coverage or convergence analysis is given.
  • domain assumption A Gaussian distribution with 10% sigma on L, W, Vdd, Tox captures the dominant process variation of the PLL.
    Section 9 adopts this distribution based on prior reports, without a sensitivity analysis for this design.
  • ad hoc to paper The ANN trained on Kriging-generated points inherits Kriging's correlation awareness.
    Section 5.3 hypothesizes this benefit; Table 3 provides weak and inconsistent support (better sigma than ANN, worse mean).

how reviews work

0 comments
Cite this review

Pith. "Pith review of iVAMS 3.0: Hierarchical-Machine-Learning-Metamodel-Integrated Intelligent Verilog-AMS for Ultra-Fast, Accurate Mixed-Signal Design Optimization." pith.science (2026). https://pith.science/paper/RK57NXI4

@misc{pith2026250601045,
  author       = {Pith},
  title        = {Pith review of: iVAMS 3.0: Hierarchical-Machine-Learning-Metamodel-Integrated Intelligent Verilog-AMS for Ultra-Fast, Accurate Mixed-Signal Design Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RK57NXI4}},
  note         = {Machine review of arXiv:2506.01045}
}
read the original abstract

Analog/Mixed-Signal (AMS) circuits and systems continually present significant challenges to designers with the increase of design complexity and aggressive technology scaling. This is due to the large number of design factors and parameters that must be taken into account as well as the process variations which are prominent in nano-CMOS circuits. Design optimization techniques while presenting an accurate and fast design flow which can perform design optimization in reasonable time are still lacking. Even with techniques such as metamodeling that aid the design phase, there is still the need to improve them for accuracy and time cost. As a trade-off of the accuracy and speed, this paper presents a design flow for ultra-fast variability-aware optimization of nano-CMOS based physical design of analog circuits. It combines a Kriging bootstrapped Artificial Neural Network (ANN) metamodel with a Particle Swarm Optimization (PSO) based algorithm in the design optimization flow. The Kriging bootstrapped ANN metamodel provides a trade-off between analog-quality accuracy and scalability and can be effectively used for large and complex AMS circuits. The proposed technique uses Kriging to bootstrap target samples used for the ANN training. This introduces Kriging characteristics, which account for correlation effects between design parameters, to the ANN. The effectiveness of the design flow is demonstrated using a PLL as a case study with as many as 21 design parameters. It is observed that the bootstrapped Kriging metamodeling is 24X faster than simple ANN metamodeling. The layout optimization for such a complex circuit can be performed effectively in a short time using this approach. The optimization flow could achieve significant reductions in the mean and standard deviation of the PLL characteristics. Thus, the proposed research is a major contribution to design for cost.

Figures

Figures reproduced from arXiv: 2506.01045 by the authors.

Figure 1
Figure 1. Techniques for fast design exploration. An automated creation technique for surrogate multivariate mathematical models by using CAD-Model construction for microwave components is proposed and tested in [6]. The technique uses mathematical formulas for the model estima￾tion and optimization and does not use sampling techniques. IP reuse for SoC interaction and microprocessor design is covered in [7]. This approach us… view at source ↗
Figure 2
Figure 2. Fast design space exploration of analog circuits thr 1: Fast design space exploration of analog circuits through acc [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 2
Figure 2. Proposed high level design flow with Hierarchical Machine Learning Modeling [25]. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figures from the paper (12 more)
Figure 3
Figure 3. Figure 3: Proposed design optimization flow [25]. SSaS maa mp mple plele3 N Data 2 1 f f ANN Metamodel Kriging Metamodel f f Kriging/ANN Metamodel ANN Training ANN Training SSaS maa mp mple plele3 N Sample 1 2 DDat aata n-1 Data 3 Data 2 1 DDat aata n-1 Da Dtaata 2 3 1 ... Krigi…
Figure 4
Figure 4. Figure 4: Proposed Kriging Bootstrapped ANN Metamodel Generation Flow [15]. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Proposed Metamodel Design Flow [15, 25]. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Flow diagram for the PSO algorithm [25]. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: High level system diagram for the PLL. The reference clock feeds the input signal to the phase detector, which compares and detects the phase difference between the input signal and the output from the VCO. The charge pump generates a supply charge in proportion to the…
Figure 8
Figure 8. Figure 8: Physical design of the 180 nm PLL [15] [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Statistical Variation Analysis 11 [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Experimental Steps and EDA/Non-EDA Tool Interactions. [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: Comparative Results with Kriging and Neural Network [15]. [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]
Figure 12
Figure 12. Figure 12: Statistical Analysis of FoMs using Kriging Bootstrapped Trained Neural Network based metamodeling [PITH_FULL_IMAGE:figures/full_fig_p014_12.png]
Figure 13
Figure 13. Figure 13: Statistical Analysis of FoMs using Kriging based metamodeling [15]. [PITH_FULL_IMAGE:figures/full_fig_p014_13.png]
Figure 14
Figure 14. Figure 14: Statistical Analysis of FoMs using Neural Network based metamodeling [15]. [PITH_FULL_IMAGE:figures/full_fig_p015_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

  1. [1]

    Fast Design Optimization through Simple Kriging Metamodeling: A Sense Amplifier Case Study,

    O. Okobiah, S. Mohanty, and E. Kougianos, “Fast Design Optimization through Simple Kriging Metamodeling: A Sense Amplifier Case Study,” IEEE Transactions on Very Large Scale Integration (VLSI) Systems , vol. 22, 16 iV AMS 3.0: Hierarchical-Machine-Learning-Metamodel-Integrated Intelligent Verilog-AMS for Ultra-Fast, Accurate Mixed-Signal Design Optimizati...

    work page 2014

  2. [2]

    A Comparative Study of Metamodels for Fast and Accurate Simulation of Nano-CMOS Circuits,

    O. Garitselov, S. P. Mohanty, and E. Kougianos, “A Comparative Study of Metamodels for Fast and Accurate Simulation of Nano-CMOS Circuits,” IEEE Transactions on Semiconductor Manufacturing, vol. 25, no. 1, pp. 26–36, Feb 2012

    work page 2012

  3. [3]

    A Comparative Study of Metamodels for Fast and Accurate Simulation of Nano-CMOS Circuits,

    O. Garitselov, S. Mohanty, and E. Kougianos, “A Comparative Study of Metamodels for Fast and Accurate Simulation of Nano-CMOS Circuits,” IEEE Transactions on Semiconductor Manufacturing, vol. 25, no. 1, pp. 26–36, 2012

    work page 2012

  4. [4]

    Towards Robust Nano-CMOS Sense Amplifier Design: A Dual-Threshold versus Dual-Oxide Perspective,

    O. Okobiah, S. P. Mohanty, E. Kougianos, and M. Poolakkaparambil, “Towards Robust Nano-CMOS Sense Amplifier Design: A Dual-Threshold versus Dual-Oxide Perspective,” in Proceedings of the 21st ACM Great Lakes Symposium on VLSI, 2011, pp. 145–150

    work page 2011

  5. [5]

    Kriging Metamodeling in Discrete-Event Simulation: An Overview,

    W. Van Beers, “Kriging Metamodeling in Discrete-Event Simulation: An Overview,” in Proceedings of the Winter Simulation Conference, 2005, pp. 202–208

    work page 2005

  6. [6]

    Yield-Aware Analog Integrated Circuit Optimization Using Geostatistics Motivated Perfor- mance Modeling,

    G. Yu and P. Li, “Yield-Aware Analog Integrated Circuit Optimization Using Geostatistics Motivated Perfor- mance Modeling,” in Computer-Aided Design, 2007. ICCAD 2007. IEEE/ACM International Conference on , Nov. 2007, pp. 464–469

    work page 2007

  7. [7]

    Kriging-Assisted Ultra-Fast Simulated-Annealing Optimization of a Clamped Bitline Sense Amplifier,

    O. Okobiah, S. P. Mohanty, E. Kougianos, and O. Garitselov, “Kriging-Assisted Ultra-Fast Simulated-Annealing Optimization of a Clamped Bitline Sense Amplifier,” VLSI Design, International Conference on , vol. 0, pp. 310–315, 2012

    work page 2012

  8. [8]

    Kriging Model Combined with Latin Hypercube Sampling for Surrogate Modeling of Analog Integrated Circuit Performance,

    H. You, M. Yang, D. Wang, and X. Jia, “Kriging Model Combined with Latin Hypercube Sampling for Surrogate Modeling of Analog Integrated Circuit Performance,” inProceedings of the International Symposium on Quality of Electronic Design, 2009, pp. 554–558

    work page 2009

Show all 41 references
  1. [9]

    Advances in Particle Swarm Optimization for Antenna Designs: Real-Number, Bi- nary, Single-Objective and Multiobjective Implementations,

    N. Jin and Y . Rahmat-Samii, “Advances in Particle Swarm Optimization for Antenna Designs: Real-Number, Bi- nary, Single-Objective and Multiobjective Implementations,” IEEE Transactions on Antennas and Propagation, vol. 55, no. 3, pp. 556–567, 2007

    2007

  2. [10]

    Handling Multiple Objectives With Particle Swarm Opti- mization,

    C. A. C. Coello, G. T. Pulido, and M. S. Lechuga, “Handling Multiple Objectives With Particle Swarm Opti- mization,” IEEE Transactions on Evolutionary Computation, vol. 8, no. 3, pp. 256–279, 2004

    2004

  3. [11]

    Using Particle Swarm Optimization in Training Neural Network for Indoor Field Strength Prediction,

    I. Vilovi ´c, N. Burum, and D. Mili´c, “Using Particle Swarm Optimization in Training Neural Network for Indoor Field Strength Prediction,” in 51st International Symposium ELMAR, 2009, pp. 275–278

    2009

  4. [12]

    Fast Statistical Analysis of Process Variation Effects Using Accurate PLL Behavioral Models,

    C.-C. Kuo, M.-J. Lee, C.-N. Liu, and C.-J. Huang, “Fast Statistical Analysis of Process Variation Effects Using Accurate PLL Behavioral Models,”IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 56, no. 6, pp. 1160–1172, June 2008

    2008

  5. [13]

    [Online]

    (2023) What Are Digital Twins? A Primer on Virtual Models. [Online]. Available: https://www.synopsys.com/ blogs/chip-design/digital-twins-semiconductor-industry.html

    2023

  6. [14]

    Exploring Kriging for Fast and Accurate Design Optimization of Nanoscale Analog Circuits,

    O. Okobiah, S. P. Mohanty, and E. Kougianos, “Exploring Kriging for Fast and Accurate Design Optimization of Nanoscale Analog Circuits,” in Proceedings of the 13th IEEE Computer Society Annual Symposium on VLSI (ISVLSI), 2014, pp. 244–247

    2014

  7. [15]

    Kriging Bootstrapped Neural Network Training for Fast and Accurate Process Variation Analysis,

    O. Okobiah, S. P. Mohanty, and E. Kougianos , “Kriging Bootstrapped Neural Network Training for Fast and Accurate Process Variation Analysis,” in Proceedings of the 15th IEEE International Symposium on Quality Electronic Design (ISQED), 2014, pp. 365–372

    2014

  8. [16]

    Kriging Metamodeling in Multi-objective Simulation Optimization,

    M. Zakerifar, W. Biles, and G. Evans, “Kriging Metamodeling in Multi-objective Simulation Optimization,” in Proceedings of the Winter Simulation Conference (WSC), 2009, pp. 2115–2122

    2009

  9. [17]

    Kriging Metamodeling in Con- strained Simulation Optimization: An Explorative Study,

    W. E. Biles, J. P. C. Kleijnen, W. C. M. van Beers, and I. van Nieuwenhuyse, “Kriging Metamodeling in Con- strained Simulation Optimization: An Explorative Study,” in Proceedings of the 39th Winter Simulation Confer- ence, 2007, pp. 355–362

    2007

  10. [18]

    Robust Simulation-Optimization using Metamodels,

    G. Dellino, J. Kleijnen, and C. Meloni, “Robust Simulation-Optimization using Metamodels,” in Proceedings of the Winter Simulation Conference (WSC), Dec. 2009, pp. 540–550

    2009

  11. [19]

    Stochastic Kriging for Simulation Metamodeling,

    B. Ankenman, B. Nelson, and J. Staum, “Stochastic Kriging for Simulation Metamodeling,” in Proceedings of the Winter Simulation Conference, 2008, pp. 362–370

    2008

  12. [20]

    Analog Circuit Optimization using Evolutionary Algorithms and Convex Optimization,

    V . Aggarwal, “Analog Circuit Optimization using Evolutionary Algorithms and Convex Optimization,” Master’s thesis, Massachusetts Institute of Technology, May 2007

    2007

  13. [21]

    A Hybrid Genetic Algorithm- Neural Network Strategy for Simulation Optimization,

    L. Wang, “A Hybrid Genetic Algorithm- Neural Network Strategy for Simulation Optimization,” Applied Math- ematics and Computation, vol. 170, no. 2, pp. 1329–1343, 2005. 17 iV AMS 3.0: Hierarchical-Machine-Learning-Metamodel-Integrated Intelligent Verilog-AMS for Ultra-Fast, Ac...

    2005

  14. [22]

    Developing Optimal Neural Network Metamodels Based on Prediction Intervals,

    A. Khosravi, S. Nahavandi, and D. Creighton, “Developing Optimal Neural Network Metamodels Based on Prediction Intervals,” inProceedings of the International Joint Conference on Neural Networks, 2009, pp. 1583– 1589

    2009

  15. [23]

    Neural Network-based Simulation Metamodels for Predicting Probability Dis- tributions,

    C. W. Zobel and K. B. Keeling, “Neural Network-based Simulation Metamodels for Predicting Probability Dis- tributions,” Computers and Industrial Engineering, vol. 54, pp. 879–888, May 2008

    2008

  16. [24]

    Simulation Metamodelling With Neural Networks: An Experimental Investi- gation

    I. Sabuncuoglu and S. Touhami, “Simulation Metamodelling With Neural Networks: An Experimental Investi- gation.” International Journal of Production Research,, vol. 40, no. 11, pp. 2483–2505, 2002

    2002

  17. [25]

    Ultra-Fast Variability-Aware Optimization of Mixed- signal Designs Using Bootstrapped Kriging,

    S. P. Mohanty, E. Kougianos, and V . P. Yanambaka, “Ultra-Fast Variability-Aware Optimization of Mixed- signal Designs Using Bootstrapped Kriging,” inSixteenth International Symposium on Quality Electronic Design ISQED, 2015, pp. 239–242

    2015

  18. [26]

    Kriging,

    G. Bohling, “Kriging,” Kansas Geological Survey, Tech. Rep., 2005

    2005

  19. [27]

    Particle Swarm Optimization over Non-Polynomial Metamodels for Fast Process Variation Resilient Design of Nano-CMOS PLL,

    O. Garitselov, S. Mohanty, E. Kougianos, and G. Zheng, “Particle Swarm Optimization over Non-Polynomial Metamodels for Fast Process Variation Resilient Design of Nano-CMOS PLL,” inProceedings of the great lakes symposium on VLSI, ser. GLSVLSI ’12, 2012, pp. 255–258

    2012

  20. [28]

    Natick, Massachusetts, United States: The MathWorks Inc., 2012

    MATLAB, MATLAB and Neural Network Toolbox Release 2012b. Natick, Massachusetts, United States: The MathWorks Inc., 2012

    2012

  21. [29]

    Analysis of the Publications on the Applications of Particle Swarm Optimisation,

    R. Poli, “Analysis of the Publications on the Applications of Particle Swarm Optimisation,” Journal of Artificial Evolution and Applications, vol. 2008, pp. 4:1–4:10, January 2008

    2008

  22. [30]

    Transforming Geocentric Cartesian Coordinates to Geodetic Coordinates by Using Differential Search Algorithm,

    P. Civicioglu, “Transforming Geocentric Cartesian Coordinates to Geodetic Coordinates by Using Differential Search Algorithm,” Computers and Geosciences, vol. 46, no. 0, pp. 229–247, 2012

    2012

  23. [31]

    Ant Colony Optimization – Artificial Ants as a Computational Intelli- gence Technique,

    M. Dorigo, M. Birattari, and T. Stutzle, “Ant Colony Optimization – Artificial Ants as a Computational Intelli- gence Technique,”IEEE Computational Intelligence Magazine, vol. 1, pp. 28–39, 2006

    2006

  24. [32]

    Application of the PSO technique to the Optimization of CMOS Operational Transconductance Amplifiers,

    S. Bennour, A. Sallem, M. Kotti, E. Gaddour, M. Fakhfakh, and M. Loulou, “Application of the PSO technique to the Optimization of CMOS Operational Transconductance Amplifiers,” inProceedings of the 5th International Conference on Design and Technology of Integrated Systems in ...

    2010

  25. [33]

    Particle Swarm Optimization over Non-Polynomial Metamodels for Fast Process Variation Resilient Design of Nano-CMOS PLL,

    O. Garitselov, S. Mohanty, E. Kougianos, and G. Zheng, “Particle Swarm Optimization over Non-Polynomial Metamodels for Fast Process Variation Resilient Design of Nano-CMOS PLL,” inProceedings of the great lakes symposium on VLSI, 2012, pp. 255–258

    2012

  26. [34]

    Statistical Timing Analysis using Levelized Covariance Propagation,

    K. Kang, B. Paul, and K. Roy, “Statistical Timing Analysis using Levelized Covariance Propagation,” inDesign, Automation and Test in Europe, 2005. Proceedings, vol. 2, 2005, pp. 764–769

    2005

  27. [35]

    Modeling and analysis of manufacturing variations,

    S. Nassif, “Modeling and analysis of manufacturing variations,” in Custom Integrated Circuits, 2001, IEEE Conference on., 2001, pp. 223–228

    2001

  28. [36]

    [Online]

    mGstat: A Geostatistical Matlab Toolbox , last accessed on 08 Apr 2025. [Online]. Available: mgstat.sourcefourge.net

    2025

  29. [37]

    A Surrogate Modeling and Adaptive Sampling Toolbox for Computer Based Design,

    D. Gorissen, I. Couckuyt, P. Demeester, T. Dhaene, and K. Crombecq, “A Surrogate Modeling and Adaptive Sampling Toolbox for Computer Based Design,”J. Mach. Learn. Res., vol. 11, pp. 2051–2055, August 2010

    2010

  30. [38]

    Accurate Polynomial Metamodeling-Based Ultra-Fast Bee Colony Optimization of a Nano-CMOS Phase-Locked Loop,

    O. Garitselov, S. Mohanty, and E. Kougianos, “Accurate Polynomial Metamodeling-Based Ultra-Fast Bee Colony Optimization of a Nano-CMOS Phase-Locked Loop,” ASP Journal of Low Power Electronics (JOLPE), vol. 8, no. 3, pp. 317–328, June 2012

    2012

  31. [39]

    Ordinary Kriging Metamodel-Assisted Ant Colony Algorithm for Fast Analog Design Optimization,

    O. Okobiah, S. Mohanty, and E. Kougianos, “Ordinary Kriging Metamodel-Assisted Ant Colony Algorithm for Fast Analog Design Optimization,” in Proceedings of the 13th International Symposium on Quality Electronic Design (ISQED), March 2012, pp. 458–463

    2012

  32. [40]

    iV AMS 1.0: Polynomial-Metamodel-Integrated Intelligent Verilog-AMS for Fast, Accurate Mixed-Signal Design Optimization,

    S. P. Mohanty and E. Kougianos, “iV AMS 1.0: Polynomial-Metamodel-Integrated Intelligent Verilog-AMS for Fast, Accurate Mixed-Signal Design Optimization,” arXiv Computer Science , vol. abs/1905.12812, 2019. [Online]. Available: http://arxiv.org/abs/1905.12812

    1905 arXiv

  33. [41]

    iV AMS 2.0: Machine-Learning-Metamodel-Integrated Intelligent Verilog-AMS for Fast and Accurate Mixed-Signal Design Optimization,

    S. P. Mohanty and E. Kougianos , “iV AMS 2.0: Machine-Learning-Metamodel-Integrated Intelligent Verilog-AMS for Fast and Accurate Mixed-Signal Design Optimization,” arXiv Electrical Engineering and Systems Science, vol. abs/1907.01526, 2019. [Online]. Available: http://arxiv.o...

    1907 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.