Deep learning reconstruction of ultrashort pulses

Zahavy, Tom; Dikopoltsev, Alex; Moss, Daniel; Haham, Gil Ilan; Cohen, Oren; Mannor, Shie; Segev, Mordechai

doi:10.1364/optica.5.000666

Cited by 152 publications

(89 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Machine learning has been widely used in areas such as control systems, speech processing, neuroscience and computer vision 24 , and has very recently been applied to predicting the behaviour of chaotic systems 25 , 26 . Applications of machine learning in the field of photonics is also relatively recent, but a number of studies have been reported in laser optimisation 27 , 28 , ultrashort pulse measurements 29 , label-free cell classification 30 , imaging 31 – 33 and coherent communications 34 . In our case, we aim to apply the techniques of machine learning to the study of chaotic nonlinear dynamics in optics, with the particular aim of studying the statistics of the maximum intensity of temporal peaks in noise-seeded modulation instability using only spectral measurements.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Machine learning analysis of extreme events in optical fibre modulation instability

Närhi¹,

Salmela²,

Toivonen³

et al. 2018

Nat Commun

118

View full text Add to dashboard Cite

A central research area in nonlinear science is the study of instabilities that drive extreme events. Unfortunately, techniques for measuring such phenomena often provide only partial characterisation. For example, real-time studies of instabilities in nonlinear optics frequently use only spectral data, limiting knowledge of associated temporal properties. Here, we show how machine learning can overcome this restriction to study time-domain properties of optical fibre modulation instability based only on spectral intensity measurements. Specifically, a supervised neural network is trained to correlate the spectral and temporal properties of modulation instability using simulations, and then applied to analyse high dynamic range experimental spectra to yield the probability distribution for the highest temporal peaks in the instability field. We also use unsupervised learning to classify noisy modulation instability spectra into subsets associated with distinct temporal dynamic structures. These results open novel perspectives in all systems exhibiting instability where direct time-domain observations are difficult.

show abstract

Section: Resultsmentioning

confidence: 99%

“…This is essential for its use in training the network to subsequently process experimental data. We also note that the use of numerical data to train a network prior to analysing experimental results has previously been used in ultrashort pulse measurement applications 29 .…”

Section: Resultsmentioning

confidence: 99%

Machine learning analysis of extreme events in optical fibre modulation instability

Närhi¹,

Salmela²,

Toivonen³

et al. 2018

Nat Commun

118

View full text Add to dashboard Cite

show abstract

“…These collected effective sensing signals (transmission depths) from the designed SIMRR sensor are processed based on a BP-ANN. The ANN is widely known machine learning techniques, which has been developed to tackle a variety of problems in many sensing areas [45][46][47][48][49]. As shown in Figure 2a, the BP-ANN consists of three types of layers, i.e.…”

Section: Multimode Sensing Based On Artificial Neuron Networkmentioning

confidence: 99%

Multi-Parameter Sensing in a Multimode Self-Interference Micro-Ring Resonator by Machine Learning

Dong

Zou

Ren

et al. 2020

Sensors

View full text Add to dashboard Cite

A universal multi-parameter sensing scheme based on a self-interference micro-ring resonator (SIMRR) is proposed. Benefit from the special intensity sensing mechanism, the SIMRR allows multimode sensing in a wide range of wavelengths but immune from frequency noise. To process the multiple mode spectra that are dependent on multiple parameters, we adopt the machine learning algorithm instead of massive asymptotic solutions of resonators. Employing the proposed multi-mode sensing approach, a two-parameter SIMRR sensor is designed. Assuming that two gases have different wavelength dependence of refractive indices, the feasibility and effectiveness of the two-parameter sensing strategy are verified numerically. Moreover, the dependence of parameter estimation accuracy on the laser intensity noises is also investigated. The numerical results indicate that our scheme of multi-parameter sensing in a multimode SIMRR holds great potential for practical high-sensitive sensing platforms compared with the single-mode sensing based on whispering gallery mode (WGM) resonators.

show abstract

“…Recently, ptychographic algorithms were adapted to reconstruct FROG spectrograms and led to so-called 'time-domain ptychography' [25,26]. A strong improvement in the reconstruction procedure was further demonstrated [27][28][29]. However, in all these time-domain ptychography studies to date, the reconstructed spectrograms result from conventional FROG measurements.…”

Section: Output Spectrummentioning

confidence: 99%

Phase-matching-free pulse retrieval based on transient absorption in solids

Leblanc¹,

Lassonde²,

Petit³

et al. 2019

Opt. Express

View full text Add to dashboard Cite

In this paper, we introduce a pulse characterization technique that is free of phasematching constraints, exploiting transient absorption in solids as an ultrafast optical switch. Based on a pump-probe setup, this technique uses pump pulses of sufficient intensity to induce the switch, while the pulses to characterize are probing the transmissivity drop of the photoexcited material. This enables the characterization of low-intensity ultra-broadband pulses at the detection limit of the spectrometer, and within the transparency range of the solid. For example, by using zinc selenide (ZnSe), pulses with wavelengths from 0.5 to 20 µm can be characterized, denoting five octaves of spectral range. Using ptychography, we retrieve the temporal profiles of both the probe pulse and the switch. To demonstrate this approach, we measure ultrashort pulses from a titanium-sapphire (Ti-Sa) amplifier which are compressed using a hollow core fiber setup, as well as infrared to mid-infrared pulses generated from an optical parametric amplifier (OPA). The characterized pulses are centered at wavelengths of 0.77, 1.53, 1.75, 4, and 10 µm, down to sub-two optical cycles duration, exceeding an octave of bandwidth, and with energy as low as a few nanojoules.

show abstract

Deep learning reconstruction of ultrashort pulses

Cited by 152 publications

References 22 publications

Machine learning analysis of extreme events in optical fibre modulation instability

Machine learning analysis of extreme events in optical fibre modulation instability

Multi-Parameter Sensing in a Multimode Self-Interference Micro-Ring Resonator by Machine Learning

Phase-matching-free pulse retrieval based on transient absorption in solids

Contact Info

Product

Resources

About