Pith. sign in

REVIEW 4 major objections 4 minor 28 references

Enhancing Field-Oriented Control of Electric Drives with Tiny Neural Network Optimized for Micro-controllers

T0 review · 4 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read A 1,400-parameter feed-forward network, TinyFC, added to a PI speed loop learns to correct the quadrature-current reference, cutting overshoot by up to 87.5% in simulation while keeping the model deployable on a micro-controller.

desk verdict The paper's own timing table sinks its central MCU claim, but the reproducible workflow and measured latency numbers are worth a second look if reframed. read the letter →

arxiv 2502.00532 v1 pith:AMGEEUPV submitted 2025-02-01 cs.LG cs.SYeess.SY

Martin Joel Mouk Elele , Danilo Pau , Shixin Zhuang , Tullio Facchinetti This is my paper · ORCID
classification cs.LGcs.SYeess.SY
keywords tinyneuralnetworksfield-orientedcontrolPMSMPIcontrollerovershootreductionmicrocontrollerdeploymentnetworkpruningquantization
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 tries to establish that a tiny feed-forward network, TinyFC, can act as a corrective add-on to a classical PI controller in field-oriented motor control. In simulation, the 1,400-parameter network learns to repair the PI controller's quadrature-current reference, cutting maximum speed overshoot by up to 87.5% in one test case and 68% in another, with a pruned version eliminating overshoot entirely in the first case. The appeal of the claim is that a network small enough to fit a micro-controller can fix a known weakness of simple PI control without the latency and complexity of heavier alternatives such as model-predictive control. The paper also reports that quantizing the pruned model to 8-bit integers keeps its weights near 1 KiB, the scale needed for embedded deployment.

What carries the argument

The load-bearing object is TinyFC, a feed-forward network with two branches of five fully connected layers plus residual connections, whose output is a current compensation $\Delta i_q$ constrained to $[-1,1]$ by a $\tanh$ activation and added to the PI controller's quadrature-current reference. Its training signal, defined by Eq. (3), is the difference between the raw PI current and a hand-adjusted ideal version: saturation at a chosen threshold $C$ for test case 1, exponential rectification with a chosen time constant $\tau$ for test case 2. The network learns this correction from a 300,001-sample-per-test-case dataset generated by the Simulink FOC model. This target-construction step is what makes the network learn to remove overshoot, and it is the fragile part of the pipeline.

What would settle it

Run the two speed profiles on a physical PMSM drive with the same motor and the NUCLEO-G474RE board, with the pruned and quantized TinyFC inserted in the loop; if the 87.5% or complete overshoot reduction does not reproduce, or the motor current becomes unstable, the central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that a small fully connected network, TinyFC, can be trained to output the quadrature-current correction $\Delta i_q$ that compensates for what a PI speed controller gets wrong, and that adding this correction to the PI reference current inside a field-oriented PMSM drive reduces overshoot and tracking error. The network takes reference speed, measured speed, and the PI-predicted quadrature current as inputs. Its training target is constructed by taking the PI's own quadrature-current signal and removing the sections the authors identify as overshoots, either by saturating at a threshold $C$ or by exponentially rectifying each interval with time constant $\tau$; the network learns the difference between the raw and the adjusted signals. In the two simulation test cases, the augmented loop reduces maximum overshoot by 87.5% and 68%, and the PCA-pruned version eliminates overshoot completely in the first case.

Load-bearing premise

The load-bearing premise is that the hand-edited quadrature-current signal, saturated at a chosen threshold $C$ or exponentially rectified with a chosen time constant $\tau$, is the ideal response the motor should follow; if that edited signal is not physically sound, the measured overshoot reduction is an artifact of the target rather than evidence that TinyFC improves control.

Editorial extensions

If this is right

  • In the simulated drive, adding TinyFC cuts maximum overshoot by 87.5% in test case 1 and 68% in test case 2, while average deviation falls by 60% and 16.7%.
  • The PCA-pruned TinyFC eliminates overshoot completely in test case 1 and lowers it to 0.03 in test case 2, with only a small increase in maximum deviation.
  • Quantization to 8-bit integers brings the pruned model's weight storage down to 1.37 KiB in case 1 and 1.07 KiB in case 2, within the flash budget of the NUCLEO-G474RE board.
  • Because the network only adds a correction to the PI reference, the PI controller remains in place; the paper explicitly describes TinyFC as supportive rather than a replacement.
  • MSE during training does not predict loop performance: the hyperparameter-optimized model with fewer parameters and low MSE raised overshoot by up to 100% in case 2, so control-loop metrics must be part of model selection.

Reading between the lines

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

  • If the hand-edited target is replaced by a signal from an established reference controller, the same pipeline could produce a TinyFC teacher without hand-chosen thresholds or time constants.
  • A loss function that penalizes overshoot and current excursions directly might fix the failure of MSE-based model selection; the paper points toward physics-informed networks but does not test them.
  • The reported MCU inference times (127.6 to 371.7 microseconds) are above the 33.33 microsecond PWM period, so whether the optimized models can actually run inside the real-time loop remains an open hardware-in-the-loop question.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 TinyFC, a 1,400-parameter feedforward neural network inserted into the speed loop of PI-based field-oriented control for a PMSM. The network takes reference speed, measured speed, and the PI quadrature-current reference as inputs and outputs a compensation value Δi_q that is added to the PI reference. Training data come from Simulink simulations of two custom speed profiles; the target Δi_q^GT is defined as the difference between the PI current reference and a manually adjusted version obtained by saturation (Eq. 1) or exponential rectification (Eq. 2). The authors compare the baseline, HPO, and pruned variants, all quantized, in terms of MSE, speed deviation and overshoot, and deployed inference time on a NUCLEO-G474RE. They report up to 87.5% overshoot reduction with TinyFC and about 100% reduction with the pruned model in test case 1, but they also report inference times from 127.6 to 371.7 µs, all exceeding the 33.33 µs period of the 30 kHz control loop.

Significance. If the central claims were supported, the paper would be a relevant demonstration of a very small neural compensator for motor control, with strengths including actual MCU inference measurements through ST Edge AI, a comparison of several compression techniques, and a public code/data repository. The paper is also honest in acknowledging that MCU timing remains a challenge. However, two load-bearing issues undermine the main claims: the training target is constructed from the PI controller's own current signal by manual overshoot removal, making the measured overshoot reduction partly a definitional artifact; and the measured inference times exceed the control period by factors of roughly 4 to 11, so the closed-loop simulation results in Tables 3 and 4 implicitly assume zero-latency compensation. These issues affect the core message of a microcontroller-optimized network that improves FOC, so the manuscript as it stands does not support its conclusions.

major comments (4)
  1. [Section 4, Eqs. (1)-(3)] The network is trained to predict Δi_q^GT = x(t) − x_adj(t), where x_adj(t) is the PI reference current after manual overshoot removal with hand-chosen parameters C and τ. Adding this prediction to the PI reference via Eq. (4) means that, by construction, the loop tracks the hand-edited reference whenever the network fits its target. The overshoot reductions in Tables 3 and 4 therefore do not validate that the network improves motor control; they mainly show that the network can reproduce a manually edited current signal. An independent validation target based on physical control criteria, such as closed-loop stability, actuator limits, or disturbance rejection, is needed before the result can be attributed to the neural compensator.
  2. [Section 5, first paragraph, and Section 6.3, Table 5] The paper correctly states that the 30 kHz PWM requires inputs every 33.33 µs and that exceeding this time causes delays and performance issues. Yet every inference time reported in Table 5 exceeds this bound: TinyFC 207.4 µs, HPO 144.8 and 127.6 µs, pruned 232.2 and 215.2 µs, and quantized-pruned 371.7 and 361.4 µs. The Simulink results in Tables 3 and 4 do not model this measured latency, so the reported overshoot improvements assume the compensation is applied instantaneously. This directly contradicts the abstract's claim that the model fits the computational constraints of a microcontroller, and it is consistent with the paper's own concluding admission that 'optimizing TinyFC inference for MCU timing remains a challenge.'
  3. [Section 6.1 and Table 4] The HPO model, one of the three proposed optimization outcomes, performs worse than the PI baseline in test case 2: Table 4 reports a maximum overshoot of 1.24 for HPO, versus 0.25 for PI in Table 3. The paper states that 'the HPO model did not perform satisfactorily within the FOC loop during this study (this case is not shown in this paper)' and then proceeds without further analysis. As a result, the claim that hyperparameter optimization is a successful route in the proposed workflow is unsupported, and the trade-off discussion among the variants is incomplete.
  4. [Section 4, Eq. (1)] Equation (1) defines x_adj(t) = min{|x(t)|, C}, which is always nonnegative. If x(t) denotes the quadrature-current reference, this definition cannot represent a valid reference during deceleration or negative-torque operation, because the sign information is discarded. Even if the test signals used here happen to avoid negative values, the equation as written is not a general saturation operator and should be replaced by a sign-preserving saturation or explicitly restricted to the positive-current case.
minor comments (4)
  1. [Section 4, Eqs. (1)-(3)] The notation x(t) is used inconsistently: it sometimes denotes the measured speed, sometimes the quadrature current, and sometimes the 'response signal.' Please define each symbol explicitly at first use.
  2. [Section 3.2] The test cases are described only qualitatively as '2 transitions per second' and '10 speed transitions per second.' The exact reference-speed sequences should be given numerically or as a data file so that the experiments are reproducible.
  3. [Section 5.3] The quantization description omits implementation details such as per-tensor versus per-channel quantization, the calibration dataset, and the integer runtime used by ST Edge AI. These details are necessary to reproduce the reported memory and timing numbers.
  4. [Section 6.3, Table 5] The table lists 'Act' and 'Lib' memory components, but the meaning of these entries and their units are not explained in the text; please clarify.

Circularity Check

2 steps flagged · score 7.0 of 10

Overshoot reduction is built into the training target: TinyFC is trained to output the hand-edited PI-current correction, so the reported improvement is the NN reproducing that edit, not an independent prediction.

  1. fitted input called prediction [Section 4, Eqs. (1)-(3)]
    "In test case 1, ... This range was defined using a threshold C, representing the maximum acceptable value within the response interval [3]. The modified signal was thus determined by the following equation: x_adj(t) = min{|x(t)|, C} (1) ... As a result of the previous considerations, the signal used as ground truth for training the TinyFC NN is Δi^GT_q defined as: Δi^GT_q = x(t) − x_adj(t) (3)"

    The ground-truth compensation is constructed as the difference between the PI current x(t) and a manually saturated/rectified version x_adj(t). The thresholds C and τ are hand-picked by inspecting the very speed overshoots and instability that the paper later reports as reduced (Tables 3-4). Training TinyFC to minimize MSE against Δi^GT_q means the network learns to output this hand-fitted correction. When the same test-case signals are replayed in closed loop, the reported overshoot reduction is therefore the reproduction of the editing rule, not an independent control-law prediction. The NN contributes function approximation of the handcrafted repair, not a new result about FOC.

  2. self definitional [Section 5, Eq. (4), combined with Section 4, Eq. (3); results in Section 6.2, Table 3]
    "The output of the TinyFC NN consists in the adjustment value Δi_q for the PI's prediction that is used to correct the reference signal i^{ref(PI)}_q based on the following equation: i^adj_q = i^{ref(PI)}_q + Δi_q^{TinyFC} (4)"

    By Eq. (3), the training target is x(t) − x_adj(t). If the NN approximates its target, then Eq. (4) gives i^adj_q ≈ x_adj(t): the actual current command in the 'augmented' FOC is, by construction, the manually repaired current. The measured overshoot reduction (up to 87.5%, and 'complete overshoot elimination' for the pruned model) is therefore the direct consequence of substituting the hand-edited signal into the loop. The claim that TinyFC 'detects the error in the PI's prediction and corrects it' is a restatement of how the target was defined; the paper does not validate the edited signal as an optimal or physically grounded reference independent of the test responses.

full rationale

The paper is not circular in the sense of load-bearing self-citation: references [16,17] to the authors' prior work are used to set up the simulation platform, not to justify the control improvement. The central derivation chain, however, reduces to construction: the NN input-output map is trained to reproduce a hand-crafted correction of the PI current, and the closed-loop evaluation measures the performance of that corrected current. The reported overshoot reductions are thus largely forced by the way the ground truth was built; the only genuinely empirical content is the network's ability to fit and generalize (or not, as in the HPO case) the manually edited signal. MCU latency issues (Section 5/Table 5) are a correctness/deployability concern, not circularity. Score 7 reflects that the central claimed improvement is mostly an artifact of the training-target construction rather than an independent control result.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities. Its load-bearing free choices are the hand-constructed training target (threshold C and decay tau) and the simulation model, both of which directly determine the reported overshoot reduction.

free parameters (3)
  • C (saturation threshold) = not reported
    Hand-chosen cap in Eq (1) used to build the adjusted training target x_adj(t) for test case 1; it defines what counts as an acceptable quadrature current and directly shapes the network's target.
  • tau (exponential decay constant) = not reported
    Hand-chosen time constant in Eq (2) used to build the rectified training target for test case 2; it controls how quickly the corrected current converges to the in-range value and is not reported or varied.
  • In-range detection window = not reported
    In test case 2, the paper identifies intervals where measured speed matches reference speed to pick initial and final in-range quadrature current values; the criterion for 'matches' is not defined numerically.
assumptions (4)
  • domain assumption Simulink PMSM model fidelity
    The entire evaluation, including the training data and the closed-loop performance metrics, is based on a Simulink model of the BR2804-1700KV-1 motor. No hardware-in-the-loop or physical measurement validates the simulation.
  • domain assumption PI controller baseline is a fair representative
    The comparison baseline is a specific PI-based FOC whose tuning gains are not reported; the claim of NN improvement is relative to this particular PI configuration.
  • domain assumption 30 kHz PWM timing budget
    The paper assumes the control loop must produce new outputs every 33.33 us (T_s = 3.3333e-5 s) and uses this to judge deployability; this is a standard assumption for the cited PWM frequency but is an external constraint, not derived.
  • standard math Universal approximation capability of feed-forward networks
    Section 2 relies on the universal approximation theorem (Hornik et al.) to justify that a feed-forward network can represent the required correction; this is a standard mathematical result but not verified for this specific architecture.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Field-Oriented Control of Electric Drives with Tiny Neural Network Optimized for Micro-controllers." pith.science (2026). https://pith.science/paper/AMGEEUPV

@misc{pith2026250200532,
  author       = {Pith},
  title        = {Pith review of: Enhancing Field-Oriented Control of Electric Drives with Tiny Neural Network Optimized for Micro-controllers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AMGEEUPV}},
  note         = {Machine review of arXiv:2502.00532}
}
read the original abstract

The deployment of neural networks on resource-constrained micro-controllers has gained momentum, driving many advancements in Tiny Neural Networks. This paper introduces a tiny feed-forward neural network, TinyFC, integrated into the Field-Oriented Control (FOC) of Permanent Magnet Synchronous Motors (PMSMs). Proportional-Integral (PI) controllers are widely used in FOC for their simplicity, although their limitations in handling nonlinear dynamics hinder precision. To address this issue, a lightweight 1,400 parameters TinyFC was devised to enhance the FOC performance while fitting into the computational and memory constraints of a micro-controller. Advanced optimization techniques, including pruning, hyperparameter tuning, and quantization to 8-bit integers, were applied to reduce the model's footprint while preserving the network effectiveness. Simulation results show the proposed approach significantly reduced overshoot by up to 87.5%, with the pruned model achieving complete overshoot elimination, highlighting the potential of tiny neural networks in real-time motor control applications.

Figures

Figures reproduced from arXiv: 2502.00532 by the authors.

Figure 1
Figure 1. Workflow diagram to deploy NN-augmented FOC [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. System model of input (in green), control (in blue) [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 4
Figure 4. Quadrature current from PI-controller simulation [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗
Figures from the paper (4 more)
Figure 3
Figure 3. Figure 3: Measured speed collected from PI-based FOC [PITH_FULL_IMAGE:figures/full_fig_p003_3.png]
Figure 5
Figure 5. Figure 5: Quadrature current adjusted based on capping and [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 6
Figure 6. Figure 6: Network model topology as shown by MATLAB [PITH_FULL_IMAGE:figures/full_fig_p004_6.png]
Figure 7
Figure 7. Figure 7: Output of pruned NN of test case 1 performance of the FOC, with ∼100% reduction in the overshoot (test case 1). The pruned model considerably reduces the overshoot in test case 2, despite slightly increasing the deviation from the reference signal in both cases. Despit…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 27 canonical work pages

  1. [1]

    A comprehensive survey on tinyml

    Youssef Abadade, Anas Temouden, Hatim Bamoumen, Nabil Benamar, Yousra Chtouki, and Abdelhakim Senhaji Hafid. A comprehensive survey on tinyml. IEEE Access, 11:96892–96922, 2023

    work page 2023

  2. [2]

    Modeling of the forward wave propagation using physics-informed neural networks

    Shaikhah Alkhadhr, Xilun Liu, and Mohamed Almekkawy. Modeling of the forward wave propagation using physics-informed neural networks. In 2021 IEEE International Ultrasonics Symposium (IUS) , pages 1–4, 2021

    work page 2021

  3. [3]

    J. Allison. The linear rectification of two amplitude-modulated signals and noise. IEEE Transactions on Information Theory , 13(1):69–75, 1967

    work page 1967

  4. [4]

    Atkins and M

    A.G. Atkins and M. Escudier. A Dictionary of Mechanical Engineering . Oxford Paperback Reference. OUP Oxford, 2013

    work page 2013

  5. [5]

    Developments in the Control Loops Bench- marking

    Grzegorz Bialic and Marian B??achuta. Developments in the Control Loops Bench- marking. 05 2008

    work page 2008

  6. [6]

    Optimal selection of pi parameters of foc for pmsm using structured h-infinity-synthesis

    Runze Cai, Ruixiang Zheng, Ming Liu, and Mian Li. Optimal selection of pi parameters of foc for pmsm using structured h-infinity-synthesis. In IECON 2017 - 43rd Annual Conference of the IEEE Industrial Electronics Society , pages 8602–8607, 2017

    work page 2017

  7. [7]

    Y. Chen, S. Hu, and D. Chen. Fast pruning strategy for neural network size optimization and its applications. Journal of Chemical Industry and Engineering (China), pages 522–526, 2001

    work page 2001

  8. [8]

    Model predictive control (mpc) design and implementation using algorithm-3 on board spartan 6 fpga sp605 evaluation kit

    Cahyantari Ekaputri and Arief Syaichu-Rohman. Model predictive control (mpc) design and implementation using algorithm-3 on board spartan 6 fpga sp605 evaluation kit. In 2013 3rd International Conference on Instrumentation Control and Automation (ICA), pages 115–120, 2013

    work page 2013

Show all 28 references
  1. [9]

    Multilayer feedforward networks are universal approximators

    Kurt Hornik, Maxwell Stinchcombe, and Halbert White. Multilayer feedforward networks are universal approximators. Neural Networks, 2(5):359–366, 1989

    1989

  2. [10]

    A mathemati- cal model and pi controller design based on indirect vector control for permanent magnet synchronous motor

    Natthawut Kongchoo, Phonsit Santiprapan, and Nattha Jindapetch. A mathemati- cal model and pi controller design based on indirect vector control for permanent magnet synchronous motor. ECTI Transactions on Computer and Information Technology (ECTI-CIT), 16:260–267, 06 2022

    2022

  3. [11]

    Krause, Oleg Wasynczuk, and Scott D

    Paul C. Krause, Oleg Wasynczuk, and Scott D. Sudhoff. Reference-Frame Theory, chapter 3, pages 86–120. John Wiley & Sons, Ltd, 2013

    2013

  4. [12]

    Kuvalekar and S.R

    Suryakant A. Kuvalekar and S.R. Mohanrajan. Pmsm torque ripple reduction in electric vehicle using neural network. In 2021 IEEE International Power and Renewable Energy Conference (IPRECON) , pages 1–6, 2021

    2021

  5. [13]

    Tiny machine learning: Progress and futures [feature]

    Ji Lin, Ligeng Zhu, Wei-Ming Chen, Wei-Chen Wang, and Song Han. Tiny machine learning: Progress and futures [feature]. IEEE Circuits and Systems Magazine, 23(3):8–34, 2023

    2023

  6. [14]

    Efficient neural networks for tiny machine learning: A comprehensive review, 2023

    Minh Tri Lê, Pierre Wolinski, and Julyan Arbel. Efficient neural networks for tiny machine learning: A comprehensive review, 2023

    2023

  7. [15]

    Bayesian optimization

    MathWorks. Bayesian optimization. https://uk.mathworks.com/help/stats/ bayesopt.html, 2024

    2024

  8. [16]

    Field oriented control dataset of a 3-phase permanent magnet synchronous motor.Data in Brief, 47:109002, 2023

    Juan Camilo Nustes, Danilo Pietro Pau, and Giambattista Gruosso. Field oriented control dataset of a 3-phase permanent magnet synchronous motor.Data in Brief, 47:109002, 2023

    2023

  9. [17]

    Modelling the field oriented control applied to a 3-phase permanent magnet synchronous motor

    Juan Camilo Nustes, Danilo Pietro Pau, and Giambattista Gruosso. Modelling the field oriented control applied to a 3-phase permanent magnet synchronous motor. Software Impacts, 15:100479, 2023

    2023

  10. [18]

    Discrete-Time Control Systems

    Katsuhiko Ogata. Discrete-Time Control Systems. Prentice Hall, Australia, Sydney, 1987

    1987

  11. [19]

    Field oriented control of surface-mount pmsm using model predictive current control

    Preeti Osekar, Sahana Kalligudd, Sachin Angadi, and A B Raju. Field oriented control of surface-mount pmsm using model predictive current control. In 2022 IEEE North Karnataka Subsection Flagship International Conference (NKCon), pages 1–5, 2022

    2022

  12. [20]

    Abdesh S

    M. Abdesh S. K. Khan and M. Azizur Rahman. An adaptive self-tuned wavelet controller for ipm motor drives. In 2008 IEEE Power and Energy Society General Meeting - Conversion and Delivery of Electrical Energy in the 21st Century , pages 1–4, 2008

    2008

  13. [21]

    Tinyml: Current progress, research challenges, and future roadmap

    Muhammad Shafique, Theocharis Theocharides, Vijay Janapa Reddy, and Boris Murmann. Tinyml: Current progress, research challenges, and future roadmap. In 2021 58th ACM/IEEE Design Automation Conference (DAC) , pages 1303–1306, 2021

    2021

  14. [22]

    A novel two-step bayesian hyperparameter optimization strategy for dos attack detection in iot

    Somnath Sinha, Shirasthi Choudhary, and Aditi Paul. A novel two-step bayesian hyperparameter optimization strategy for dos attack detection in iot. In 2024 2nd International Conference on Intelligent Data Communication Technologies and Internet of Things (IDCIoT), pages 112–119, 2024

    2024

  15. [23]

    Motor control nucleo pack with nucleo-f302r8 and x-nucleo- ihm07m1

    STMicroelectronics. Motor control nucleo pack with nucleo-f302r8 and x-nucleo- ihm07m1. https://www.st.com/en/evaluation-tools/p-nucleo-ihm001.html. Last accessed: November 10, 2024

    2024

  16. [24]

    Stm32 nucleo-64 development board with stm32g474re mcu

    STMicroelectronics. Stm32 nucleo-64 development board with stm32g474re mcu. https://www.st.com/en/evaluation-tools/nucleo-g474re.html. Last accessed: November 10, 2024

    2024

  17. [25]

    St edge ai developer cloud

    STMicroelectronics. St edge ai developer cloud. https://stedgeai-dc.st.com/home, 2024

    2024

  18. [26]

    O’Brien, and Dakota Roberson

    Jonathan Tacke, John F. O’Brien, and Dakota Roberson. Auto-synthesis of high- performance power system compensators. IEEE Transactions on Power Systems , 38(2):1219–1228, 2023

    2023

  19. [27]

    Enforcing exact physics in scientific machine learning: A data-driven exterior calculus on graphs

    Nathaniel Trask, Andy Huang, and Xiaozhe Hu. Enforcing exact physics in scientific machine learning: A data-driven exterior calculus on graphs. Journal of Computational Physics, 456:110969, 2022

    2022

  20. [28]

    Linear pruning tech- niques for neural networks-based on projection latent structure

    Lei Xie, Quan-Ling Zhang, Ming Guo, and Shu-Qing Wang. Linear pruning tech- niques for neural networks-based on projection latent structure. In SMC’03 Con- ference Proceedings. 2003 IEEE International Conference on Systems, Man and Cy- bernetics. Conference Theme - System Secu...

    2003

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.