PhD defence public presentation, Bayesian methods for inverse problems with point clouds: applications to single-photon lidar, ENSEEHIT, Toulouse, France
Compressive sampling (CS) aims at acquiring a signal at a sampling rate below the Nyquist rate by exploiting prior knowledge that a signal is sparse or correlated in some domain. Despite the remarkable progress in the theory of CS, the sampling rate on a single image required by CS is still very high in practice. In this presentation, a non-local compressive sampling (NLCS) recovery method is proposed to further reduce the sampling rate by exploiting non-local patch correlation and local piecewise smoothness present in natural images. Two non-local sparsity measures, i.e., non-local wavelet sparsity and non-local joint sparsity, are proposed to exploit the patch correlation in NLCS. An efficient iterative algorithm is developed to solve the NLCS recovery problem, which is shown to have stable convergence behavior in experiments. The experimental results show that our NLCS significantly improves the state-of-the-art of image compressive sampling.
PhD defence public presentation, Bayesian methods for inverse problems with point clouds: applications to single-photon lidar, ENSEEHIT, Toulouse, France
Compressive sampling (CS) aims at acquiring a signal at a sampling rate below the Nyquist rate by exploiting prior knowledge that a signal is sparse or correlated in some domain. Despite the remarkable progress in the theory of CS, the sampling rate on a single image required by CS is still very high in practice. In this presentation, a non-local compressive sampling (NLCS) recovery method is proposed to further reduce the sampling rate by exploiting non-local patch correlation and local piecewise smoothness present in natural images. Two non-local sparsity measures, i.e., non-local wavelet sparsity and non-local joint sparsity, are proposed to exploit the patch correlation in NLCS. An efficient iterative algorithm is developed to solve the NLCS recovery problem, which is shown to have stable convergence behavior in experiments. The experimental results show that our NLCS significantly improves the state-of-the-art of image compressive sampling.
DTAM: Dense Tracking and Mapping in Real-Time, Robot vision GroupLihang Li
This is the slides about DTAM for my group meeting report, hope it does help to anyone who will want to implement DTAM and need to understand it deeply.
Ill-posedness formulation of the emission source localization in the radio- d...Ahmed Ammar Rebai PhD
To contact the authors : tarek.salhi@gmail.com and ahmed.rebai2@gmail.com
In the field of radio detection in astroparticle physics, many studies have shown the strong dependence of the solution of the radio-transient sources localization problem (the radio-shower time of arrival on antennas) such solutions are purely numerical artifacts. Based on a detailed analysis of some already published results of radio-detection experiments like : CODALEMA 3 in France, AERA in Argentina and TREND in China, we demonstrate the ill-posed character of this problem in the sens of Hadamard. Two approaches have been used as the existence of solutions degeneration and the bad conditioning of the mathematical formulation problem. A comparison between experimental results and simulations have been made, to highlight the mathematical studies. Many properties of the non-linear least square function are discussed such as the configuration of the set of solutions and the bias.
ARRAY FACTOR OPTIMIZATION OF AN ACTIVE PLANAR PHASED ARRAY USING EVOLUTIONARY...jantjournal
Evolutionary algorithms (EAs) have the potential to handle complex, multi-dimensional optimization problems in the field of phased array. Out of different EAs, particle swarm optimization (PSO) is a popular choice. In a phased array, antenna element failure is a common phenomenon and this leads to degradation
of the array factor (AF) pattern, primarily in terms of increased side lobe levels (SLLs), displacement of nulls and reduction in the null depths. The recovery of a degraded pattern using a cost and time-effective approach is on demand. In this context, an attempt made to obtain an optimized AF pattern after fault in a
49 elements quasi-circular aperture equilateral triangular grid active planar phased array using PSO. In the paper, multiple cases on recovery are discussed having a maximum 20% element failure. Each recovery is also further evaluated by different statistical analyses. A dedicated software tool was developed to carry out the work presented in this paper.
Space-time adaptive processing (STAP) is a signal
processing technique most commonly used in radar systems where
interference is a problem. The radar signal processor is used to
remove the unintentional cluttering effects caused by ground
reflections and echoes due to sea, desert, forest, etc. and intentional
jamming and make the received signal useful. In this paper a new
approach to STAP based on subspace projection has been described
in detail. According to linear algebra and three dimensional
geometry, if we project a range space on to a subspace spanned by
linearly independent vectors, we can suppress data which is
perpendicular to that subspace. In subspace based technique, the
received data is projected on to a subspace which is orthogonal to
clutter subspace to remove the clutter. The probability of target
detection can be find out in order to analyse the performance of the
proposed algorithm. Two existing algorithms, SMI and DPCA are
chosen to do the comparison. while plotting the detection Probability
against SINR , the results obtained are better for subspace technique
than DPCA and SMI. We got the SINR improved for subspace based
technique for same detection probability. The effect of subspace rank
on SINR was also analysed for understanding the computational load
caused by the technique. We also analysed the convergence of the
algorithm by taking plots of SINR against range snapshots.
Here is a new 9-point scheme for finite difference solution of acoustic waves in frequency domain. The algorithm honors both accuracy and computational efficiency.
We investigate the use of stochastic parametrization to account for model errors due to sub-grid scales in data assimilation of chaotic systems. Using data from fine simulations of the system, the stochastic parametrization leads to a non-Markovian model that captures the ket statistical and dynamical properties of the full system. The non-Markovian model can then be used in data assimilation algorithms to improve the performance of state estimation and prediction. Tests on the two-layer Lorenz 96 model show that such a non-Markovian stochastic parametrization approach improves data assimilation, and it outperforms the techniques of localization and inflation in the ensemble Kalman filter with perturbed observations.
Abstract : Motivated by the recovery and prediction of electricity consumption time series, we extend Nonnegative Matrix Factorization to take into account external features as side information. We consider general linear measurement settings, and propose a framework which models non-linear relationships between external features and the response variable. We extend previous theoretical results to obtain a sufficient condition on the identifiability of NMF with side information. Based on the classical Hierarchical Alternating Least Squares (HALS) algorithm, we propose a new algorithm (HALSX, or Hierarchical Alternating Least Squares with eXogeneous variables) which estimates NMF in this setting. The algorithm is validated on both simulated and real electricity consumption datasets as well as a recommendation system dataset, to show its performance in matrix recovery and prediction for new rows and columns.
Many algorithms have been developed to find sparse representation over redundant dictionaries or
transform. This paper presents a novel method on compressive sensing (CS)-based image compression
using sparse basis on CDF9/7 wavelet transform. The measurement matrix is applied to the three levels of
wavelet transform coefficients of the input image for compressive sampling. We have used three different
measurement matrix as Gaussian matrix, Bernoulli measurement matrix and random orthogonal matrix.
The orthogonal matching pursuit (OMP) and Basis Pursuit (BP) are applied to reconstruct each level of
wavelet transform separately. Experimental results demonstrate that the proposed method given better
quality of compressed image than existing methods in terms of proposed image quality evaluation indexes
and other objective (PSNR/UIQI/SSIM) measurements.
A walk through the intersection between machine learning and mechanistic mode...JuanPabloCarbajal3
Talk at EURECOM, France.
It overviews regression in several of its forms: regularized, constrained, and mixed. It builds the bridge between machine learning and dynamical models.
Robust Image Denoising in RKHS via Orthogonal Matching PursuitPantelis Bouboulis
We present a robust method for the image denoising task based on kernel ridge regression and sparse modeling. Added noise is assumed to consist of two parts. One part is impulse noise assumed to be sparse (outliers), while the other part is bounded noise. The noisy image is divided into small regions of interest, whose pixels are regarded as points of a two-dimensional surface. A kernel based ridge regression method, whose parameters are selected adaptively, is employed to fit the data, whereas the outliers are detected via the use of the increasingly popular orthogonal matching pursuit (OMP) algorithm. To this end, a new variant of the OMP rationale is employed that has the additional advantage to automatically terminate, when all outliers have been selected.
DTAM: Dense Tracking and Mapping in Real-Time, Robot vision GroupLihang Li
This is the slides about DTAM for my group meeting report, hope it does help to anyone who will want to implement DTAM and need to understand it deeply.
Ill-posedness formulation of the emission source localization in the radio- d...Ahmed Ammar Rebai PhD
To contact the authors : tarek.salhi@gmail.com and ahmed.rebai2@gmail.com
In the field of radio detection in astroparticle physics, many studies have shown the strong dependence of the solution of the radio-transient sources localization problem (the radio-shower time of arrival on antennas) such solutions are purely numerical artifacts. Based on a detailed analysis of some already published results of radio-detection experiments like : CODALEMA 3 in France, AERA in Argentina and TREND in China, we demonstrate the ill-posed character of this problem in the sens of Hadamard. Two approaches have been used as the existence of solutions degeneration and the bad conditioning of the mathematical formulation problem. A comparison between experimental results and simulations have been made, to highlight the mathematical studies. Many properties of the non-linear least square function are discussed such as the configuration of the set of solutions and the bias.
ARRAY FACTOR OPTIMIZATION OF AN ACTIVE PLANAR PHASED ARRAY USING EVOLUTIONARY...jantjournal
Evolutionary algorithms (EAs) have the potential to handle complex, multi-dimensional optimization problems in the field of phased array. Out of different EAs, particle swarm optimization (PSO) is a popular choice. In a phased array, antenna element failure is a common phenomenon and this leads to degradation
of the array factor (AF) pattern, primarily in terms of increased side lobe levels (SLLs), displacement of nulls and reduction in the null depths. The recovery of a degraded pattern using a cost and time-effective approach is on demand. In this context, an attempt made to obtain an optimized AF pattern after fault in a
49 elements quasi-circular aperture equilateral triangular grid active planar phased array using PSO. In the paper, multiple cases on recovery are discussed having a maximum 20% element failure. Each recovery is also further evaluated by different statistical analyses. A dedicated software tool was developed to carry out the work presented in this paper.
Space-time adaptive processing (STAP) is a signal
processing technique most commonly used in radar systems where
interference is a problem. The radar signal processor is used to
remove the unintentional cluttering effects caused by ground
reflections and echoes due to sea, desert, forest, etc. and intentional
jamming and make the received signal useful. In this paper a new
approach to STAP based on subspace projection has been described
in detail. According to linear algebra and three dimensional
geometry, if we project a range space on to a subspace spanned by
linearly independent vectors, we can suppress data which is
perpendicular to that subspace. In subspace based technique, the
received data is projected on to a subspace which is orthogonal to
clutter subspace to remove the clutter. The probability of target
detection can be find out in order to analyse the performance of the
proposed algorithm. Two existing algorithms, SMI and DPCA are
chosen to do the comparison. while plotting the detection Probability
against SINR , the results obtained are better for subspace technique
than DPCA and SMI. We got the SINR improved for subspace based
technique for same detection probability. The effect of subspace rank
on SINR was also analysed for understanding the computational load
caused by the technique. We also analysed the convergence of the
algorithm by taking plots of SINR against range snapshots.
Here is a new 9-point scheme for finite difference solution of acoustic waves in frequency domain. The algorithm honors both accuracy and computational efficiency.
We investigate the use of stochastic parametrization to account for model errors due to sub-grid scales in data assimilation of chaotic systems. Using data from fine simulations of the system, the stochastic parametrization leads to a non-Markovian model that captures the ket statistical and dynamical properties of the full system. The non-Markovian model can then be used in data assimilation algorithms to improve the performance of state estimation and prediction. Tests on the two-layer Lorenz 96 model show that such a non-Markovian stochastic parametrization approach improves data assimilation, and it outperforms the techniques of localization and inflation in the ensemble Kalman filter with perturbed observations.
Abstract : Motivated by the recovery and prediction of electricity consumption time series, we extend Nonnegative Matrix Factorization to take into account external features as side information. We consider general linear measurement settings, and propose a framework which models non-linear relationships between external features and the response variable. We extend previous theoretical results to obtain a sufficient condition on the identifiability of NMF with side information. Based on the classical Hierarchical Alternating Least Squares (HALS) algorithm, we propose a new algorithm (HALSX, or Hierarchical Alternating Least Squares with eXogeneous variables) which estimates NMF in this setting. The algorithm is validated on both simulated and real electricity consumption datasets as well as a recommendation system dataset, to show its performance in matrix recovery and prediction for new rows and columns.
Many algorithms have been developed to find sparse representation over redundant dictionaries or
transform. This paper presents a novel method on compressive sensing (CS)-based image compression
using sparse basis on CDF9/7 wavelet transform. The measurement matrix is applied to the three levels of
wavelet transform coefficients of the input image for compressive sampling. We have used three different
measurement matrix as Gaussian matrix, Bernoulli measurement matrix and random orthogonal matrix.
The orthogonal matching pursuit (OMP) and Basis Pursuit (BP) are applied to reconstruct each level of
wavelet transform separately. Experimental results demonstrate that the proposed method given better
quality of compressed image than existing methods in terms of proposed image quality evaluation indexes
and other objective (PSNR/UIQI/SSIM) measurements.
A walk through the intersection between machine learning and mechanistic mode...JuanPabloCarbajal3
Talk at EURECOM, France.
It overviews regression in several of its forms: regularized, constrained, and mixed. It builds the bridge between machine learning and dynamical models.
Robust Image Denoising in RKHS via Orthogonal Matching PursuitPantelis Bouboulis
We present a robust method for the image denoising task based on kernel ridge regression and sparse modeling. Added noise is assumed to consist of two parts. One part is impulse noise assumed to be sparse (outliers), while the other part is bounded noise. The noisy image is divided into small regions of interest, whose pixels are regarded as points of a two-dimensional surface. A kernel based ridge regression method, whose parameters are selected adaptively, is employed to fit the data, whereas the outliers are detected via the use of the increasingly popular orthogonal matching pursuit (OMP) algorithm. To this end, a new variant of the OMP rationale is employed that has the additional advantage to automatically terminate, when all outliers have been selected.
Regularized Compression of A Noisy Blurred Image ijcsa
Both regularization and compression are important issues in image processing and have been widely
approached in the literature. The usual procedure to obtain the compression of an image given through a
noisy blur requires two steps: first a deblurring step of the image and then a factorization step of the
regularized image to get an approximation in terms of low rank nonnegative factors. We examine here the
possibility of swapping the two steps by deblurring directly the noisy factors or partially denoised factors.
The experimentation shows that in this way images with comparable regularized compression can be
obtained with a lower computational cost.
Detection of Neural Activities in FMRI Using Jensen-Shannon DivergenceCSCJournals
In this paper, we present a statistical technique based on Jensen-Shanon divergence for detecting the regions of activity in fMRI images. The method is model free and we exploit the metric property of the square root of Jensen-Shannon divergence to accumulate the variations between successive time frames of fMRI images. Theoretically and experimentally we show the effectiveness of our algorithm.
The retrieval algorithms in remote sensing generally involve complex physical forward models that are nonlinear and computationally expensive to evaluate. Statistical emulation provides an alternative with cheap computation and can be used to calibrate model parameters and to improve computational efficiency of the retrieval algorithms. We introduce a framework of combining dimension reduction of input and output spaces and Gaussian process emulation
technique. The functional principal component analysis (FPCA) is chosen to reduce to the output space of thousands of dimensions by orders of magnitude. In addition, instead of making restrictive assumptions regarding the correlation structure of the high-dimensional input space,
we identity and exploit the most important directions of this space and thus construct a Gaussian process emulator with feasible computation. We will present preliminary results obtained from applying our method to OCO-2 data, and discuss how our framework can be generalized in
distributed systems. This is joint work with Jon Hobbs, Alex Konomi, Pulong Ma, and Anirban Mondal, and Joon Jin Song.
The MAIN CONTRIBUTION is an on-line heuristic law to set the training process and to modify the NN topology based on the Levenberg-Marquardt method.
An Area Predictor Filter using nonlinear autoregressive model based on neural networks for time series forecasting is introduced.
The core of the proposal is to analyze the roughness (long or short term stochastic dependence) of time series evaluated by the Hurst parameter (H).
The proposed law adapts in real time the topology of the filter at each stage of time series, changing the number of pattern, the number of iterations and the input vector length.
The main results show a good performance of the predictor, considering in particular to time series whose H parameter has a high roughness of signal, which is evaluated by HS and HA, respectively.
These results encouraged to continue working on new adjustment algorithms for time series modeling natural phenomena.
Data Driven Choice of Threshold in Cepstrum Based Spectrum Estimatesipij
The technique of cepstrum thresholding, which is shown to be an effective, yet simple, way of obtaining a smoothed non parametric spectrum estimate of a stationary signal. The major problem of this method is the choice of the threshold value for variance reduction of spectrum estimates. This paper proposes a new threshold selection method which is based on cross validation schemes such as Leave-One-Out, LeaveTwo-Out and Leave-Half-Out. This new methods are easy to describe, simple to implement, and does not impose severe conditions on the unknown spectrum. Numerical results suggest that this new methods are shown to be in agreement with those obtained when the spectrum is fully known.
ACEP Magazine edition 4th launched on 05.06.2024Rahul
This document provides information about the third edition of the magazine "Sthapatya" published by the Association of Civil Engineers (Practicing) Aurangabad. It includes messages from current and past presidents of ACEP, memories and photos from past ACEP events, information on life time achievement awards given by ACEP, and a technical article on concrete maintenance, repairs and strengthening. The document highlights activities of ACEP and provides a technical educational article for members.
Water billing management system project report.pdfKamal Acharya
Our project entitled “Water Billing Management System” aims is to generate Water bill with all the charges and penalty. Manual system that is employed is extremely laborious and quite inadequate. It only makes the process more difficult and hard.
The aim of our project is to develop a system that is meant to partially computerize the work performed in the Water Board like generating monthly Water bill, record of consuming unit of water, store record of the customer and previous unpaid record.
We used HTML/PHP as front end and MYSQL as back end for developing our project. HTML is primarily a visual design environment. We can create a android application by designing the form and that make up the user interface. Adding android application code to the form and the objects such as buttons and text boxes on them and adding any required support code in additional modular.
MySQL is free open source database that facilitates the effective management of the databases by connecting them to the software. It is a stable ,reliable and the powerful solution with the advanced features and advantages which are as follows: Data Security.MySQL is free open source database that facilitates the effective management of the databases by connecting them to the software.
Low power architecture of logic gates using adiabatic techniquesnooriasukmaningtyas
The growing significance of portable systems to limit power consumption in ultra-large-scale-integration chips of very high density, has recently led to rapid and inventive progresses in low-power design. The most effective technique is adiabatic logic circuit design in energy-efficient hardware. This paper presents two adiabatic approaches for the design of low power circuits, modified positive feedback adiabatic logic (modified PFAL) and the other is direct current diode based positive feedback adiabatic logic (DC-DB PFAL). Logic gates are the preliminary components in any digital circuit design. By improving the performance of basic gates, one can improvise the whole system performance. In this paper proposed circuit design of the low power architecture of OR/NOR, AND/NAND, and XOR/XNOR gates are presented using the said approaches and their results are analyzed for powerdissipation, delay, power-delay-product and rise time and compared with the other adiabatic techniques along with the conventional complementary metal oxide semiconductor (CMOS) designs reported in the literature. It has been found that the designs with DC-DB PFAL technique outperform with the percentage improvement of 65% for NOR gate and 7% for NAND gate and 34% for XNOR gate over the modified PFAL techniques at 10 MHz respectively.
Understanding Inductive Bias in Machine LearningSUTEJAS
This presentation explores the concept of inductive bias in machine learning. It explains how algorithms come with built-in assumptions and preferences that guide the learning process. You'll learn about the different types of inductive bias and how they can impact the performance and generalizability of machine learning models.
The presentation also covers the positive and negative aspects of inductive bias, along with strategies for mitigating potential drawbacks. We'll explore examples of how bias manifests in algorithms like neural networks and decision trees.
By understanding inductive bias, you can gain valuable insights into how machine learning models work and make informed decisions when building and deploying them.
6th International Conference on Machine Learning & Applications (CMLA 2024)ClaraZara1
6th International Conference on Machine Learning & Applications (CMLA 2024) will provide an excellent international forum for sharing knowledge and results in theory, methodology and applications of on Machine Learning & Applications.
Fundamentals of Electric Drives and its applications.pptx
Bayesian restoration of high-dimensional photon-starved images
1. Bayesian restoration of high-dimensional
photon-starved images
J. Tachella1, Y. Altmann1, M. Pereyra1, J.-Y. Tourneret2
and S. McLaughlin1
1Heriot-Watt University, Edinburgh, UK
2INP-ENSEEHIT-IRIT-T´eSA, Toulouse, France
European Signal Processing Conference, Rome
September 4, 2018
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 1 / 29
2. Outline
In this work, we compare 6 state-of-the-art Monte Carlo samplers in the
context of Poisson image restoration.
A Bayesian model for restoration of images with Poisson noise.
Appropriate posterior statistics.
We evaluate the performance of the samplers in a series of
experiments, varying the image size and the total number of collected
photons.
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 2 / 29
3. Photon-limited imaging
Photon-limited imaging arises in many applications where the light
flux changes rapidly or is extremely limited.
The measurements generally follow Poisson statistics.
Examples:
Lidar [1]
Medical imaging [2]
Astronomy [3]
1 Altmann et al. (2016) Lidar waveform-based analysis of depth images constructed using sparse single-photon data, IEEE
Trans. on Image Proc., 25(5), 1935-1946
2 Willett et al. (2003) Platelets: a multiscale approach for recovering edges and surfaces in photon-limited medical
imaging, IEEE Trans. on Medical Imag., 22(3), 332-350.
3 Zhang et al. (2008) Wavelets, ridgelets, and curvelets for Poisson noise removal, IEEE Trans. on Image Proc., 17(7),
1093-1108.
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 3 / 29
4. Observational model
In this work, we study the following inverse problem:
y|z ∼ P(Az)
K discrete measurements: y ∈ ZK
+
N-pixel image to recover: z ∈ RN
+
Linear operator A ∈ RK×N
+ , generally rank-deficient
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 4 / 29
5. A low-photon toy example (104
photons)
In this example, A is a 2 × 2 blur
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 5 / 29
6. Bayesian approach
In the Bayesian paradigm, we are interested in the posterior distribution
of the unknown image z given the measurements y, i.e.,
p(z|y, λ) ∝ p(y|z)p(z|λ), where
p(y|z) is the likelihood (or data fidelity term)
p(z|λ) is the prior distribution, which contains the a priori
assumptions on the image z (sparsity under some basis, smoothness,
etc).
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 6 / 29
7. Bayesian model
Hierarchical formulation:
y|z ∼ P(Az)
z|λ ∼ LN(0, (λP)−1
)
The log-normal Gaussian Markov random field prior with P = D D,
where D is the 2D laplacian filter promotes smooth and positive
images z.
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 7 / 29
8. Posterior distribution
Following Bayes theorem, the posterior distribution is
p(z|y, λ) =
p(y|z)p(z|λ)
p(y|z)p(z|λ)dz
∝ exp(−Uy (z))
where the negative log-posterior density (or energy) is
Uy (z) = 1K Az − y log(Az) +
λ
2
log(z) P log(z) + 1N log(z)
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 8 / 29
9. Inference
The convex optimization approach finds the MAP estimate
ˆzMAP = arg max
z
p(z|y, λ) = arg min
z
Uy (z). (1)
However, when we have very few photons...
0 1 2 3 4 5
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 9 / 29
10. Inference
To alleviate these problems, we consider other statistics of p(z|y, λ).
In particular, we compute numerical estimates of:
The minimum mean squared error estimator (MMSE) of zn [4]
E{zn|y, λ} =
+∞
0
znp(zn|y, λ)dzn
The marginal posterior variance
var{zn|y, λ} =
+∞
0
(zn − E{zn|y, λ})2
p(zn|y, λ)dzn.
4 Robert, C. (2007) The Bayesian choice: from decision-theoretic foundations to
computational implementation, Springer Science
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 10 / 29
11. Inference
How do we compute these posterior statistics?
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 11 / 29
12. Markov Chain Monte Carlo sampling
MCMC samplers construct a Markov chain of NMC samples
{x(s)|s = 1, . . . , NMC} that provide full insight of p(z|y, λ).
We compute the numerical estimates as
E{zn|y, λ} ≈
1
NMC
NMC
s=1
z(s)
n (2)
and
var{zn|y, λ} ≈
1
NMC
NMC
s=1
z(s)
n
2
− ˆµ2
n. (3)
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 12 / 29
13. Markov chain Monte Carlo sampling
We cannot obtain independent samples, as we are not using a
conjugate prior.
Many samplers are available in the statistics literature to overcome
this problem, but their behaviour is not well studied for
photon-limited imaging problems:
The dimension is very high
The number of data samples K is generally lower than the parameter
dimension N.
The likelihood term has longer tails as the photon-count reduces
(higher uncertainty).
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 13 / 29
14. Markov chain Monte Carlo sampling
In this work, we compare 6 state-of-the-art Monte Carlo samplers
Random walk Metropolis (RWM) [5]
Unadjusted Langevin algorithm (ULA) [5]
Metropolis adjusted Langevin algorithm (MALA) [5]
Hamiltonian Monte Carlo (HMC) [5]
No U-turn HMC (NUTS) [6]
Bouncy particle sampler (BPS) [7]
5 Brooks et al. (2011) Handbook of Markov chain Monte Carlo, CRC press.
6 Hoffman and Gelman (2014) The No-U-turn sampler: adaptively setting path lengths in
Hamiltonian Monte Carlo, Journal of Machine Learning Research, 15(1), 1593-1623.
7 Bouchard-Cˆot´e et al. (2018) The bouncy particle sampler: A nonreversible rejection-free
Markov chain Monte Carlo method, Journal of the American Statistical Association, 1-13.
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 14 / 29
15. Experiments
We evaluate the stochastic estimators as the dimension N of the
problem increases and the number of photons ( K
k=1 yk)
decreases
We assessed the bias and variance of the estimators as a function
of computing time.
Each algorithm was ran 10 times using different random seeds.
The reference estimate was computed using a long chain.
The parameters of all samplers were adjusted adaptively to obtain the
best performance.
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 15 / 29
16. Posterior mean estimates vs image size
10 60 110 160 210 260
3
4
5
64x64
10 60 110 160 210 260
Computing time [sec]
0
0.05
0.1
0.15
0.2
256x256
RWM ULA MALA HMC NUTS BPS
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 16 / 29
18. Marginal posterior variance estimates vs total photons
10 70 130 190 250
0
10
20
30
40
100k
photons
10 70 130 190 250
Computing time [sec]
0
0.01
0.02
1k
photons
RWM ULA MALA HMC NUTS BPS
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 18 / 29
19. Conclusions
HMC obtains the best scaling properties.
NUTS fails to provide good estimates in very high dimensions.
BPS achieves good mean estimates but shows slow convergence in
variance estimates
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 19 / 29
20. Conclusions
These results are in agreement with the theoretical scaling of the
effective sample size (ESS):
Method
O(N) per independent sample
MC iterations Total computations1
RWM N1 N2 log N
ULA N0.33 N1.33 log N
MALA N0.33 N1.33 log N
HMC N0.25 N1.25 log N
NUTS N0.25 N1.25 log N
BPS N0.47 N1.47 log N
The scaling of the MCMC samplers is still worse than an optimization
alternative, which would scale as N log N.
1
Obtaining one sample involves order N log N computations in all methods
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 20 / 29
21. Thanks for your attention!
Contact: jat3@hw.ac.uk
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 21 / 29
23. Random Walk Metropolis (RWM)
A proposal is generated by
x∗
∼ N(x(s−1)
, δI)
x∗ is accepted with probability
ρ = min{1, exp Uy (x(s−1)
) − Uy (x∗
) }
The value of δ is adapted to achieve the optimal acceptance ratio of
23%.
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 23 / 29
24. Unadjusted Langevin algorithm (ULA)
A new sample is obtained using gradient information
x(s)
∼ N(x(s−1)
−
δ
2
Uy (x(s−1)
), δI)
As no acceptance rule is used, the samples converge to a distribution
that is close to p(x|y, λ)
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 24 / 29
25. Metropolis adjusted Langevin algorithm (MALA)
Similarly to ULA, a proposal is generated by
x∗
∼ N x(s−1)
−
δ
2
Uy (x(s−1)
), δI
To converge to the exact posterior, x∗ is accepted with probability
ρ = min{1, exp(α)}
α = Uy (x(s−1)
) − Uy (x∗
) +
1
2δ
||x∗
− x(s−1)
+
δ
2
Uy (x(s−1)
)||2
2
−
1
2δ
||x(s−1)
− x∗
+
δ
2
Uy (x∗
)||2
2
The value of δ is adapted to achieve the optimal acceptance ratio of
57%.
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 25 / 29
26. Hamiltonian Monte Carlo (HMC)
We augment the target distribution
log p(x, p|y, λ) ∝ −Uy (x) −
1
2
||p||2
2
Sampling (x∗, p∗) ∼ p(x, p|y, λ) is done by simulating Hamiltonian
dynamics starting from (x(s−1), p0) and using a leap-frog integrator
of L steps of size .
As the leap-frog integrator is an inexact approximation, we accept x
with probability
ρ = min{1, exp Uy (x(s−1)
) − Uy (x∗
) −
1
2
||p∗
||2
2 +
1
2
||p0
||2
2}
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 26 / 29
27. No U-turn HMC (NUTS)
Adjusting L steps of size in HMC is generally cumbersome and
requires many trial runs.
The no U-turn sampler [6] automatically adapts the values of L and ,
such that the optimal acceptance rate of ≈ 65% is attained.
6 Hoffman and Gelman (2014) The No-U-turn sampler: adaptively setting path lengths in
Hamiltonian Monte Carlo, Journal of Machine Learning Research, 15(1), 1593-1623.
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 27 / 29
28. Bouncy particle sampler (BPS)
Recently, a ’rejection-free’ MCMC algorithm was introduced in [7].
The algorithm simulates a particle x that bounces around the
posterior with constant velocity v.
Each bouncing time τ(s) is obtained by sampling V ∼ U(0, 1) and
solving the equations:
Uy (x(s)
+ v(s)
τ(s)
) − Uy (x(s)
+ v(s)
τ∗
) = − log V
τ∗
= arg min
t≥0
Uy (x(s)
+ v(s)
t)
that are 1D and can be efficiently solved using Newton’s method!
7 Bouchard-Cˆot´e et al. (2018) The bouncy particle sampler: A nonreversible
rejection-free Markov chain Monte Carlo method, Journal of the American
Statistical Association, 1-13.
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 28 / 29
29. Bouncy particle sampler (BPS)
As the particle trajectory is ergodic, we compute estimators as
time-averages:
ˆµn =
NB
s=1
1
v
(s)
n
exp x(s)
n + v(s)
n τ(s)
ˆσ2
n =
NB
s=1
1
2v
(s)
n
exp 2x(s)
n + 2v(s)
n τ(s)
− ˆµ2
n
Juli´an Tachella (HWU-T´eSA) Restoration of photon-starved images September 4, 2018 29 / 29