MBMaxime Bassenne

Maxime Bassenne

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Maxime Bassenne has a total of 36 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Human-level comparable control volume mapping with a deep unsupervised-learning model for image-guided radiation therapy

    by Xiaokun Liang, Maxime Bassenne, Dimitre H. Hristov, Tauhidul Islam, Wei Zhao, Mengyu Jia, Zhicheng Zhang, Michael Gensheimer, Beth Beadle, Quynh Le, et al.


  • Open Access

    Deep Learning Prediction of Cancer Prevalence from Satellite Imagery

    by Jean-Emmanuel Bibault, Maxime Bassenne, Hongyi Ren, Lei Xing

    The worldwide growth of cancer incidence can be explained in part by changes in the prevalence and distribution of risk factors. There are geographical gaps in the estimates of cancer prevalence, which could be filled with innovative methods. We used deep learning (DL) features extracted from satellite images to predict cancer prevalence at the census tract level in seven cities in the United States. We trained the model using detailed cancer prevalence estimates from 2018 available in the CDC (Center for Disease Control) 500 Cities project. Data from 3500 census tracts covering 14,483,366 inhabitants were included. Features were extracted from 170,210 satellite images with deep learning. This method explained up to 64.37% (median = 43.53%) of the variation of cancer prevalence. Satellite features are highly correlated with individual socioeconomic and health measures that are linked to cancer prevalence (age, smoking and drinking status, and obesity). A higher similarity between two environments is associated with better generalization of the model (p = 1.10–6). This method can be used to accurately estimate cancer prevalence at a high spatial resolution without using surveys at a fraction of the cost.


  • Open Access

    Time-Accurate and highly-Stable Explicit operators for stiff differential equations

    by Maxime Bassenne, Lin Fu, Ali Mani

    Unconditionally stable implicit time-marching methods are powerful in solving stiff differential equations efficiently. In this work, a novel framework to handle stiff physical terms implicitly is proposed. Both physical and numerical stiffness originating from convection, diffusion and source terms (typically related to reaction) can be handled by a set of predefined Time-Accurate and highly-Stable Explicit (TASE) operators in a unified framework. The proposed TASE operators act as preconditioners on the stiff terms and can be deployed to any existing explicit time-marching methods straightforwardly. The resulting time integration methods remain the original explicit time-marching schemes, yet with nearly unconditional stability. The TASE operators can be designed to be arbitrarily high-order accurate with Richardson extrapolation such that the accuracy order of original explicit time-marching method is preserved. Theoretical analyses and stability diagrams show that the $s$-stages $s$th-order explicit Runge-Kutta (RK) methods are unconditionally stable when preconditioned by the TASE operators with order $p \leq s$ and $p \leq 2$. On the other hand, the $s$th-order RK methods preconditioned by the TASE operators with order of $p \leq s$ and $p > 2$ are nearly unconditionally stable. The only free parameter in TASE operators can be determined a priori based on stability arguments. Unlike classical implicit methods, the TASE methodology allows for solving non-linear problems with arbitrary order without requiring solving a nonlinear system of equations. A set of benchmark problems with strong stiffness is simulated to assess the performance of the TASE method. Numerical results suggest that the proposed framework preserves the high-order accuracy of the explicit time-marching methods with very-large time steps for all the considered cases.

  • Open Access

    Computational model discovery with reinforcement learning

    by Maxime Bassenne, Adrián Lozano-Durán

    The motivation of this study is to leverage recent breakthroughs in artificial intelligence research to unlock novel solutions to important scientific problems encountered in computational science. To address the human intelligence limitations in discovering reduced-order models, we propose to supplement human thinking with artificial intelligence. Our three-pronged strategy consists of learning (i) models expressed in analytical form, (ii) which are evaluated a posteriori, and iii) using exclusively integral quantities from the reference solution as prior knowledge. In point (i), we pursue interpretable models expressed symbolically as opposed to black-box neural networks, the latter only being used during learning to efficiently parameterize the large search space of possible models. In point (ii), learned models are dynamically evaluated a posteriori in the computational solver instead of based on a priori information from preprocessed high-fidelity data, thereby accounting for the specificity of the solver at hand such as its numerics. Finally in point (iii), the exploration of new models is solely guided by predefined integral quantities, e.g., averaged quantities of engineering interest in Reynolds-averaged or large-eddy simulations (LES). We use a coupled deep reinforcement learning framework and computational solver to concurrently achieve these objectives. The combination of reinforcement learning with objectives (i), (ii) and (iii) differentiate our work from previous modeling attempts based on machine learning. In this report, we provide a high-level description of the model discovery framework with reinforcement learning. The method is detailed for the application of discovering missing terms in differential equations. An elementary instantiation of the method is described that discovers missing terms in the Burgers' equation.

  • Mitigation of turbophoresis in particle-laden turbulent channel flows by using incident electric fields

    by M. Di Renzo, P. L. Johnson, M. Bassenne, L. Villafañe, J. Urzay

    It is known that the dynamics of particles dispersed in turbulent flows can be significantly altered by electric charges and external electric fields. The next step therefore involves the investigation of whether these electric interactions can be exploited in engineering applications to control processes involving the transport of particles in turbulence. In this study, direct numerical simulations of incompressible turbulent channel flows laden with a positively isocharged suspension of monodisperse small inertial particles are employed to investigate the effect of electric charges carried by the particles on their dispersion, and to illustrate the intentional abatement of turbophoretic effects using incident electric fields. An Eulerian-Lagrangian formulation is employed along with a fast multipole method for the electric potential conveniently corrected with wall boundary conditions. Operating conditions are identified in terms of characteristic dimensionless parameters where an AC electric field applied across the channel walls decreases the time-averaged concentration of particles near the walls by up to two orders of magnitude.


  • Turbophoresis of small inertial particles: theoretical considerations and application to wall-modelled large-eddy simulations

    by Perry L. Johnson, Maxime Bassenne, Parviz Moin

    In wall-bounded turbulent flows, the wall-normal gradient in turbulence intensity causes inertial particles to move preferentially toward the wall, leading to elevated concentration levels in the viscous sublayer. At first glance, wall-modelled large-eddy simulations may seem ill suited for accurately simulating this behaviour, given that the sharp gradients and coherent structures in the viscous sublayer and buffer region are unresolved in this approach. In this paper, a detailed inspection of conservation equations describing the influence of turbophoresis and near-wall structures on particle concentration profiles reveals a more nuanced view depending on the friction Stokes number. The dynamics of low and moderate Stokes number particles indeed depends strongly on the complex spatio-temporal details of streaks, ejections, and sweeps in the near-wall region. This significantly impacts the near-wall particle concentration through a biased sampling effect which provides a net force away from the wall on the particle ensemble caused by the tendency of inertial particles to accumulate in low-speed ejection regions. At higher Stokes numbers, however, this biased sampling is of minimal importance, and the particle concentration becomes inversely proportional to the wall-normal particle velocity variance at a given distance from the wall. As a result, wall-modelled large-eddy simulations can predict concentration profiles with more accuracy in the high Stokes number regime than low Stokes numbers simply by modifying the interpolation scheme for particles between the first grid point and the boundary. However, accurate representation of low and moderate Stokes number particles depends critically on information not present in standard wall-modelled large-eddy simulations.


  • Open Access

    Painting fluid motion using convolutional neural networks: An album of fluid motion 2.0

    by Maxime Bassenne, Andrew Banko, Sadaf Sobhani

    This paper is associated with a poster winner of a 2018 APS/DFD Gallery of Fluid Motion Award for work presented at the DFD Gallery of Fluid Motion. The original poster is available online at the Gallery of Fluid Motion, https://doi.org/10.1103/APS.DFD.2018.GFM.P0004

  • Open Access

    Mandala-inspired representation of the turbulent energy cascade

    by Maxime Bassenne, Hyunji Jane Bae, Adrián Lozano-Durán

    This paper is associated with a poster winner of a 2017 APS/DFD Milton van Dyke Award for work presented at the DFD Gallery of Fluid Motion. The original poster is available from the Gallery of Fluid Motion, https://doi.org/10.1103/APS.DFD.2017.GFM.P0026


  • Open Access

    Wavelet multiresolution analysis of particle-laden turbulence

    by Maxime Bassenne, Parviz Moin, Javier Urzay

    Direct numerical simulations of incompressible homogeneous-isotropic turbulence laden with a dilute suspension of inertial point particles are performed in conjunction with a wavelet multiresolution analysis of the results. The use of spatially localized wavelet basis functions enables the simultaneous consideration of physical and scale spaces in the spectral characterization of the flow field of the carrier phase and the concentration field of the dispersed phase. The multiresolution analysis of the dispersed phase provides statistical information about the spatial variabilities of a scale-dependent coarse-grained number density field and the local energy spectra of its fluctuations, characterizing the sensitivities of those quantities to variations in scale and Stokes number. In particular, the spatial variabilities of the wavelet energy spectrum of the particle concentration fluctuations are observed to be maximum in regimes where the particles preferentially concentrate. The results highlight the scale-dependent inhomogeneities of the structures in the concentration field generated by preferential concentration, and the existence of characteristic scales of interaction between the dispersed and carrier phases. Additionally, an interphase multiresolution analysis is performed that indicates the occurrence of a spatial anticorrelation between the enstrophy and kinetic-energy spectra of the carrier phase and the particle concentration at small scales in regimes where preferential concentration is important. This anticorrelation vanishes as the scale is increased and is largely suppressed when the preferential-concentration effect is negligible.