pith. sign in

arxiv: 2404.10425 · v2 · submitted 2024-04-16 · 💻 cs.RO · cs.AI

Optimizing BioTac Simulation for Realistic Tactile Perception

Wadhah Zai El Amri , Nicol\'as Navarro-Guerrero This is my paper

Pith reviewed 2026-05-24 02:25 UTC · model grok-4.3

classification 💻 cs.RO cs.AI
keywords BioTac simulationtactile sensingXGBoost regressortransformer encoderneural network comparisonforce and contact datasensor output predictionrobot perception
0
0 comments X

The pith

BioTac simulation models achieve better accuracy by training without temperature readings.

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

The paper first shows that including BioTac temperature readings in a simulation model fails to produce accurate sensor output predictions once deployed on a robot. It then evaluates three alternatives—an XGBoost regressor, a standard neural network, and a transformer encoder—trained solely on force and contact point positions while varying the size of the input window. These models deliver statistically significant gains over the baseline, and the XGBoost and transformer versions outperform the feed-forward neural network. A sympathetic reader would care because realistic tactile simulation lets robots handle physical contact more reliably without extra sensor channels.

Core claim

Excluding temperature from the training data lets an XGBoost regressor and a transformer encoder predict BioTac sensor outputs more accurately than a baseline feed-forward network; the improvement holds across tested input window sizes and reaches statistical significance.

What carries the argument

Input window size applied to force and contact point sequences fed into an XGBoost regressor, neural network, or transformer encoder.

If this is right

  • XGBoost and transformer models become the preferred choice over feed-forward networks for this prediction task.
  • Input window size can be tuned to further improve prediction quality without temperature.
  • Simulation accuracy increases enough to support more reliable robot tactile responses.
  • Temperature readings can be omitted from the deployed model without loss of performance.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same exclusion of temperature might help simulation of other nonlinear tactile sensors that also suffer from thermal drift.
  • Real-time robot control loops could run faster if the model no longer needs to read or process temperature.
  • Combining the improved simulation with vision or proprioception might yield more robust manipulation policies.

Load-bearing premise

Models trained only on force and contact data will still produce accurate outputs when the robot is deployed in the real world.

What would settle it

Record actual BioTac readings during a physical robot grasping or sliding task and compare them directly to the outputs of the temperature-free XGBoost or transformer model.

Figures

Figures reproduced from arXiv: 2404.10425 by Nicol\'as Navarro-Guerrero, Wadhah Zai El Amri.

Figure 1
Figure 1. Figure 1: (a): Map of the electrodes on the BioTac sensor [ [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Average normalized MAE over all channels for ten folds for Network [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 2
Figure 2. Figure 2: Temperature values of the BioTac sensor in the Ruppel et al. [ [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Visualization of the training and validation loss values for all [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of our used transformer architecture, based on Dosovitskiy [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 8
Figure 8. Figure 8: The Normalized MAE was calculated over all channels for the four [PITH_FULL_IMAGE:figures/full_fig_p006_8.png] view at source ↗
Figure 7
Figure 7. Figure 7: Distribution of nearest contact points for each electrode. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visualization of inference time against the number of parameters for [PITH_FULL_IMAGE:figures/full_fig_p007_9.png] view at source ↗
read the original abstract

Tactile sensing presents a promising opportunity for enhancing the interaction capabilities of today's robots. BioTac is a commonly used tactile sensor that enables robots to perceive and respond to physical tactile stimuli. However, the sensor's non-linearity poses challenges in simulating its behavior. In this paper, we first investigate a BioTac simulation that uses temperature, force, and contact point positions to predict the sensor outputs. We show that training with BioTac temperature readings does not yield accurate sensor output predictions during deployment. Consequently, we tested three alternative models, i.e., an XGBoost regressor, a neural network, and a transformer encoder. We train these models without temperature readings and provide a detailed investigation of the window size of the input vectors. We demonstrate that we achieve statistically significant improvements over the baseline network. Furthermore, our results reveal that the XGBoost regressor and transformer outperform traditional feed-forward neural networks in this task. We make all our code and results available online on https://github.com/wzaielamri/Optimizing_BioTac_Simulation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper investigates BioTac tactile sensor simulation, finding that models trained with temperature, force, and contact positions fail to produce accurate outputs during real-robot deployment. It then evaluates three temperature-free alternatives (XGBoost regressor, feed-forward neural network, transformer encoder) on force and contact data, reporting statistically significant gains over a baseline network on held-out data, with XGBoost and transformer performing best. Window-size effects are analyzed and code/results are released.

Significance. If the temperature-free models prove effective in deployment (where the temperature-inclusive baseline failed), the work would meaningfully advance practical tactile simulation for robotics by mitigating non-linearity and generalization issues. The open release of code and results is a clear strength supporting reproducibility and follow-on work.

major comments (2)
  1. [Deployment evaluation] Deployment evaluation (abstract and results): The manuscript shows temperature-inclusive training yields inaccurate predictions on deployment but provides no corresponding real-robot deployment metrics for the XGBoost, NN, or transformer models. Since the central motivation is 'realistic tactile perception' in robots and the temperature-free approach is motivated precisely by the baseline's deployment failure, explicit deployment results (or a clear statement that none were collected) are required to support the claims.
  2. [Results] Results and abstract: The claims of 'statistically significant improvements' and superior performance of XGBoost/transformer lack reported dataset size, number of samples/trials, error bars, exact baseline architecture, exclusion criteria, or the specific statistical test used. These omissions make it impossible to assess whether the held-out gains are robust or load-bearing for the performance conclusions.
minor comments (2)
  1. [Abstract] The abstract refers to 'a neural network' as one of the tested models but later distinguishes the baseline feed-forward network; clarify whether the tested NN is distinct from the baseline.
  2. [Figures/Tables] Figure and table captions should explicitly state whether metrics are on held-out collected data or deployment data.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful review and constructive feedback. We address each major comment below. We agree that additional details on the experimental setup are needed and will revise the manuscript accordingly. Regarding deployment, we will clarify the scope of our evaluations.

read point-by-point responses

  1. Referee: [Deployment evaluation] Deployment evaluation (abstract and results): The manuscript shows temperature-inclusive training yields inaccurate predictions on deployment but provides no corresponding real-robot deployment metrics for the XGBoost, NN, or transformer models. Since the central motivation is 'realistic tactile perception' in robots and the temperature-free approach is motivated precisely by the baseline's deployment failure, explicit deployment results (or a clear statement that none were collected) are required to support the claims.

    Authors: We agree that real-robot deployment results for the XGBoost, neural network, and transformer models would provide stronger support for the claims. The manuscript reports that the temperature-inclusive baseline produced inaccurate predictions during real-robot deployment, which motivated the temperature-free models. However, the new models were only evaluated on held-out test data from the collected dataset; no corresponding real-robot deployment metrics were collected for them. In the revision, we will add an explicit statement clarifying that deployment evaluations were not performed for the proposed models and discuss this as a limitation of the current work. revision: yes

  2. Referee: [Results] Results and abstract: The claims of 'statistically significant improvements' and superior performance of XGBoost/transformer lack reported dataset size, number of samples/trials, error bars, exact baseline architecture, exclusion criteria, or the specific statistical test used. These omissions make it impossible to assess whether the held-out gains are robust or load-bearing for the performance conclusions.

    Authors: We acknowledge that the manuscript does not provide sufficient details on the experimental protocol to fully assess robustness. The dataset consists of force and contact position inputs paired with BioTac outputs, collected across multiple trials. In the revised version, we will report the total dataset size and number of samples/trials, include error bars on all performance metrics, specify the exact architecture and hyperparameters of the baseline feed-forward neural network, describe any data exclusion criteria, and state the statistical test (along with p-values) used to establish significance. The released code and results on GitHub already contain the raw data and scripts, but these details will now be added to the paper text for clarity. revision: yes

Circularity Check

0 steps flagged

No circularity: standard empirical ML training and held-out evaluation

full rationale

The paper reports an empirical machine-learning study: models (XGBoost, feed-forward NN, transformer) are trained on collected BioTac force/contact data to predict sensor outputs, with performance measured via statistical tests on held-out data. No equations, derivations, or parameter-fitting steps are described that reduce a claimed prediction to its own inputs by construction. No self-citation load-bearing arguments, uniqueness theorems, or ansatzes appear in the reported claims. The approach is self-contained against external benchmarks (held-out test sets and significance testing) and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Only the abstract is available, so the full set of modeling assumptions cannot be audited. The work appears to rest on standard supervised regression assumptions without introducing new free parameters or invented entities.

axioms (1)
  • domain assumption Standard machine learning assumptions that XGBoost, feed-forward networks, and transformers are appropriate regressors for time-windowed sensor data.
    Invoked implicitly by training and comparing these models on the BioTac task.

pith-pipeline@v0.9.0 · 5715 in / 1247 out tokens · 22312 ms · 2026-05-24T02:25:42.399113+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

43 extracted references · 43 canonical work pages

  1. [1]

    Visuo- Haptic Object Perception for Robots: An Overview,

    N. Navarro-Guerrero, S. Toprak, J. Josifovski, and L. Jamone, “Visuo- Haptic Object Perception for Robots: An Overview,” Autonomous Robots, vol. 47, no. 4, pp. 377–403, 2023

    work page 2023

  2. [2]

    Tactile Sensors,

    A. Parmiggiani, S. Dussoni, L. Natale, and G. Metta, “Tactile Sensors,” in Encyclopedia of Robotics . Springer, 2020, pp. 1–11

    work page 2020

  3. [3]

    Analysis of Randomization Effects on Sim2Real Transfer in Reinforcement Learning for Robotic Manipulation Tasks,

    J. Josifovski, M. Malmir, N. Klarmann, B. L. Zagar, N. Navarro-Guerrero, and A. Knoll, “Analysis of Randomization Effects on Sim2Real Transfer in Reinforcement Learning for Robotic Manipulation Tasks,” in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2022, pp. 10 193–10 200

    work page 2022

  4. [4]

    Continual Learning from Demonstration of Robotics Skills,

    S. Auddy, J. Hollenstein, M. Saveriano, A. Rodr ´ıguez-S´anchez, and J. Piater, “Continual Learning from Demonstration of Robotics Skills,” Robotics and Autonomous Systems , vol. 165, p. 104427, 2023

    work page 2023

  5. [5]

    TACTO: A Fast, Flexible, and Open-Source Simulator for High-Resolution Vision-Based Tactile Sensors,

    S. Wang, M. Lambeta, P.-W. Chou, and R. Calandra, “TACTO: A Fast, Flexible, and Open-Source Simulator for High-Resolution Vision-Based Tactile Sensors,” IEEE Robotics and Automation Letters , vol. 7, no. 2, pp. 3930–3937, 2022

    work page 2022

  6. [6]

    DIGIT: A Novel Design for a Low-Cost Compact High- Resolution Tactile Sensor With Application to In-Hand Manipulation,

    M. Lambeta, P.-W. Chou, S. Tian, B. Yang, B. Maloon, V . R. Most, D. Stroud, R. Santos, A. Byagowi, G. Kammerer, D. Jayaraman, and R. Calandra, “DIGIT: A Novel Design for a Low-Cost Compact High- Resolution Tactile Sensor With Application to In-Hand Manipulation,” IEEE Robotics and Automation Letters , vol. 5, no. 3, pp. 3838–3845, 2020

    work page 2020

  7. [7]

    OmniTact: A Multi-Directional High-Resolution Touch Sensor,

    A. Padmanabha, F. Ebert, S. Tian, R. Calandra, C. Finn, and S. Levine, “OmniTact: A Multi-Directional High-Resolution Touch Sensor,” in IEEE International Conference on Robotics and Automation (ICRA) , 2020, pp. 618–624

    work page 2020

  8. [8]

    Using Tactile Sensing to Improve the Sample Efficiency and Performance of Deep Deterministic Policy Gradients for Simulated in-Hand Manipulation Tasks,

    A. Melnik, L. Lach, M. Plappert, T. Korthals, R. Haschke, and H. Ritter, “Using Tactile Sensing to Improve the Sample Efficiency and Performance of Deep Deterministic Policy Gradients for Simulated in-Hand Manipulation Tasks,” Frontiers in Robotics and AI , vol. 8, no. 538773, 2021

    work page 2021

  9. [9]

    Gazebo plugin for the icub skin,

    J, Geukes, M. Nakatenus and R. Calandra. “Gazebo plugin for the icub skin,” 2017. Online Available: https://github.com/robertocalandra/ icub-gazebo-skin

    work page 2017

  10. [10]

    Simulation of the SynTouch BioTac Sensor,

    P. Ruppel, Y . Jonetzko, M. G ¨orner, N. Hendrich, and J. Zhang, “Simulation of the SynTouch BioTac Sensor,” inInternational Conference on Intelligent Autonomous Systems (IAS) , ser. AISC, vol. 867. Springer International Publishing, 2019, pp. 374–387

    work page 2019

  11. [11]

    Interpreting and Predicting Tactile Signals for the SynTouch BioTac,

    Y . S. Narang, B. Sundaralingam, K. Van Wyk, A. Mousavian, and D. Fox, “Interpreting and Predicting Tactile Signals for the SynTouch BioTac,” The International Journal of Robotics Research , vol. 40, no. 12-14, pp. 1467–1487, 2021

    work page 2021

  12. [12]

    Generation of Tactile Data From 3D Vision and Target Robotic Grasps,

    B. S. Zapata-Impata, P. Gil, Y . Mezouar, and F. Torres, “Generation of Tactile Data From 3D Vision and Target Robotic Grasps,” IEEE Transactions on Haptics , vol. 14, no. 1, pp. 57–67, 2021

    work page 2021

  13. [13]

    Biomimetic Tactile Sensor for Control of Grip,

    N. Wettels, D. Popovic, V . J. Santos, R. S. Johansson, and G. E. Loeb, “Biomimetic Tactile Sensor for Control of Grip,” in IEEE International Conference on Rehabilitation Robotics (ICORR) , 2007, pp. 923–932

    work page 2007

  14. [14]

    Multimodal Tactile Sensor,

    N. Wettels, J. A. Fishel, and G. E. Loeb, “Multimodal Tactile Sensor,” in The Human Hand as an Inspiration for Robot Hand Development , ser. Springer Tracts in Advanced Robotics. Springer International Publishing, 2014, no. 95, pp. 405–429

    work page 2014

  15. [15]

    Tactile Sensing,

    L. Natale and G. Cannata, “Tactile Sensing,” in Humanoid Robotics: A Reference. Springer Netherlands, 2019, pp. 2539–2561

    work page 2019

  16. [16]

    Self- Supervised Regrasping Using Spatio-Temporal Tactile Features and Reinforcement Learning,

    Y . Chebotar, K. Hausman, Z. Su, G. Sukhatme, and S. Schaal, “Self- Supervised Regrasping Using Spatio-Temporal Tactile Features and Reinforcement Learning,” in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2016, pp. 1960–1966

    work page 2016

  17. [17]

    Tactile Identification of Objects Using Bayesian Exploration,

    D. Xu, G. E. Loeb, and J. A. Fishel, “Tactile Identification of Objects Using Bayesian Exploration,” in IEEE International Conference on Robotics and Automation (ICRA) , 2013, pp. 3056–3061

    work page 2013

  18. [18]

    Using the BioTac as a Tumor Localization Tool,

    M. S. Arian, C. A. Blaine, G. E. Loeb, and J. A. Fishel, “Using the BioTac as a Tumor Localization Tool,” in IEEE Haptics Symposium (HAPTICS), 2014, pp. 443–448

    work page 2014

  19. [19]

    Cutaneous Feedback of Fingertip Deformation and Vibration for Palpation in Robotic Surgery,

    C. Pacchierotti, D. Prattichizzo, and K. J. Kuchenbecker, “Cutaneous Feedback of Fingertip Deformation and Vibration for Palpation in Robotic Surgery,” IEEE Transactions on Biomedical Engineering , vol. 63, no. 2, pp. 278–287, 2016

    work page 2016

  20. [20]

    SynTouch, BioTac ® Product Manual, August, 2018, V21: https://www.syntouchinc.com/wp-content/uploads/2018/08/ BioTac-Manual-V .21.pdf

    work page 2018

  21. [21]

    The Shadow Dexterous Hand: https://www.shadowrobot.com/ dexterous-hand-series/

    work page

  22. [22]

    com/products/ft/ft models.aspx?id=Nano17

    ATI Industrial Automation Inc.: F/T Sensor Nano17 https://www.ati-ia. com/products/ft/ft models.aspx?id=Nano17

    work page

  23. [23]

    Design and Use Paradigms for Gazebo, an Open-Source Multi-Robot Simulator,

    N. Koenig and A. Howard, “Design and Use Paradigms for Gazebo, an Open-Source Multi-Robot Simulator,” in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , vol. 3, 2004, pp. 2149–2154

    work page 2004

  24. [24]

    PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space,

    C. R. Qi, L. Yi, H. Su, and L. J. Guibas, “PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space,” in International Conference on Neural Information Processing Systems (NIPS) , ser. NIPS’17. Curran Associates Inc., 2017, pp. 5105–5114

    work page 2017

  25. [25]

    NVIDIA Isaac Gym Simulator: https://developer.nvidia.com/isaac-gym

    work page

  26. [26]

    Sim-to-Real for Robotic Tactile Sensing Via Physics-Based Simulation and Learned Latent Projections,

    Y . Narang, B. Sundaralingam, M. Macklin, A. Mousavian, and D. Fox, “Sim-to-Real for Robotic Tactile Sensing Via Physics-Based Simulation and Learned Latent Projections,” in IEEE International Conference on Robotics and Automation (ICRA) , 2021, pp. 6444–6451

    work page 2021

  27. [27]

    Extended Tactile Perception: Vibration Sensing through Tools and Grasped Objects,

    T. Taunyazov, L. S. Song, E. Lim, H. H. See, D. Lee, B. C. Tee, and H. Soh, “Extended Tactile Perception: Vibration Sensing through Tools and Grasped Objects,” in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2021, pp. 1755–1762

    work page 2021

  28. [28]

    SynTouch, BioTac ® SP Product Manual, August, 2018, V3: https://www.syntouchinc.com/wp-content/uploads/2018/08/ BioTac-SP-Manual-V .3.pdf

    work page 2018

  29. [29]

    Generalizing Semi-Supervised Generative Adversarial Networks to Regression Using Feature Contrast- ing,

    G. Olmschenk, Z. Zhu, and H. Tang, “Generalizing Semi-Supervised Generative Adversarial Networks to Regression Using Feature Contrast- ing,” Computer Vision and Image Understanding , vol. 186, pp. 1–12, 2019

    work page 2019

  30. [30]

    AprilTag: A Robust and Flexible Visual Fiducial System,

    E. Olson, “AprilTag: A Robust and Flexible Visual Fiducial System,” in IEEE International Conference on Robotics and Automation (ICRA) , 2011, pp. 3400–3407

    work page 2011

  31. [31]

    Estimating point of contact, force and torque in a biomimetic tactile sensor with deformable skin,

    C.-H. Lin, J. A. Fishel, and G. E. Loeb, “Estimating point of contact, force and torque in a biomimetic tactile sensor with deformable skin,” in ”Technical report, SynTouch LLC” , 2013

    work page 2013

  32. [32]

    Approximate Statistical Tests for Comparing Supervised Classification Learning Algorithms,

    T. G. Dietterich, “Approximate Statistical Tests for Comparing Supervised Classification Learning Algorithms,” Neural Computation, vol. 10, no. 7, pp. 1895–1923, Oct. 1998

    work page 1923

  33. [33]

    Inference for the Generalization Error,

    C. Nadeau and Y . Bengio, “Inference for the Generalization Error,” Machine Learning, vol. 52, no. 3, pp. 239–281, Sep. 2003

    work page 2003

  34. [34]

    XGBoost: A Scalable Tree Boosting System,

    T. Chen and C. Guestrin, “XGBoost: A Scalable Tree Boosting System,” in ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD) , ser. KDD ’16. ACM, 2016, pp. 785–794

    work page 2016

  35. [35]

    A Comparative Analysis of Gradient Boosting Algorithms,

    C. Bent ´ejac, A. Cs ¨org˝o, and G. Mart ´ınez-Mu˜noz, “A Comparative Analysis of Gradient Boosting Algorithms,” Artificial Intelligence Review, vol. 54, no. 3, pp. 1937–1967, 2021

    work page 1937

  36. [36]

    Hidden Representations in Deep Neural Networks: Part 2. Regression Problems,

    L. Das, A. Sivaram, and V . Venkatasubramanian, “Hidden Representations in Deep Neural Networks: Part 2. Regression Problems,” Computers & Chemical Engineering, vol. 139, p. 106895, 2020

    work page 2020

  37. [37]

    Trans- formers in Time Series: A Survey,

    Q. Wen, T. Zhou, C. Zhang, W. Chen, Z. Ma, J. Yan, and L. Sun, “Trans- formers in Time Series: A Survey,” in International Joint Conference on Artificial Intelligence (IJCAI) , vol. 6, 2023, pp. 6778–6786

    work page 2023

  38. [38]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Un- terthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale,” in International Conference on Learning Representations (ICLR), 2021

    work page 2021

  39. [39]

    SMAC3: A Versatile Bayesian Optimization Package for Hyperparameter Optimization,

    M. Lindauer, K. Eggensperger, M. Feurer, A. Biedenkapp, D. Deng, C. Benjamins, T. Ruhkopf, R. Sass, and F. Hutter, “SMAC3: A Versatile Bayesian Optimization Package for Hyperparameter Optimization,” Journal of Machine Learning Research , vol. 23, no. 54, pp. 1–9, 2022

    work page 2022

  40. [40]

    Hyperband: A Novel Bandit-Based Approach to Hyperparameter Opti- mization,

    L. Li, K. Jamieson, G. DeSalvo, A. Rostamizadeh, and A. Talwalkar, “Hyperband: A Novel Bandit-Based Approach to Hyperparameter Opti- mization,” Journal of Machine Learning Research , vol. 18, no. 185, pp. 1–52, 2018

    work page 2018

  41. [41]

    Adam: A Method for Stochastic Optimization,

    D. P. Kingma and J. Ba, “Adam: A Method for Stochastic Optimization,” in International Conference on Learning Representations (ICLR) , ser. 3rd, 2015, p. 15

    work page 2015

  42. [42]

    Efficient Transformers: A Survey,

    Y . Tay, M. Dehghani, D. Bahri, and D. Metzler, “Efficient Transformers: A Survey,” ACM Computing Surveys , vol. 55, no. 6, pp. 109:1–109:28, 2022

    work page 2022

  43. [43]

    An Experimental Analysis of the Power Consumption of Convolutional Neural Networks for Keyword Spotting,

    R. Tang, W. Wang, Z. Tu, and J. Lin, “An Experimental Analysis of the Power Consumption of Convolutional Neural Networks for Keyword Spotting,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Calgary, AB, Canada, 2018, pp. 5479–5483. TABLE VI SUMMARY OF THE HYPERPARAMETER SEARCH SPACE USED FOR THE XGB OOST REGRESSOR ,...

    work page 2018