pith. sign in

arxiv: 2605.18014 · v1 · pith:XV6XJ3Q6new · submitted 2026-05-18 · ⚛️ physics.optics

Optoelectronic Chromatic Dispersion in a Single Photodiode for Machine-Learning-Based Computational Spectroscopy

Endalamaw Ewnu Kassa , Ziv Glasser , Uttama K. Saint , Roi Yozevitch , Shmuel Sternklar This is my paper

Pith reviewed 2026-05-20 00:57 UTC · model grok-4.3

classification ⚛️ physics.optics
keywords optoelectronic chromatic dispersioncomputational spectroscopysingle photodiodeGaussian Process RegressionRF amplitude and phasespectral reconstructionmachine learningC and L bands
0
0 comments X

The pith

A single photodiode encodes spectral information through optoelectronic chromatic dispersion and recovers wavelengths to 0.178 nm using Gaussian Process Regression.

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

The paper demonstrates a compact spectrometer that relies on a single germanium photodiode to capture wavelength-dependent carrier diffusion delays. These delays appear as distinct RF amplitude and phase patterns across fifteen modulation frequencies, which together with the DC voltage form a 31-dimensional feature set. Machine-learning models, especially Gaussian Process Regression, invert this feature set to reconstruct single input wavelengths across the C and L bands while spanning seven optical power levels. The approach achieves 0.178 nm accuracy on held-out, wavelength-grouped test data and maintains 0.342 nm RMSE under five-fold cross-validation, extending to dual-wavelength cases as well. This replaces bulky dispersive optics with a simple detector plus computation.

Core claim

Optoelectronic chromatic dispersion in a single photodiode produces carrier diffusion delays whose wavelength dependence is read out as repeatable RF amplitude and phase signatures at multiple modulation frequencies. These signatures, combined with DC voltage, supply a high-dimensional feature vector that machine-learning models invert to recover the input optical spectrum. Gaussian Process Regression yields 0.178 nm accuracy for single-wavelength reconstruction on a held-out test set that includes seven power levels, and 0.362 nm and 0.434 nm accuracies for the swept and fixed components of dual-wavelength inputs.

What carries the argument

Optoelectronic chromatic dispersion (OED), in which wavelength-dependent absorption depth creates measurable carrier diffusion delays that appear as RF amplitude and phase signatures at multiple modulation frequencies.

If this is right

  • Single-wavelength reconstruction reaches 0.178 nm accuracy on a wavelength-grouped held-out test set spanning seven optical power levels.
  • Five-fold cross-validation produces an RMSE of 0.342 plus or minus 0.117 nm across the C- and L-bands.
  • Dual-wavelength inputs are reconstructed to 0.362 nm for the swept wavelength and 0.434 nm for the fixed wavelength.
  • The method requires no alignment-sensitive optics and is compatible with on-chip integration.
  • The same 31-dimensional feature space supports portable optical sensing and field-deployable monitoring.

Where Pith is reading between the lines

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

  • The same OED signatures could be collected at higher modulation frequencies to increase feature dimensionality and potentially improve multi-wavelength resolution.
  • Integration with a CMOS-compatible photodiode array would allow parallel sampling of multiple spatial points while retaining the single-detector spectral capability.
  • The approach could be tested on broadband or pulsed sources to determine whether the diffusion-delay encoding remains invertible outside continuous-wave C- and L-band conditions.
  • Transfer learning from the current training set might allow rapid adaptation to new wavelength ranges without full retraining.

Load-bearing premise

The RF amplitude and phase signatures produced by optoelectronic chromatic dispersion remain sufficiently distinct and repeatable across different optical powers and wavelength combinations for machine-learning models to invert them without systematic confusion.

What would settle it

A new wavelength-power pair that produces RF signatures overlapping those of an already-trained combination and causes the Gaussian Process Regression model to output an error larger than the reported 0.342 nm RMSE would falsify the claim that the signatures are reliably invertible.

Figures

Figures reproduced from arXiv: 2605.18014 by Endalamaw Ewnu Kassa, Roi Yozevitch, Shmuel Sternklar, Uttama K. Saint, Ziv Glasser.

Figure 1
Figure 1. Figure 1: Schematic of Wavelength-dependent optical absorption and minority-carrier diffusion in a p-n photodiode: physical origin of optoelec￾tronic chromatic dispersion (OED). A sinusoidally modulated optical beam of input power Pin is incident on a p-n photodiode comprising an entrance (absorbing) region of width WE and a substrate region of width WS, separated by a depletion boundary at position d. The optical p… view at source ↗
Figure 2
Figure 2. Figure 2: Experimental setup for measuring optoelectronic chromatic dispersion (OED) in a photodiode using the modulation phase-shift (MPS) technique. A tunable laser (TL) generates a continuous-wave optical carrier whose wavelength is swept across the measurement band; a Mach￾Zehnder modulator (MZM) imposes a sinusoidal intensity modulation at angular frequency Ω. The amplitude-modulated signal propagates through a… view at source ↗
Figure 3
Figure 3. Figure 3: Dataset overview. (a) Target wavelength distribution (1530–1610 nm), showing near-uniform sampling across the C+L bands with a minor gap near 1565–1570 nm. (b) Pearson correlation matrix of the four input features—Amplitude, Phase, Voltage, and Wavelength revealing strong inter-feature correlations (|r| ≥ 0.86), with Phase–Wavelength showing the highest positive correlation (r = 0.94) and Phase-Amplitude t… view at source ↗
Figure 4
Figure 4. Figure 4: Performance comparison of five machine learning models for single-wavelength spectrum reconstruction. (a–e) Scatter plots of predicted versus true wavelength λ for Ridge, SVR, GBR, GPR, and RF; the dashed diagonal line represents the ideal fit (y = x). (f) Normalised radar chart comparing R 2 , RMSE, and MAE for all models. (g) CDF of absolute prediction errors for λ, showing that ∼90% of GPR predictions a… view at source ↗
Figure 5
Figure 5. Figure 5: Predicted versus true wavelengths for λ1 (panels a-e) and λ2 (panels f-j) across five machine learning regressors Ridge, SVR, GBR, GPR, and RF trained on the combined OED DC feature set under dual-wavelength illumination. Each panel plots the model-predicted wavelength against the ground-truth value for the held-out test set; the dashed diagonal represents the locus of ideal prediction (ˆλ = λ), and deviat… view at source ↗
Figure 6
Figure 6. Figure 6: Cumulative distribution function (CDF) of absolute reconstruction errors for (a) λ1 and (b) λ2 across five machine learning regressors - Ridge, SVR, GBR, GPR, and RF - trained on the combined OED DC feature set under dual-wavelength illumination. Horizontal dashed and dotted lines mark the 50th and 90th error percentiles, respectively; a steeper CDF rise indicates higher accuracy and greater prediction con… view at source ↗
Figure 7
Figure 7. Figure 7: Feature importance rankings derived from four machine learning models trained on the OED DC feature set, comprising frequency￾resolved phase and amplitude responses measured at 15 discrete frequencies from 100 kHz to 1.5 MHz in 100 kHz increments, together with a frequency-independent DC voltage feature. Phase1 and Amplitude1 correspond to 100 kHz, while Phase15 and Amplitude15 correspond to 1.5 MHz, with … view at source ↗
Figure 8
Figure 8. Figure 8: Residual analysis of spectrum reconstruction using Ridge, SVR, RF, GBR, and GPR. (a) Single-wavelength reconstruction residuals (∆λ = λmeas − λpred) versus true wavelength. Ridge shows the largest spread (MAE = 0.931 nm, σ = 1.182 nm), while GPR achieves the highest accuracy (MAE = 0.107 nm, σ = 0.178 nm). Dashed/dotted lines indicate zero residual and ±1σ. (b) Dual-wavelength residual maps for channel pai… view at source ↗
Figure 9
Figure 9. Figure 9: OED characterisation in a Ge p-n photodiode across the telecom C+L band. (a) OED phase shift θ(λ), referenced to λref = 1530 nm, versus wavelength at modulation frequencies of 0.1, 0.5, 1.0, and 1.5 MHz. θ increases monotonically with wavelength at all frequencies, confirming substrate-dominant device operation, wherein longer-wavelength photons generate carriers at greater absorption depths, extending the… view at source ↗
Figure 10
Figure 10. Figure 10: Five-fold cross-validation (CV) performance comparison of five regression models - Ridge, SVR, GBR, GPR, and RF -for wave￾length prediction in optical sensing system. (a) Per-fold RMSE (left) and MAE (right) for each model across the five CV folds, with mean ± standard deviation annotated in the legend. (b) Fold-resolved heatmaps of CV RMSE (left) and CV MAE (right), with color scale transitioning from ye… view at source ↗
Figure 11
Figure 11. Figure 11: Computational efficiency vs. predictive accuracy for five regression models (Ridge, SVR, GBR, GPR, RF). GPR achieves the lowest test errors while maintaining moderate training time, demonstrating an optimal balance between accuracy and computational cost. Ridge trains fastest but suffers from high prediction errors. The plot highlights the trade-off between speed and reliability across models. 5.1. Spectr… view at source ↗
Figure 12
Figure 12. Figure 12: Uncertainty quantification of the Gaussian Process Regression (GPR) model for wavelength prediction. (a) Parity plot of GPR￾predicted versus true wavelengths (λ), with error bars showing one predicted standard deviation (σ). The dashed line represents perfect agreement. (b) Absolute prediction error versus predicted σ, color-coded by true wavelength. A positive Spearman correlation (r = 0.567, p = 1.04×10… view at source ↗
read the original abstract

Spectroscopy requires high-precision wavelength discrimination but typically requires bulky, alignment-sensitive instrumentation. To address this, we present a compact computational spectrometer built from a single germanium PN photodiode. The system exploits optoelectronic chromatic dispersion (OED), a phenomenon whereby wavelength-dependent absorption depth produces carrier diffusion delays that encode spectral information as measurable RF amplitude and phase signatures in the photodiode output. We extract DC voltage, RF amplitude, and RF phase across 15 modulation frequencies (0.1-1.5 MHz), forming a 31-dimensional feature vector per optical input. Spectral reconstruction was formulated as a high-dimensional inverse problem and solved using five machine learning models, utilizing group-wavelength splitting and k-fold cross-validation to prevent spectral leakage and ensure unbiased evaluation. Across the C- and L-bands, single-wavelength reconstruction using Gaussian Process Regression (GPR) achieves an accuracy of 0.178 nm on a wavelength-grouped, held-out test set spanning seven optical power levels. Five-fold cross-validation yields a robust Root Mean Square Error (RMSE) of (0.342 +/- 0.117) nm, confirming excellent generalization under wavelength and power variations. For dual-wavelength inputs, GPR yields accuracies of 0.362 nm for the swept wavelength and 0.434 nm for the fixed wavelength. This is the first spectral reconstruction method exploiting a multi-frequency OED feature space from a single photodiode. By merging the physics of OED with data-driven learning, this work enables alignment-free, on-chip-compatible spectrometers suitable for portable optical sensing, smartphone-integrated diagnostics, and field-deployable environmental monitoring.

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

1 major / 2 minor

Summary. The manuscript introduces a compact computational spectrometer based on a single germanium PN photodiode that exploits optoelectronic chromatic dispersion (OED). Wavelength-dependent carrier diffusion delays are encoded as RF amplitude and phase signatures across 15 modulation frequencies (0.1–1.5 MHz), yielding 31-dimensional feature vectors (DC + 15 amplitudes + 15 phases). These features are inverted for spectral reconstruction using five machine-learning models, with Gaussian Process Regression (GPR) achieving 0.178 nm accuracy on single-wavelength, wavelength-grouped held-out tests spanning seven optical power levels and a 5-fold cross-validation RMSE of 0.342 ± 0.117 nm. Dual-wavelength results are also reported. The work claims this is the first use of multi-frequency OED from a single photodiode for alignment-free spectroscopy.

Significance. If the OED signatures remain sufficiently distinct and repeatable, the approach offers a genuinely compact, on-chip-compatible route to computational spectroscopy that merges device physics with data-driven inversion. The use of wavelength-grouped splitting and k-fold CV to block leakage is a positive methodological choice, and the reported single-wavelength accuracy is competitive for a single-detector system. Reproducible feature extraction from a standard photodiode is a practical strength.

major comments (1)
  1. [Results (single- and dual-wavelength reconstruction) and Methods (feature extraction and cross-validation)] The central claim that GPR reconstructs wavelength to 0.178 nm (and CV RMSE 0.342 ± 0.117 nm) across seven optical power levels rests on the unverified assumption that the 31-dimensional OED feature vectors remain linearly scalable and free of power-correlated nonlinearities (e.g., space-charge or junction-capacitance effects). No explicit check—such as power-normalized feature stability plots, linearity tests, or ablation of power as a covariate—is provided in the results or supplementary material. If such confounds exist and correlate with wavelength, the model could achieve low RMSE by fitting power-induced artifacts rather than pure OED physics.
minor comments (2)
  1. [Abstract] The abstract states 'five-fold cross-validation yields a robust Root Mean Square Error (RMSE) of (0.342 +/- 0.117) nm' but does not clarify whether the reported uncertainty is the standard deviation across folds or a different metric; this should be stated explicitly.
  2. [Introduction / Methods] Notation for the modulation frequencies and the exact definition of the 31-dimensional vector (DC + amplitudes + phases) is introduced in the abstract but would benefit from a compact equation or table in the main text for immediate reference.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We address the major comment below and describe the revisions that will be incorporated to strengthen the presentation of our results.

read point-by-point responses

  1. Referee: [Results (single- and dual-wavelength reconstruction) and Methods (feature extraction and cross-validation)] The central claim that GPR reconstructs wavelength to 0.178 nm (and CV RMSE 0.342 ± 0.117 nm) across seven optical power levels rests on the unverified assumption that the 31-dimensional OED feature vectors remain linearly scalable and free of power-correlated nonlinearities (e.g., space-charge or junction-capacitance effects). No explicit check—such as power-normalized feature stability plots, linearity tests, or ablation of power as a covariate—is provided in the results or supplementary material. If such confounds exist and correlate with wavelength, the model could achieve low RMSE by fitting power-induced artifacts rather than pure OED physics.

    Authors: We thank the referee for raising this important methodological point. Our experimental protocol varied optical power independently across seven discrete levels while sweeping wavelength, and both the wavelength-grouped held-out test set and the 5-fold cross-validation were performed on the full multi-power dataset. This design already requires the model to generalize across power variations rather than exploit power-specific correlations. Nevertheless, we agree that explicit verification of feature linearity and power independence would further substantiate the claim. In the revised manuscript we will add (i) power-normalized RF amplitude and phase stability plots at fixed wavelengths across the seven power levels, (ii) linearity tests of the 31-dimensional features versus optical power, and (iii) an ablation experiment that includes optical power as an explicit covariate. These additions, to be placed in the results section and supplementary material, will directly demonstrate that the reported reconstruction accuracy arises from OED physics rather than power-induced artifacts. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical ML reconstruction on measured OED signatures

full rationale

The paper's central result is an empirical accuracy (0.178 nm single-wavelength, RMSE 0.342 ± 0.117 nm via 5-fold CV) obtained by feeding experimentally measured 31-dimensional feature vectors (DC + 15 amplitudes + 15 phases at 0.1-1.5 MHz) from a single photodiode into standard GPR and other ML models. Wavelength-grouped splitting and cross-validation are used to block leakage. No derivation chain, equations, or self-citations are shown that reduce the reported accuracies to fitted parameters, self-definitions, or ansatzes by construction. The approach is self-contained against external benchmarks because the inputs are physical RF measurements and the outputs are predictive performance on held-out data; any hyperparameter tuning is ordinary ML practice and does not create circularity.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The approach rests on standard semiconductor physics for carrier diffusion and standard machine-learning assumptions for generalization; no new physical constants or entities are introduced beyond the measured OED effect itself.

free parameters (1)
  • modulation frequencies
    Fifteen discrete frequencies between 0.1 and 1.5 MHz are selected to sample the dispersion response; their exact values are chosen rather than derived from first principles.
axioms (1)
  • domain assumption Carrier diffusion delay varies monotonically with absorption depth and therefore with wavelength in the germanium photodiode.
    Invoked when mapping RF phase and amplitude signatures to wavelength information.

pith-pipeline@v0.9.0 · 5845 in / 1437 out tokens · 49676 ms · 2026-05-20T00:57:08.788342+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

52 extracted references · 52 canonical work pages

  1. [1]

    A review on spectroscopic techniques for analysis of nanomaterials and biomaterials.ES Energy & Environment, 27:1264, 2024

    Rai Dhirendra Prasad, Prashant D Sarvalkar, Nirmala Prasad, Saurabh R Prasad, Rai Surendra Prasad, Rai Bishwendra Prasad, Rai Rajnarayan Prasad, CB Desai, Anil Kumar Vaidya, B Teli, et al. A review on spectroscopic techniques for analysis of nanomaterials and biomaterials.ES Energy & Environment, 27:1264, 2024

    work page 2024

  2. [2]

    Spectroscopy—principle, types, and applications

    Sweta Sinha, Christine Jeyaseelan, Gunjan Singh, Tanya Munjal, and Debarati Paul. Spectroscopy—principle, types, and applications. InBasic biotechniques for bioprocess and bioentrepreneurship, pages 145–164. Elsevier, 2023

    work page 2023

  3. [3]

    Machine learning spectroscopy to advance computation and analysis.Chemical Science, 2025

    Julia Westermayr and P Marquetand. Machine learning spectroscopy to advance computation and analysis.Chemical Science, 2025

    work page 2025

  4. [4]

    Advances in miniaturized computational spectrometers.Advanced Science, 11(47):2404448, 2024

    Qian Xue, Yang Yang, Wenkai Ma, Hanqiu Zhang, Daoli Zhang, Xinzheng Lan, Liang Gao, Jianbing Zhang, and Jiang Tang. Advances in miniaturized computational spectrometers.Advanced Science, 11(47):2404448, 2024

    work page 2024

  5. [5]

    Miniaturization of optical spectrometers.Sci- ence, 371(6528):eabe0722, 2021

    Zongyin Yang, Tom Albrow-Owen, Weiwei Cai, and Tawfique Hasan. Miniaturization of optical spectrometers.Sci- ence, 371(6528):eabe0722, 2021

    work page 2021

  6. [6]

    Recent development of ultrafast optical characterizations for quantum materials.Advanced Materials, 35(27):2110068, 2023

    Tao Dong, Si-Jie Zhang, and Nan-Lin Wang. Recent development of ultrafast optical characterizations for quantum materials.Advanced Materials, 35(27):2110068, 2023

    work page 2023

  7. [7]

    Review of miniaturized computational spectrometers.Sensors, 23(21):8768, 2023

    Qingze Guan, Zi Heng Lim, Haoyang Sun, Jeremy Xuan Yu Chew, and Guangya Zhou. Review of miniaturized computational spectrometers.Sensors, 23(21):8768, 2023

    work page 2023

  8. [8]

    Advances in cost-effective integrated spectrometers.Light: Science & Applications, 11(1):174, 2022

    Ang Li, Chunhui Yao, Junfei Xia, Huijie Wang, Qixiang Cheng, Richard Penty, Yeshaiahu Fainman, and Shilong Pan. Advances in cost-effective integrated spectrometers.Light: Science & Applications, 11(1):174, 2022

    work page 2022

  9. [9]

    High resolution imaging spectrometry using dispersion shearing interferometer with amici prisms

    Huitong Huang, Liansheng Li, Yu Yang, Xiaoqun Xing, Xin Meng, Yixuan Xu, Yixuan Liu, Wenxuan Wu, Haibo Zhu, Yi Zong, et al. High resolution imaging spectrometry using dispersion shearing interferometer with amici prisms. Optics Express, 32(26):45816–45832, 2024

    work page 2024

  10. [10]

    Ryan D Muir, Daniel E Mittelberger, and John E Heebner. Three phase spectral interferometry for recording high dynamic range optical waveforms with¡ 1 ps resolution over¿ 2 ns records applied to closed-loop pulse shaping.Applied Optics, 63(22):5991–5999, 2024

    work page 2024

  11. [11]

    Metasurface spectrometers beyond resolution-sensitivity constraints.Science Advances, 10(49):eadr7155, 2024

    Feng Tang, Jingjun Wu, Tom Albrow-Owen, Hanxiao Cui, Fujia Chen, Yaqi Shi, Lan Zou, Jun Chen, Xuhan Guo, Yijun Sun, et al. Metasurface spectrometers beyond resolution-sensitivity constraints.Science Advances, 10(49):eadr7155, 2024

    work page 2024

  12. [12]

    Integrated optical spectrometers on silicon photonics platforms.Laser & Photonics Reviews, 19(7):2400155, 2025

    Zunyue Zhang, Shujiao Zhang, Xingyu Liu, Zhijie Wei, Tarun Sharma, Ganapathy Senthil Murugan, Hon Ki Tsang, Tiegen Liu, and Zhenzhou Cheng. Integrated optical spectrometers on silicon photonics platforms.Laser & Photonics Reviews, 19(7):2400155, 2025

    work page 2025

  13. [13]

    Multi-mode interference waveguide chip-scale spectrometer.APL Photonics, 9(10), 2024

    Md Nafiz Amin, Vahid Ganjalizadeh, Tyler J Adams, Porter B Dixon, Zoe Weber, Matthew DeMartino, Kevin Bundy, Aaron R Hawkins, and Holger Schmidt. Multi-mode interference waveguide chip-scale spectrometer.APL Photonics, 9(10), 2024

    work page 2024

  14. [14]

    Compact spectrometer based on a disordered photonic chip.Nature Photonics, 7(9):746–751, 2013

    Brandon Redding, Seng Fatt Liew, Raktim Sarma, and Hui Cao. Compact spectrometer based on a disordered photonic chip.Nature Photonics, 7(9):746–751, 2013

    work page 2013

  15. [15]

    One dimensional disordered photonic structures charac- terized by uniform distributions of clusters.Optical Materials, 39:235–238, 2015

    Michele Bellingeri, Ilka Kriegel, and Francesco Scotognella. One dimensional disordered photonic structures charac- terized by uniform distributions of clusters.Optical Materials, 39:235–238, 2015

    work page 2015

  16. [16]

    Arrayed waveguide grating spectrometers for astronomical applications: new results.Optics express, 25(15):17918–17935, 2017

    Pradip Gatkine, Sylvain Veilleux, Yiwen Hu, Joss Bland-Hawthorn, and Mario Dagenais. Arrayed waveguide grating spectrometers for astronomical applications: new results.Optics express, 25(15):17918–17935, 2017

    work page 2017

  17. [17]

    Integrated scanning spectrometer with a tunable micro-ring resonator and an arrayed waveguide grating.Photonics Research, 10(5):A74– A81, 2022

    Zunyue Zhang, Yi Wang, Jiarui Wang, Dan Yi, David Weng U Chan, Wu Yuan, and Hon Ki Tsang. Integrated scanning spectrometer with a tunable micro-ring resonator and an arrayed waveguide grating.Photonics Research, 10(5):A74– A81, 2022

    work page 2022

  18. [18]

    Miniaturized disordered photonic molecule spectrometer.Light: Science & Applica- tions, 14(1):144, 2025

    Yujia Zhang, Tom Albrow-Owen, Zhenyu Zhao, Yinpeng Chen, Yaotian Zhao, Hannah Joyce, Tawfique Hasan, Zongyin Yang, Yikai Su, and Xuhan Guo. Miniaturized disordered photonic molecule spectrometer.Light: Science & Applica- tions, 14(1):144, 2025

    work page 2025

  19. [19]

    Fast spectrum reconstruction based-on interpretable network with physical model involved.Optics & Laser Technology, 182:112079, 2025

    Xinyu Su, Shuangli Liu, Hui Wu, Peng Chen, Jiangnan Yang, and Jingjun Wu. Fast spectrum reconstruction based-on interpretable network with physical model involved.Optics & Laser Technology, 182:112079, 2025

    work page 2025

  20. [20]

    A survey on computational spectral reconstruction methods from rgb to hyperspectral imaging.Scientific reports, 12(1):11905, 2022

    Jingang Zhang, Runmu Su, Qiang Fu, Wenqi Ren, Felix Heide, and Yunfeng Nie. A survey on computational spectral reconstruction methods from rgb to hyperspectral imaging.Scientific reports, 12(1):11905, 2022

    work page 2022

  21. [21]

    Reconstruc- tive spectrometers: hardware miniaturization and computational reconstruction.eLight, 5(1):23, 2025

    Yiru Zhang, Enbo Yang, Hoon Hahn Yoon, Qixiang Cheng, Zhipei Sun, Tawfique Hasan, and Weiwei Cai. Reconstruc- tive spectrometers: hardware miniaturization and computational reconstruction.eLight, 5(1):23, 2025

    work page 2025

  22. [22]

    Miniaturized on-chip spectrometer enabled by electrochromic modulation.Light: Science & Applications, 13(1):278, 2024

    Menghan Tian, Baolei Liu, Zelin Lu, Yao Wang, Ze Zheng, Jiaqi Song, Xiaolan Zhong, and Fan Wang. Miniaturized on-chip spectrometer enabled by electrochromic modulation.Light: Science & Applications, 13(1):278, 2024

    work page 2024

  23. [23]

    Ultracompact computational spectroscopy with a detour-phased planar lens.Light: Advanced Manufacturing, 5(4):523– 531, 2025

    Wenkai Yang, Zijian Wang, Jian Xu, Dashan Dong, Guiyuan Cao, Han Lin, Baohua Jia, Lige Liu, and Kebin Shi. Ultracompact computational spectroscopy with a detour-phased planar lens.Light: Advanced Manufacturing, 5(4):523– 531, 2025

    work page 2025

  24. [24]

    Opto- electronic chromatic dispersion and wavelength monitoring in a photodiode.Optics Express, 29(13):19839–19852, 2021

    Ziv Glasser, Pawan Kumar Dubey, Ayuushi Dutta, Egor Liokumovitch, Rita Abramov, and Shmuel Sternklar. Opto- electronic chromatic dispersion and wavelength monitoring in a photodiode.Optics Express, 29(13):19839–19852, 2021

    work page 2021

  25. [25]

    Large and tunable optoelectronic chromatic dispersion in pin-type photodiodes.Optics Letters, 49(8):2057–2060, 2024

    Ayuushi Dutta, Egor Liokumovitch, Ziv Glaser, and Shmuel Sternklar. Large and tunable optoelectronic chromatic dispersion in pin-type photodiodes.Optics Letters, 49(8):2057–2060, 2024

    work page 2057

  26. [26]

    Large optoelectronic chromatic dispersion in pn-type silicon photodiodes and photovoltaic cells.Optics Letters, 49(8):2185–2188, 2024

    Sapna Mudgal, Pawan K Dubey, Ziv Glasser, and Shmuel Sternklar. Large optoelectronic chromatic dispersion in pn-type silicon photodiodes and photovoltaic cells.Optics Letters, 49(8):2185–2188, 2024

    work page 2024

  27. [27]

    Deep learning spectroscopy: Neural networks for molecular excitation spectra.Advanced science, 6(9):1801367, 2019

    Kunal Ghosh, Annika Stuke, Milica Todorovi ´c, Peter Bjørn Jørgensen, Mikkel N Schmidt, Aki Vehtari, and Patrick Rinke. Deep learning spectroscopy: Neural networks for molecular excitation spectra.Advanced science, 6(9):1801367, 2019

    work page 2019

  28. [28]

    Multicollinearity in regression analysis: problems, detection, and solutions.Statistics by Jim, 2:2, 2017

    Jim Frost et al. Multicollinearity in regression analysis: problems, detection, and solutions.Statistics by Jim, 2:2, 2017

    work page 2017

  29. [29]

    Fractional ridge regression: a fast, interpretable reparameterization of ridge regression

    Ariel Rokem and Kendrick Kay. Fractional ridge regression: a fast, interpretable reparameterization of ridge regression. GigaScience, 9(12):giaa133, 2020

    work page 2020

  30. [30]

    A tutorial on support vector regression.Statistics and computing, 14(3):199– 222, 2004

    Alex J Smola and Bernhard Sch ¨olkopf. A tutorial on support vector regression.Statistics and computing, 14(3):199– 222, 2004

    work page 2004

  31. [31]

    On the noise model of support vector machines regression

    Massimiliano Pontil, Sayan Mukherjee, and Federico Girosi. On the noise model of support vector machines regression. InInternational Conference on Algorithmic Learning Theory, pages 316–324. Springer, 2000

    work page 2000

  32. [32]

    Random forests.Machine learning, 45(1):5–32, 2001

    Leo Breiman. Random forests.Machine learning, 45(1):5–32, 2001

    work page 2001

  33. [33]

    Greedy function approximation: a gradient boosting machine.Annals of statistics, pages 1189– 1232, 2001

    Jerome H Friedman. Greedy function approximation: a gradient boosting machine.Annals of statistics, pages 1189– 1232, 2001

    work page 2001

  34. [34]

    Gaussian processes for machine learning.International journal of neural systems, 14(02):69–106, 2004

    Matthias Seeger. Gaussian processes for machine learning.International journal of neural systems, 14(02):69–106, 2004

    work page 2004

  35. [35]

    An intuitive tutorial to gaussian process regression.Computing in Science & Engineering, 25(4):4–11, 2023

    Jie Wang. An intuitive tutorial to gaussian process regression.Computing in Science & Engineering, 25(4):4–11, 2023

    work page 2023

  36. [36]

    Carrier dynamic identification enables wavelength and intensity sensitivity in perovskite photodetectors.Light: Science & Applications, 13(1):280, 2024

    Liangliang Min, Yicheng Zhou, Haoxuan Sun, Linqi Guo, Meng Wang, Fengren Cao, Wei Tian, and Liang Li. Carrier dynamic identification enables wavelength and intensity sensitivity in perovskite photodetectors.Light: Science & Applications, 13(1):280, 2024

    work page 2024

  37. [37]

    A survey on multi-output regression.Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 5(5):216–233, 2015

    Hanen Borchani, Gherardo Varando, Concha Bielza, and Pedro Larranaga. A survey on multi-output regression.Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 5(5):216–233, 2015

    work page 2015

  38. [38]

    Survey on multi-output learning

    Donna Xu, Yaxin Shi, Ivor W Tsang, Yew-Soon Ong, Chen Gong, and Xiaobo Shen. Survey on multi-output learning. IEEE transactions on neural networks and learning systems, 31(7):2409–2429, 2019

    work page 2019

  39. [39]

    Dual-wavelength visible photodetector based on vertical (in, ga) n nanowires grown by molec- ular beam epitaxy.RSC advances, 11(26):15632–15638, 2021

    Jianya Zhang, Min Zhou, Dongmin Wu, Lifeng Bian, Yukun Zhao, Hua Qin, Wenxian Yang, Yuanyuan Wu, Zhiwei Xing, and Shulong Lu. Dual-wavelength visible photodetector based on vertical (in, ga) n nanowires grown by molec- ular beam epitaxy.RSC advances, 11(26):15632–15638, 2021

    work page 2021

  40. [40]

    Nonlinear self-calibrated spectrometer with single gese-inse heterojunction device.Science Advances, 10(20):eadn6028, 2024

    Rana Darweesh, Rajesh Kumar Yadav, Elior Adler, Michal Poplinger, Adi Levi, Jea-Jung Lee, Amir Leshem, Ashwin Ramasubramaniam, Fengnian Xia, and Doron Naveh. Nonlinear self-calibrated spectrometer with single gese-inse heterojunction device.Science Advances, 10(20):eadn6028, 2024

    work page 2024

  41. [41]

    The frequency response characteristics of ge-on-si photodetectors under high incident power.Nanomaterials, 15(5):398, 2025

    Jin Jiang, Hongmin Chen, Fenghe Yang, Chunlai Li, Jin He, Xiumei Wang, and Jishi Cui. The frequency response characteristics of ge-on-si photodetectors under high incident power.Nanomaterials, 15(5):398, 2025

    work page 2025

  42. [42]

    Temperature-dependent anomalous dispersion of poly- crystalline germanium film near absorption threshold.Materials Chemistry and Physics, 243:122562, 2020

    Jiawei Xu, Wenjun Shi, Xiaoyan Wen, Haifei Lu, and Min Li. Temperature-dependent anomalous dispersion of poly- crystalline germanium film near absorption threshold.Materials Chemistry and Physics, 243:122562, 2020

    work page 2020

  43. [43]

    Characterization of power-to-phase conversion in high-speed pin photodiodes.IEEE Photonics Journal, 3(1):140–151, 2011

    Jennifer Taylor, Shubahshish Datta, Archita Hati, C Nelson, F Quinlan, Abhay Joshi, and S Diddams. Characterization of power-to-phase conversion in high-speed pin photodiodes.IEEE Photonics Journal, 3(1):140–151, 2011

    work page 2011

  44. [44]

    Self-calibrated measurement of frequency response for broadband photodetectors based on two-tone photonic sampling.Frontiers in Physics, 9:678777, 2021

    Mengke Wang, Ying Xu, Yutong He, Zhao Liu, Yali Zhang, Zhiyao Zhang, Heping Li, Shangjian Zhang, and Yong Liu. Self-calibrated measurement of frequency response for broadband photodetectors based on two-tone photonic sampling.Frontiers in Physics, 9:678777, 2021

    work page 2021

  45. [45]

    Sparse spectrum gaussian process regression.The Journal of Machine Learning Research, 11:1865–1881, 2010

    Miguel L ´azaro-Gredilla, Joaquin Quinonero-Candela, Carl Edward Rasmussen, and An´ıbal R Figueiras-Vidal. Sparse spectrum gaussian process regression.The Journal of Machine Learning Research, 11:1865–1881, 2010

    work page 2010

  46. [46]

    Optimization of the phase generated carrier demodulation algorithm based on support vector regression.Applied Optics, 60(31):9818– 9827, 2021

    Huyong Ma, Binai Li, Min Xue, Wenbin Zhao, Shuangjiao Wei, Zhenglan Bian, and Fenghong Chu. Optimization of the phase generated carrier demodulation algorithm based on support vector regression.Applied Optics, 60(31):9818– 9827, 2021

    work page 2021

  47. [47]

    Model-agnostic feature importance and effects with dependent features: a conditional subgroup approach.Data Mining and Knowledge Discovery, 38(5):2903– 2941, 2024

    Christoph Molnar, Gunnar K ¨onig, Bernd Bischl, and Giuseppe Casalicchio. Model-agnostic feature importance and effects with dependent features: a conditional subgroup approach.Data Mining and Knowledge Discovery, 38(5):2903– 2941, 2024

    work page 2024

  48. [48]

    Cundiff et al

    Steven T. Cundiff et al. Analysis of shot noise in the detection of ultrashort optical pulse trains.Applied Physics Letters, 102(16):163507, 2013

    work page 2013

  49. [49]

    John Wiley & Sons, 2012

    Govind P Agrawal.Fiber-optic communication systems. John Wiley & Sons, 2012

    work page 2012

  50. [50]

    Characteriza- tion of germanium photodiodes and trap detector.Measurement Science and Technology, 17(4):908–912, 2006

    Antti Lamminp ¨a¨a, Mart Noorma, Tuomas Hyypp¨a, Farshid Manoocheri, Petri K¨arh¨a, and Erkki Ikonen. Characteriza- tion of germanium photodiodes and trap detector.Measurement Science and Technology, 17(4):908–912, 2006

    work page 2006

  51. [51]

    Optoelectronic chromatic disper- sion in germanium pn photodiodes: wavelength monitoring and fbg interrogation.Optics Letters, 46(16):4061–4064, 2021

    Egor Liokumovitch, Ziv Glasser, Leeju Singh, Rita Abramov, and Shmuel Sternklar. Optoelectronic chromatic disper- sion in germanium pn photodiodes: wavelength monitoring and fbg interrogation.Optics Letters, 46(16):4061–4064, 2021

    work page 2021

  52. [52]

    Employing gaussian process regression with bayesian inference to predict the living- materials producer price index in china.Quality & Quantity, pages 1–39, 2025

    Bingzi Jin and Xiaojie Xu. Employing gaussian process regression with bayesian inference to predict the living- materials producer price index in china.Quality & Quantity, pages 1–39, 2025

    work page 2025