LB
L. Bliek
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
16 records found
1
Real-Time Data-Driven Maintenance Logistics
A Public-Private Collaboration
Conference paper
(2024)
-
Willem van Jaarsveld, Laurens Bliek, Mathijs de Weerdt, Stella Kapodistria, Verus Pronk, Peter Verleijsdonk, Simon Voorberg, Sicco Verwer, Yingqian Zhang, More authors...
The project “Real-time data-driven maintenance logistics” was initiated with the purpose of bringing innovations in data-driven decision making to maintenance logistics, by bringing problem owners in the form of three innovative companies together with researchers at two leading knowledge institutions. This paper reviews innovations in three related areas: How the innovations were inspired by practice, how they materialized, and how the results impact practice.
...
The project “Real-time data-driven maintenance logistics” was initiated with the purpose of bringing innovations in data-driven decision making to maintenance logistics, by bringing problem owners in the form of three innovative companies together with researchers at two leading knowledge institutions. This paper reviews innovations in three related areas: How the innovations were inspired by practice, how they materialized, and how the results impact practice.
Journal article
(2022)
-
Ligia Maria Moreira Zorello, Laurens Bliek, Sebastian Troia, Tias Guns, Sicco Verwer, Guido Maier
The 5G Radio Access Network (RAN) virtualization aims to improve network quality and lower the operator's costs. One of its main features is the functional split, i.e., dividing the instantiation of RAN baseband functions into different units over metro-network nodes. However, its optimal placement is non-trivial: it depends on the application requirements and on the expected traffic volume, whose daily variation highly impacts the total power consumption. Current optimization solutions fail to provide a placement solution capable of handling traffic fluctuations. In fact, the standard machine learning algorithms used in the literature for planning the network resources in advance result in an allocation that is inadequate to carry the actual traffic at all the time-slots. Hence, we must reserve an artificial buffer capacity in the nodes to ensure feasibility. Instead, our proposed method exploits a fine-grained two-step multi-task algorithm that predicts the mean and quantile traffic, making the artificial capacity no longer necessary. The subsequent placement uses mixed-integer linear programming and a heuristic. The former considers the expected traffic in the objective function (to estimate costs) and the quantile in the constraints (to enforce capacity limits). The heuristic combines the mean and quantile results to minimize the power and comply with the requirements. While using sufficiently large artificial buffers guarantees robustness with a mild power increase compared to the oracle, the fine-grained multi-task model improves the results, reducing the power consumption compared to the mean and meets all constraints. The heuristic enables significant computational time reduction.
...
The 5G Radio Access Network (RAN) virtualization aims to improve network quality and lower the operator's costs. One of its main features is the functional split, i.e., dividing the instantiation of RAN baseband functions into different units over metro-network nodes. However, its optimal placement is non-trivial: it depends on the application requirements and on the expected traffic volume, whose daily variation highly impacts the total power consumption. Current optimization solutions fail to provide a placement solution capable of handling traffic fluctuations. In fact, the standard machine learning algorithms used in the literature for planning the network resources in advance result in an allocation that is inadequate to carry the actual traffic at all the time-slots. Hence, we must reserve an artificial buffer capacity in the nodes to ensure feasibility. Instead, our proposed method exploits a fine-grained two-step multi-task algorithm that predicts the mean and quantile traffic, making the artificial capacity no longer necessary. The subsequent placement uses mixed-integer linear programming and a heuristic. The former considers the expected traffic in the objective function (to estimate costs) and the quantile in the constraints (to enforce capacity limits). The heuristic combines the mean and quantile results to minimize the power and comply with the requirements. While using sufficiently large artificial buffers guarantees robustness with a mild power increase compared to the oracle, the fine-grained multi-task model improves the results, reducing the power consumption compared to the mean and meets all constraints. The heuristic enables significant computational time reduction.
One method to solve expensive black-box optimization problems is to use a surrogate model that approximates the objective based on previous observed evaluations. The surrogate, which is cheaper to evaluate, is optimized instead to find an approximate solution to the original problem. In the case of discrete problems, recent research has revolved around discrete surrogate models that are specifically constructed to deal with these problems. A main motivation is that literature considers continuous methods, such as Bayesian optimization with Gaussian processes as the surrogate, to be sub-optimal (especially in higher dimensions) because they ignore the discrete structure by, e.g., rounding off real-valued solutions to integers. However, we claim that this is not true. In fact, we present empirical evidence showing that the use of continuous surrogate models displays competitive performance on a set of high-dimensional discrete benchmark problems, including a real-life application, against state-of-the-art discrete surrogate-based methods. Our experiments with different kinds of discrete decision variables and time constraints also give more insight into which algorithms work well on which type of problem.
...
One method to solve expensive black-box optimization problems is to use a surrogate model that approximates the objective based on previous observed evaluations. The surrogate, which is cheaper to evaluate, is optimized instead to find an approximate solution to the original problem. In the case of discrete problems, recent research has revolved around discrete surrogate models that are specifically constructed to deal with these problems. A main motivation is that literature considers continuous methods, such as Bayesian optimization with Gaussian processes as the surrogate, to be sub-optimal (especially in higher dimensions) because they ignore the discrete structure by, e.g., rounding off real-valued solutions to integers. However, we claim that this is not true. In fact, we present empirical evidence showing that the use of continuous surrogate models displays competitive performance on a set of high-dimensional discrete benchmark problems, including a real-life application, against state-of-the-art discrete surrogate-based methods. Our experiments with different kinds of discrete decision variables and time constraints also give more insight into which algorithms work well on which type of problem.
A challenging problem in both engineering and computer science is that of minimising a function for which we have no mathematical formulation available, that is expensive to evaluate, and that contains continuous and integer variables, for example in automatic algorithm configuration. Surrogate-based algorithms are very suitable for this type of problem, but most existing techniques are designed with only continuous or only discrete variables in mind. Mixed-Variable ReLU-based Surrogate Modelling (MVRSM) is a surrogate-based algorithm that uses a linear combination of rectified linear units, defined in such a way that (local) optima satisfy the integer constraints. Unlike other methods, it also has a constant run-time per iteration. This method outperforms the state of the art on several synthetic benchmarks with up to 238 continuous and integer variables, and achieves competitive performance on two real-life benchmarks: XG-Boost hyperparameter tuning and Electrostatic Precipitator optimisation.
...
A challenging problem in both engineering and computer science is that of minimising a function for which we have no mathematical formulation available, that is expensive to evaluate, and that contains continuous and integer variables, for example in automatic algorithm configuration. Surrogate-based algorithms are very suitable for this type of problem, but most existing techniques are designed with only continuous or only discrete variables in mind. Mixed-Variable ReLU-based Surrogate Modelling (MVRSM) is a surrogate-based algorithm that uses a linear combination of rectified linear units, defined in such a way that (local) optima satisfy the integer constraints. Unlike other methods, it also has a constant run-time per iteration. This method outperforms the state of the art on several synthetic benchmarks with up to 238 continuous and integer variables, and achieves competitive performance on two real-life benchmarks: XG-Boost hyperparameter tuning and Electrostatic Precipitator optimisation.
Conference paper
(2020)
-
S.E. Verwer, A. Nadeem, C.A. Hammerschmidt, L. Bliek, Abdullah Al-Dujaili, Una-May O’Reilly
Training classifiers that are robust against adversarially modified examples is becoming increasingly important in practice. In the field of malware detection, adversaries modify malicious binary files to seem benign while preserving their malicious behavior. We report on the results of a recently held robust malware detection challenge. There were two tracks in which teams could participate: the attack track asked for adversarially modified malware samples and the defend track asked for trained neural network classifiers that are robust to such modifications. The teams were unaware of the attacks/defenses they had to detect/evade. Although only 9 teams participated, this unique setting allowed us to make several interesting observations. We also present the challenge winner: GRAMS, a family of novel techniques to train adversarially robust networks that preserve the intended (malicious) functionality and yield high-quality adversarial samples. These samples are used to iteratively train a robust classifier. We show that our techniques, based on discrete optimization techniques, beat purely gradient-based methods. GRAMS obtained first place in both the attack and defend tracks of the competition.
...
Training classifiers that are robust against adversarially modified examples is becoming increasingly important in practice. In the field of malware detection, adversaries modify malicious binary files to seem benign while preserving their malicious behavior. We report on the results of a recently held robust malware detection challenge. There were two tracks in which teams could participate: the attack track asked for adversarially modified malware samples and the defend track asked for trained neural network classifiers that are robust to such modifications. The teams were unaware of the attacks/defenses they had to detect/evade. Although only 9 teams participated, this unique setting allowed us to make several interesting observations. We also present the challenge winner: GRAMS, a family of novel techniques to train adversarially robust networks that preserve the intended (malicious) functionality and yield high-quality adversarial samples. These samples are used to iteratively train a robust classifier. We show that our techniques, based on discrete optimization techniques, beat purely gradient-based methods. GRAMS obtained first place in both the attack and defend tracks of the competition.
When a black-box optimization objective can only be evaluated with costly or noisy measurements, most standard optimization algorithms are unsuited to find the optimal solution. Specialized algorithms that deal with exactly this situation make use of surrogate models. These models are usually continuous and smooth, which is beneficial for continuous optimization problems, but not necessarily for combinatorial problems. However, by choosing the basis functions of the surrogate model in a certain way, we show that it can be guaranteed that the optimal solution of the surrogate model is integer. This approach outperforms random search, simulated annealing and a Bayesian optimization algorithm on the problem of finding robust routes for a noise-perturbed traveling salesman benchmark problem, with similar performance as another Bayesian optimization algorithm, and outperforms all compared algorithms on a convex binary optimization problem with a large number of variables.
...
When a black-box optimization objective can only be evaluated with costly or noisy measurements, most standard optimization algorithms are unsuited to find the optimal solution. Specialized algorithms that deal with exactly this situation make use of surrogate models. These models are usually continuous and smooth, which is beneficial for continuous optimization problems, but not necessarily for combinatorial problems. However, by choosing the basis functions of the surrogate model in a certain way, we show that it can be guaranteed that the optimal solution of the surrogate model is integer. This approach outperforms random search, simulated annealing and a Bayesian optimization algorithm on the problem of finding robust routes for a noise-perturbed traveling salesman benchmark problem, with similar performance as another Bayesian optimization algorithm, and outperforms all compared algorithms on a convex binary optimization problem with a large number of variables.
Journal article
(2019)
-
Laurens Bliek, Sander Wahls, Ilka Visscher, Caterina Taddei, Roelof Bernardus Timens, Ruud Oldenbeuving, Chris Roeloffzen, Michel Verhaegen
We present a novel photonic beamformer for a fully integrated transmit phased array antenna, together with an automatic procedure for tuning the delays in this system. Such an automatic tuning procedure is required because the large number of actuators makes manual tuning practically impossible. The antenna system is designed for the purpose of the broadband aircraft-satellite communication in the \mathrm{K_u}-band to provide satellite Internet connections on board the aircraft. The goal of the beamformer is to automatically steer the transmit antenna electronically in the direction of the satellite. This is done using a mix of phase shifters and tunable optical delay lines, which are all integrated on a chip and laid out in a tree structure. The \mathrm{K_u}-band has a bandwidth of 0.5 GHz. We show how an optical delay line is automatically configured over this bandwidth, providing a delay of approximately 0.4 ns. The tuning algorithm calculates the best actuator voltages based on past measurements. This is the first time that such an automatic tuning scheme is used on a photonic beamformer for this type of transmit phased array antenna. We show that the proposed method is able to provide accurate beamforming ({<}\text{11.25}^{\circ} phase error over the whole bandwidth) for two different delay settings.
...
We present a novel photonic beamformer for a fully integrated transmit phased array antenna, together with an automatic procedure for tuning the delays in this system. Such an automatic tuning procedure is required because the large number of actuators makes manual tuning practically impossible. The antenna system is designed for the purpose of the broadband aircraft-satellite communication in the \mathrm{K_u}-band to provide satellite Internet connections on board the aircraft. The goal of the beamformer is to automatically steer the transmit antenna electronically in the direction of the satellite. This is done using a mix of phase shifters and tunable optical delay lines, which are all integrated on a chip and laid out in a tree structure. The \mathrm{K_u}-band has a bandwidth of 0.5 GHz. We show how an optical delay line is automatically configured over this bandwidth, providing a delay of approximately 0.4 ns. The tuning algorithm calculates the best actuator voltages based on past measurements. This is the first time that such an automatic tuning scheme is used on a photonic beamformer for this type of transmit phased array antenna. We show that the proposed method is able to provide accurate beamforming ({<}\text{11.25}^{\circ} phase error over the whole bandwidth) for two different delay settings.
Automatic tuning of photonic beamformers
A data-driven approach
Beamforming is a signal processing technique used in highly directional antennas. An array of antenna elements transmits the same signal, but with a different time delay for each element. By providing the right time delays for each antenna element, the whole array transmits a high-powered signal in one desired direction. This technique can be used for example to provide satellite television and Internet connections on board of aircrafts. Recently, developments in the field of integrated microwave photonics have paved the way for broadband, low-loss, and low-weight beamformer systems. These photonic beamformers convert the signals to be transmitted to the optical domain, provide the correct time delays with tunable optical delay lines, and then convert the signal back to the radio frequency domain. The main challenge here lies in tuning the actuators of the tunable optical delay lines in such a way that they provide the desired time delays. Challenges like actuator crosstalk, parameter sensitivity, noise and model errors cause complications when traditional tuning algorithms are used, such as nonlinear optimization routines. All results obtained with these photonic beamformers in the literature so far have been achieved by tuning the whole system by hand, or by applying nonlinear
optimization techniques to a simplified simulation of the system rather than the actual system. In order to find a practical way of tuning a photonic beamformer in real time, this thesis takes a data-driven approach. Instead of relying on perfectly accurate physical models, a surrogate function is used that approximates the relation between the system actuators and a cost function, namely the difference between the measured and desired time delay of each antenna element. By performing nonlinear optimization techniques on this surrogate cost function and by continuously updating the approximation as new measurements are obtained, the time delays of each antenna element should converge towards the desired values. The Data-based Online Nonlinear Extremum-seeker (DONE) algorithm is used to update and optimize the surrogate function in real time. This algorithm is especially designed to optimize cost functions that are costly to evaluate (for example in terms of time), that contain noise, and for which derivatives cannot be easily computed or approximated. The DONE algorithm is applied to a simulation of a photonic beamformer and to the real system, as well as to several other applications. It is shown that the algorithm outperforms comparable methods on several fronts, especially computation time. Furthermore, the theory behind the algorithm is investigated, but practical results are also given, for example rules of thumb for choosing the hyper-parameters. Finally, variations to the DONE algorithm have been developed that are easier to use, can be implemented more efficiently, and can deal with time-varying objective functions.
...
Beamforming is a signal processing technique used in highly directional antennas. An array of antenna elements transmits the same signal, but with a different time delay for each element. By providing the right time delays for each antenna element, the whole array transmits a high-powered signal in one desired direction. This technique can be used for example to provide satellite television and Internet connections on board of aircrafts. Recently, developments in the field of integrated microwave photonics have paved the way for broadband, low-loss, and low-weight beamformer systems. These photonic beamformers convert the signals to be transmitted to the optical domain, provide the correct time delays with tunable optical delay lines, and then convert the signal back to the radio frequency domain. The main challenge here lies in tuning the actuators of the tunable optical delay lines in such a way that they provide the desired time delays. Challenges like actuator crosstalk, parameter sensitivity, noise and model errors cause complications when traditional tuning algorithms are used, such as nonlinear optimization routines. All results obtained with these photonic beamformers in the literature so far have been achieved by tuning the whole system by hand, or by applying nonlinear
optimization techniques to a simplified simulation of the system rather than the actual system. In order to find a practical way of tuning a photonic beamformer in real time, this thesis takes a data-driven approach. Instead of relying on perfectly accurate physical models, a surrogate function is used that approximates the relation between the system actuators and a cost function, namely the difference between the measured and desired time delay of each antenna element. By performing nonlinear optimization techniques on this surrogate cost function and by continuously updating the approximation as new measurements are obtained, the time delays of each antenna element should converge towards the desired values. The Data-based Online Nonlinear Extremum-seeker (DONE) algorithm is used to update and optimize the surrogate function in real time. This algorithm is especially designed to optimize cost functions that are costly to evaluate (for example in terms of time), that contain noise, and for which derivatives cannot be easily computed or approximated. The DONE algorithm is applied to a simulation of a photonic beamformer and to the real system, as well as to several other applications. It is shown that the algorithm outperforms comparable methods on several fronts, especially computation time. Furthermore, the theory behind the algorithm is investigated, but practical results are also given, for example rules of thumb for choosing the hyper-parameters. Finally, variations to the DONE algorithm have been developed that are easier to use, can be implemented more efficiently, and can deal with time-varying objective functions.
Conference paper
(2018)
-
Chris Roeloffzen, Ilka Visscher, Robert Grootjans, Laurens Bliek, Sander Wahls, Michel Verhaegen, Caterina Taddei, Dimitri Geskus, Ruud Oldenbeuving, Jörn Epping, Roelof Bernardus Timens, Paul van Dijk, René Heideman, Marcel Hoekman
This paper describes the design, fabrication, packaging, testing and automated tuning of an integrated 1x4 optical beamforming network. It consists of hybridly integrated InP and TriPleX chips, where end-facet coupling is used for optical interfacing.
...
This paper describes the design, fabrication, packaging, testing and automated tuning of an integrated 1x4 optical beamforming network. It consists of hybridly integrated InP and TriPleX chips, where end-facet coupling is used for optical interfacing.
Journal article
(2017)
-
Paolo Pozzi, Dean Wilding, Oleg Soloviev, Hans Verstraete, Laurens Bliek, Gleb Vdovine, Michel Verhaegen
The quality of fluorescence microscopy images is often impaired by the presence
of sample induced optical aberrations. Adaptive optical elements such as deformable mirrors or spatial light modulators can be used to correct aberrations. However, previously reported techniques either require special sample preparation, or time consuming optimization procedures for the correction of static aberrations. This paper reports a technique for optical sectioning fluorescence microscopy capable of correcting dynamic aberrations in any fluorescent sample during the acquisition. This is achieved by implementing adaptive optics in a non conventional confocal microscopy setup, with multiple programmable confocal apertures, in which out of focus light can be separately detected, and used to optimize the correction performance with a sampling frequency an order of magnitude faster than the imaging rate of the system. The
paper reports results comparing the correction performances to traditional image optimization algorithms, and demonstrates how the system can compensate for dynamic changes in the aberrations, such as those introduced during a focal stack acquisition though a thick sample. ...
of sample induced optical aberrations. Adaptive optical elements such as deformable mirrors or spatial light modulators can be used to correct aberrations. However, previously reported techniques either require special sample preparation, or time consuming optimization procedures for the correction of static aberrations. This paper reports a technique for optical sectioning fluorescence microscopy capable of correcting dynamic aberrations in any fluorescent sample during the acquisition. This is achieved by implementing adaptive optics in a non conventional confocal microscopy setup, with multiple programmable confocal apertures, in which out of focus light can be separately detected, and used to optimize the correction performance with a sampling frequency an order of magnitude faster than the imaging rate of the system. The
paper reports results comparing the correction performances to traditional image optimization algorithms, and demonstrates how the system can compensate for dynamic changes in the aberrations, such as those introduced during a focal stack acquisition though a thick sample. ...
The quality of fluorescence microscopy images is often impaired by the presence
of sample induced optical aberrations. Adaptive optical elements such as deformable mirrors or spatial light modulators can be used to correct aberrations. However, previously reported techniques either require special sample preparation, or time consuming optimization procedures for the correction of static aberrations. This paper reports a technique for optical sectioning fluorescence microscopy capable of correcting dynamic aberrations in any fluorescent sample during the acquisition. This is achieved by implementing adaptive optics in a non conventional confocal microscopy setup, with multiple programmable confocal apertures, in which out of focus light can be separately detected, and used to optimize the correction performance with a sampling frequency an order of magnitude faster than the imaging rate of the system. The
paper reports results comparing the correction performances to traditional image optimization algorithms, and demonstrates how the system can compensate for dynamic changes in the aberrations, such as those introduced during a focal stack acquisition though a thick sample.
of sample induced optical aberrations. Adaptive optical elements such as deformable mirrors or spatial light modulators can be used to correct aberrations. However, previously reported techniques either require special sample preparation, or time consuming optimization procedures for the correction of static aberrations. This paper reports a technique for optical sectioning fluorescence microscopy capable of correcting dynamic aberrations in any fluorescent sample during the acquisition. This is achieved by implementing adaptive optics in a non conventional confocal microscopy setup, with multiple programmable confocal apertures, in which out of focus light can be separately detected, and used to optimize the correction performance with a sampling frequency an order of magnitude faster than the imaging rate of the system. The
paper reports results comparing the correction performances to traditional image optimization algorithms, and demonstrates how the system can compensate for dynamic changes in the aberrations, such as those introduced during a focal stack acquisition though a thick sample.
Web publication
(2017)
-
H.R.G.W. Verstraete, Morgan Heisler, Meyeong Jin Ju, Daniel J. Wahl, L. Bliek, Jeroen Kalkman, Stefano Bonora, Marinko V. Sarunic, Michel Verhaegen, Yifan Jian
Optical Coherence Tomography (OCT) has revolutionized modern ophthalmology, providing depth resolved images of the retinal layers in a system that is suited to a clinical environment. A limitation of the performance and utilization of the OCT systems has been the lateral resolution. Through the combination of wavefront sensorless adaptive optics with dual variable optical elements, we present a compact lens based OCT system that is capable of imaging the photoreceptor mosaic. We utilized a commercially available variable focal length lens to correct for a wide range of defocus commonly found in patient eyes, and a multi-actuator adaptive lens after linearization of the hysteresis in the piezoelectric actuators for aberration correction to obtain near diffraction limited imaging at the retina. A parallel processing computational platform permitted real-time image acquisition and display. The Data-based Online Nonlinear Extremum seeker (DONE) algorithm was used for real time optimization of the wavefront sensorless adaptive optics OCT, and the performance was compared with a coordinate search algorithm. Cross sectional images of the retinal layers and en face images of the cone photoreceptor mosaic acquired in vivo from research volunteers before and after WSAO optimization are presented. Applying the DONE algorithm in vivo for wavefront sensorless AO-OCT demonstrates that the DONE algorithm succeeds in drastically improving the signal while achieving a computational time of 1 ms per iteration, making it applicable for high speed real time applications.
...
Optical Coherence Tomography (OCT) has revolutionized modern ophthalmology, providing depth resolved images of the retinal layers in a system that is suited to a clinical environment. A limitation of the performance and utilization of the OCT systems has been the lateral resolution. Through the combination of wavefront sensorless adaptive optics with dual variable optical elements, we present a compact lens based OCT system that is capable of imaging the photoreceptor mosaic. We utilized a commercially available variable focal length lens to correct for a wide range of defocus commonly found in patient eyes, and a multi-actuator adaptive lens after linearization of the hysteresis in the piezoelectric actuators for aberration correction to obtain near diffraction limited imaging at the retina. A parallel processing computational platform permitted real-time image acquisition and display. The Data-based Online Nonlinear Extremum seeker (DONE) algorithm was used for real time optimization of the wavefront sensorless adaptive optics OCT, and the performance was compared with a coordinate search algorithm. Cross sectional images of the retinal layers and en face images of the cone photoreceptor mosaic acquired in vivo from research volunteers before and after WSAO optimization are presented. Applying the DONE algorithm in vivo for wavefront sensorless AO-OCT demonstrates that the DONE algorithm succeeds in drastically improving the signal while achieving a computational time of 1 ms per iteration, making it applicable for high speed real time applications.
We propose CDONE, a convex version of the DONE algorithm. DONE is a derivative-free online optimization algorithm that uses surrogate modeling with noisy measurements to find a minimum of objective functions that are expensive to evaluate. Inspired by their success in deep learning, CDONE makes use of rectified linear units, together with a nonnegativity constraint to enforce convexity of the surrogate model. This leads to a sparse and cheap to evaluate surrogate model of the unknown optimization objective that is still accurate and that can be minimized with convex optimization algorithms. The CDONE algorithm is demonstrated on a toy example and on the problem of hyper-parameter optimization for a deep learning example on handwritten digit classification.
...
We propose CDONE, a convex version of the DONE algorithm. DONE is a derivative-free online optimization algorithm that uses surrogate modeling with noisy measurements to find a minimum of objective functions that are expensive to evaluate. Inspired by their success in deep learning, CDONE makes use of rectified linear units, together with a nonnegativity constraint to enforce convexity of the surrogate model. This leads to a sparse and cheap to evaluate surrogate model of the unknown optimization objective that is still accurate and that can be minimized with convex optimization algorithms. The CDONE algorithm is demonstrated on a toy example and on the problem of hyper-parameter optimization for a deep learning example on handwritten digit classification.
Web publication
(2017)
-
Yifan Jian, H.R.G.W. Verstraete, Morgan Heisler, Meyeong Jin Ju, Daniel J. Wahl, L. Bliek, Jeroen Kalkman, Stefano Bonora, Michel Verhaegen, Marinko V. Sarunic
Adaptive optics has been successfully applied to cellular resolution imaging of the retina, enabling visualization of the characteristic mosaic patterns of the outer retina. Wavefront sensorless adaptive optics (WSAO) is a novel technique that facilitates high resolution ophthalmic imaging; it replaces the Hartmann-Shack Wavefront Sensor with an image-driven optimization algorithm and mitigates some the challenges encountered with sensor-based designs. However, WSAO generally requires longer time to perform aberrations correction than the conventional closed-loop adaptive optics. When used for in vivo retinal imaging applications, motion artifacts during the WSAO optimization process will affect the quality of the aberration correction. A faster converging optimization scheme needs to be developed to account for rapid temporal variation of the wavefront and continuously apply corrections. In this project, we investigate the Databased Online Nonlinear Extremum-seeker (DONE), a novel non-linear multivariate optimization algorithm in combination with in vivo human WSAO OCT imaging. We also report both hardware and software updates of our compact lens based WSAO 1060nm swept source OCT human retinal imaging system, including real time retinal layer segmentation and tracking (ILM and RPE), hysteresis correction for the multi-actuator adaptive lens, precise synchronization control for the 200kHz laser source, and a zoom lens unit for rapid switching of the field of view. Cross sectional images of the retinal layers and en face images of the cone photoreceptor mosaic acquired in vivo from research volunteers before and after WSAO optimization are presented.
...
Adaptive optics has been successfully applied to cellular resolution imaging of the retina, enabling visualization of the characteristic mosaic patterns of the outer retina. Wavefront sensorless adaptive optics (WSAO) is a novel technique that facilitates high resolution ophthalmic imaging; it replaces the Hartmann-Shack Wavefront Sensor with an image-driven optimization algorithm and mitigates some the challenges encountered with sensor-based designs. However, WSAO generally requires longer time to perform aberrations correction than the conventional closed-loop adaptive optics. When used for in vivo retinal imaging applications, motion artifacts during the WSAO optimization process will affect the quality of the aberration correction. A faster converging optimization scheme needs to be developed to account for rapid temporal variation of the wavefront and continuously apply corrections. In this project, we investigate the Databased Online Nonlinear Extremum-seeker (DONE), a novel non-linear multivariate optimization algorithm in combination with in vivo human WSAO OCT imaging. We also report both hardware and software updates of our compact lens based WSAO 1060nm swept source OCT human retinal imaging system, including real time retinal layer segmentation and tracking (ILM and RPE), hysteresis correction for the multi-actuator adaptive lens, precise synchronization control for the 200kHz laser source, and a zoom lens unit for rapid switching of the field of view. Cross sectional images of the retinal layers and en face images of the cone photoreceptor mosaic acquired in vivo from research volunteers before and after WSAO optimization are presented.
Conference paper
(2017)
-
Laurens Bliek, Hans Verstraete, Sander Wahls, Roelof Bernardus Timens, Ruud Oldenbeuving, Chris Roeloffzen, Michel Verhaegen
In recent years we have seen a rise in the amount of fitness tracking and self monitoring devices. These devices which often work in conjunction with a smartphone are becoming more accurate and are becoming widely adopted. This trend goes hand in hand with Electronic Health Care (e-health): the shift of health care to the digital domain. E-health would allow patients to measure their medical condition at home, allowing a diagnosis to be made based on measurements taken over a longer period of time, while reducing the work performed by a doctor. Measurements are tored in the cloud, simplifying the way in which they can be shared with healthcare providers and possibly research nstitutions. Modernizing healthcare this way should give the patient more insight and control over his/her healthcare and medical data. Furthermore the amount of visits required to the hospital can be reduced, an endeavor which can be demanding for many less fit for elderly individuals.
However, handling medical data this way causes concern for privacy. Often the data handled by these devices is very sensitive and could easily be used to identify the user and monitor many of their behaviours. In order to achieve privacy there are several approaches. One way is to enforce involved parties through legislation to use the data for specific purposes only. However, this relies on the party being semi-trusted and does not guarantee safety in case of a data-breach.
In this work the way in which the integration of wearables into the medical domain is currently taking place and how privacy and security is handled will be explored. Furthermore we will show the current state of research regarding improving this security.
...
In recent years we have seen a rise in the amount of fitness tracking and self monitoring devices. These devices which often work in conjunction with a smartphone are becoming more accurate and are becoming widely adopted. This trend goes hand in hand with Electronic Health Care (e-health): the shift of health care to the digital domain. E-health would allow patients to measure their medical condition at home, allowing a diagnosis to be made based on measurements taken over a longer period of time, while reducing the work performed by a doctor. Measurements are tored in the cloud, simplifying the way in which they can be shared with healthcare providers and possibly research nstitutions. Modernizing healthcare this way should give the patient more insight and control over his/her healthcare and medical data. Furthermore the amount of visits required to the hospital can be reduced, an endeavor which can be demanding for many less fit for elderly individuals.
However, handling medical data this way causes concern for privacy. Often the data handled by these devices is very sensitive and could easily be used to identify the user and monitor many of their behaviours. In order to achieve privacy there are several approaches. One way is to enforce involved parties through legislation to use the data for specific purposes only. However, this relies on the party being semi-trusted and does not guarantee safety in case of a data-breach.
In this work the way in which the integration of wearables into the medical domain is currently taking place and how privacy and security is handled will be explored. Furthermore we will show the current state of research regarding improving this security.
Journal article
(2017)
-
Hans Verstraete, M. Heisler, M.J. Ju, D. Wahl, Laurens Bliek, Jeroen Kalkman, S. Bonora, Y Jian, Michel Verhaegen, M.V. Sarunic
In this report, which is an international collaboration of OCT, adaptive optics, and control research, we demonstrate the data-based online nonlinear extremum-seeker (DONE) algorithm to guide the image based optimization for wavefront sensorless adaptive optics (WFSL-AO) OCT for in vivo human retinal imaging. The ocular aberrations were corrected using a multi-actuator adaptive lens after linearization of the hysteresis in the piezoelectric actuators. The DONE algorithm succeeded in drastically improving image quality and the OCT signal intensity, up to a factor seven, while achieving a computational time of 1 ms per iteration, making it applicable for many high speed applications. We demonstrate the correction of five aberrations using 70 iterations of the DONE algorithm performed over 2.8 s of continuous volumetric OCT acquisition. Data acquired from an imaging phantom and in vivo from human research volunteers are presented.
...
In this report, which is an international collaboration of OCT, adaptive optics, and control research, we demonstrate the data-based online nonlinear extremum-seeker (DONE) algorithm to guide the image based optimization for wavefront sensorless adaptive optics (WFSL-AO) OCT for in vivo human retinal imaging. The ocular aberrations were corrected using a multi-actuator adaptive lens after linearization of the hysteresis in the piezoelectric actuators. The DONE algorithm succeeded in drastically improving image quality and the OCT signal intensity, up to a factor seven, while achieving a computational time of 1 ms per iteration, making it applicable for many high speed applications. We demonstrate the correction of five aberrations using 70 iterations of the DONE algorithm performed over 2.8 s of continuous volumetric OCT acquisition. Data acquired from an imaging phantom and in vivo from human research volunteers are presented.
This paper analyzes data-based online nonlinear extremum-seeker (DONE), an online optimization algorithm that iteratively minimizes an unknown function based on costly and noisy measurements. The algorithm maintains a surrogate of the unknown function in the form of a random Fourier expansion. The surrogate is updated whenever a new measurement is available, and then used to determine the next measurement point. The algorithm is comparable with Bayesian optimization algorithms, but its computational complexity per iteration does not depend on the number of measurements. We derive several theoretical results that provide insight on how the hyperparameters of the algorithm should be chosen. The algorithm is compared with a Bayesian optimization algorithm for an analytic benchmark problem and three applications, namely, optical coherence tomography, optical beam-forming network tuning, and robot arm control. It is found that the DONE algorithm is significantly faster than Bayesian optimization in the discussed problems while achieving a similar or better performance.
...
This paper analyzes data-based online nonlinear extremum-seeker (DONE), an online optimization algorithm that iteratively minimizes an unknown function based on costly and noisy measurements. The algorithm maintains a surrogate of the unknown function in the form of a random Fourier expansion. The surrogate is updated whenever a new measurement is available, and then used to determine the next measurement point. The algorithm is comparable with Bayesian optimization algorithms, but its computational complexity per iteration does not depend on the number of measurements. We derive several theoretical results that provide insight on how the hyperparameters of the algorithm should be chosen. The algorithm is compared with a Bayesian optimization algorithm for an analytic benchmark problem and three applications, namely, optical coherence tomography, optical beam-forming network tuning, and robot arm control. It is found that the DONE algorithm is significantly faster than Bayesian optimization in the discussed problems while achieving a similar or better performance.