Toward Node Liability in Federated Learning: Computational Cost and Network Overhead

We present a high-quality sky segmentation model for depth refinement and investigate residual architecture performance to inform optimally shrinking the network. We describe a model that runs in near real-time on mobile device, present a new, highquality dataset, and detail a unique weighing to trade off false positives and false negatives in binary classifiers. We show how the optimizations improve bokeh rendering by correcting stereo depth misprediction in sky regions. We detail techniques used to preserve edges, reject false positives, and ensure generalization to the diversity of sky scenes. Finally, we present a compact model and compare performance of four popular residual architectures (ShuffleNet, MobileNetV2, Resnet-101, and Resnet-34-like) at constant computational cost.

Download Full-text

An Alternate Algorithm for (3x3) Median Filtering of Digital Images

INTERNATIONAL JOURNAL OF COMPUTERS & TECHNOLOGY ◽

10.24297/ijct.v1i1.6732 ◽

2012 ◽

Vol 2 (1) ◽

pp. 7-9 ◽

Cited By ~ 2

Author(s):

Satinderjit Singh

Keyword(s):

Median Filter ◽

Computational Cost ◽

Spatial Coherence ◽

General Purpose ◽

Median Filtering ◽

Basic Algorithm ◽

Temporal Complexity ◽

Filter Kernel ◽

One Step ◽

High Computational Cost

Median filtering is a commonly used technique in image processing. The main problem of the median filter is its high computational cost (for sorting N pixels, the temporal complexity is O(NÂ·log N), even with the most efficient sorting algorithms). When the median filter must be carried out in real time, the software implementation in general-purpose processorsdoes not usually give good results. This Paper presents an efficient algorithm for median filtering with a 3x3 filter kernel with only about 9 comparisons per pixel using spatial coherence between neighboring filter computations. The basic algorithm calculates two medians in one step and reuses sorted slices of three vertical neighboring pixels. An extension of this algorithm for 2D spatial coherence is also examined, which calculates four medians per step.

Download Full-text

Parameterization of Divalent Cations for Coarse-Grained Simulations

10.26434/chemrxiv.11881716 ◽

2020 ◽

Author(s):

Florencia Klein ◽

Daniela Cáceres-Rojas ◽

Monica Carrasco ◽

Juan Carlos Tapia ◽

Julio Caballero ◽

...

Keyword(s):

Molecular Dynamics ◽

Metal Ions ◽

Molecular Dynamics Simulations ◽

Divalent Cations ◽

Computational Cost ◽

Data Bank ◽

Coarse Grained ◽

Interaction Parameters ◽

Dynamics Simulations ◽

Dynamical Description

<p>Although molecular dynamics simulations allow for the study of interactions among virtually all biomolecular entities, metal ions still pose significant challenges to achieve an accurate structural and dynamical description of many biological assemblies. This is particularly the case for coarse-grained (CG) models. Although the reduced computational cost of CG methods often makes them the technique of choice for the study of large biomolecular systems, the parameterization of metal ions is still very crude or simply not available for the vast majority of CG- force fields. Here, we show that incorporating statistical data retrieved from the Protein Data Bank (PDB) to set specific Lennard-Jones interactions can produce structurally accurate CG molecular dynamics simulations. Using this simple approach, we provide a set of interaction parameters for Calcium, Magnesium, and Zinc ions, which cover more than 80% of the metal-bound structures reported on the PDB. Simulations performed using the SIRAH force field on several proteins and DNA systems show that using the present approach it is possible to obtain non-bonded interaction parameters that obviate the use of topological constraints. </p>

Download Full-text

Active Learning Accelerates Ab Initio Molecular Dynamics on Pericyclic Reactive Energy Surfaces

10.26434/chemrxiv.11910948 ◽

2020 ◽

Author(s):

Shi Jun Ang ◽

Wujie Wang ◽

Daniel Schwalbe-Koda ◽

Simon Axelrod ◽

Rafael Gomez-Bombarelli

Keyword(s):

Molecular Dynamics ◽

Potential Energy ◽

Transition State ◽

Potential Energy Surface ◽

Ab Initio ◽

Energy Surface ◽

Computational Cost ◽

Reactive Trajectories ◽

Dynamics Simulations ◽

Energy Surfaces

<div>Modeling dynamical effects in chemical reactions, such as post-transition state bifurcation, requires <i>ab initio</i> molecular dynamics simulations due to the breakdown of simpler static models like transition state theory. However, these simulations tend to be restricted to lower-accuracy electronic structure methods and scarce sampling because of their high computational cost. Here, we report the use of statistical learning to accelerate reactive molecular dynamics simulations by combining high-throughput ab initio calculations, graph-convolution interatomic potentials and active learning. This pipeline was demonstrated on an ambimodal trispericyclic reaction involving 8,8-dicyanoheptafulvene and 6,6-dimethylfulvene. With a dataset size of approximately</div><div>31,000 M062X/def2-SVP quantum mechanical calculations, the computational cost of exploring the reactive potential energy surface was reduced by an order of magnitude. Thousands of virtually costless picosecond-long reactive trajectories suggest that post-transition state bifurcation plays a minor role for the reaction in vacuum. Furthermore, a transfer-learning strategy effectively upgraded the potential energy surface to higher</div><div>levels of theory ((SMD-)M06-2X/def2-TZVPD in vacuum and three other solvents, as well as the more accurate DLPNO-DSD-PBEP86 D3BJ/def2-TZVPD) using about 10% additional calculations for each surface. Since the larger basis set and the dynamic correlation capture intramolecular non-covalent interactions more accurately, they uncover longer lifetimes for the charge-separated intermediate on the more accurate potential energy surfaces. The character of the intermediate switches from entropic to thermodynamic upon including implicit solvation effects, with lifetimes increasing with solvent polarity. Analysis of 2,000 reactive trajectories on the chloroform PES shows a qualitative agreement with the experimentally-reported periselectivity for this reaction. This overall approach is broadly applicable and opens a door to the study of dynamical effects in larger, previously-intractable reactive systems.</div>

Download Full-text

Driving Torsion Scans with Wavefront Propagation

10.26434/chemrxiv.12044673.v1 ◽

2020 ◽

Cited By ~ 1

Author(s):

Yudong Qiu ◽

Daniel Smith ◽

Chaya Stern ◽

mudong feng ◽

Lee-Ping Wang

Keyword(s):

Potential Energy ◽

Force Field ◽

Degrees Of Freedom ◽

Computational Cost ◽

Initial Guess ◽

Field Development ◽

Force Field Development ◽

Wavefront Propagation ◽

Open Source Software Package

<div>The parameterization of torsional / dihedral angle potential energy terms is a crucial part of developing molecular mechanics force fields.</div><div>Quantum mechanical (QM) methods are often used to provide samples of the potential energy surface (PES) for fitting the empirical parameters in these force field terms.</div><div>To ensure that the sampled molecular configurations are thermodynamically feasible, constrained QM geometry optimizations are typically carried out, which relax the orthogonal degrees of freedom while fixing the target torsion angle(s) on a grid of values.</div><div>However, the quality of results and computational cost are affected by various factors on a non-trivial PES, such as dependence on the chosen scan direction and the lack of efficient approaches to integrate results started from multiple initial guesses.</div><div>In this paper we propose a systematic and versatile workflow called \textit{TorsionDrive} to generate energy-minimized structures on a grid of torsion constraints by means of a recursive wavefront propagation algorithm, which resolves the deficiencies of conventional scanning approaches and generates higher quality QM data for force field development.</div><div>The capabilities of our method are presented for multi-dimensional scans and multiple initial guess structures, and an integration with the MolSSI QCArchive distributed computing ecosystem is described.</div><div>The method is implemented in an open-source software package that is compatible with many QM software packages and energy minimization codes.</div>

Download Full-text

Message Passing Neural Networks for Partial Charge Assignment to Metal-Organic Frameworks

10.26434/chemrxiv.12298487 ◽

2020 ◽

Author(s):

Ali Raza ◽

Arni Sturluson ◽

Cory Simon ◽

Xiaoli Fern

Keyword(s):

Electronic Structure ◽

Message Passing ◽

Gas Adsorption ◽

Metal Organic Frameworks ◽

Computational Cost ◽

Electronic Structure Calculations ◽

High Fidelity ◽

Partial Charges ◽

Metal Organic ◽

Structure Calculations

Virtual screenings can accelerate and reduce the cost of discovering metal-organic frameworks (MOFs) for their applications in gas storage, separation, and sensing. In molecular simulations of gas adsorption/diffusion in MOFs, the adsorbate-MOF electrostatic interaction is typically modeled by placing partial point charges on the atoms of the MOF. For the virtual screening of large libraries of MOFs, it is critical to develop computationally inexpensive methods to assign atomic partial charges to MOFs that accurately reproduce the electrostatic potential in their pores. Herein, we design and train a message passing neural network (MPNN) to predict the atomic partial charges on MOFs under a charge neutral constraint. A set of ca. 2,250 MOFs labeled with high-fidelity partial charges, derived from periodic electronic structure calculations, serves as training examples. In an end-to-end manner, from charge-labeled crystal graphs representing MOFs, our MPNN machine-learns features of the local bonding environments of the atoms and learns to predict partial atomic charges from these features. Our trained MPNN assigns high-fidelity partial point charges to MOFs with orders of magnitude lower computational cost than electronic structure calculations. To enhance the accuracy of virtual screenings of large libraries of MOFs for their adsorption-based applications, we make our trained MPNN model and MPNN-charge-assigned computation-ready, experimental MOF structures publicly available.<br>

Download Full-text

Nanosecond Photodynamics Simulations of a Cis-Trans Isomerization Are Enabled by Machine Learning

10.26434/chemrxiv.13047863 ◽

2020 ◽

Author(s):

Jingbai Li ◽

Patrick Reiser ◽

André Eberhard ◽

Pascal Friederich ◽

Steven Lopez

Keyword(s):

Machine Learning ◽

Neural Networks ◽

Excited State ◽

Adaptive Sampling ◽

Computational Cost ◽

Ground Truth ◽

Absolute Error ◽

Photochemical Reactions ◽

Computational Techniques ◽

Full Potential

<p>Photochemical reactions are being increasingly used to construct complex molecular architectures with mild and straightforward reaction conditions. Computational techniques are increasingly important to understand the reactivities and chemoselectivities of photochemical isomerization reactions because they offer molecular bonding information along the excited-state(s) of photodynamics. These photodynamics simulations are resource-intensive and are typically limited to 1–10 picoseconds and 1,000 trajectories due to high computational cost. Most organic photochemical reactions have excited-state lifetimes exceeding 1 picosecond, which places them outside possible computational studies. Westermeyr <i>et al.</i> demonstrated that a machine learning approach could significantly lengthen photodynamics simulation times for a model system, methylenimmonium cation (CH<sub>2</sub>NH<sub>2</sub><sup>+</sup>).</p><p>We have developed a Python-based code, Python Rapid Artificial Intelligence <i>Ab Initio</i> Molecular Dynamics (PyRAI<sup>2</sup>MD), to accomplish the unprecedented 10 ns <i>cis-trans</i> photodynamics of <i>trans</i>-hexafluoro-2-butene (CF<sub>3</sub>–CH=CH–CF<sub>3</sub>) in 3.5 days. The same simulation would take approximately 58 years with ground-truth multiconfigurational dynamics. We proposed an innovative scheme combining Wigner sampling, geometrical interpolations, and short-time quantum chemical trajectories to effectively sample the initial data, facilitating the adaptive sampling to generate an informative and data-efficient training set with 6,232 data points. Our neural networks achieved chemical accuracy (mean absolute error of 0.032 eV). Our 4,814 trajectories reproduced the S<sub>1</sub> half-life (60.5 fs), the photochemical product ratio (<i>trans</i>: <i>cis</i> = 2.3: 1), and autonomously discovered a pathway towards a carbene. The neural networks have also shown the capability of generalizing the full potential energy surface with chemically incomplete data (<i>trans</i> → <i>cis</i> but not <i>cis</i> → <i>trans</i> pathways) that may offer future automated photochemical reaction discoveries.</p>

Download Full-text

Automated, Accurate, and Scalable Relative Protein-Ligand Binding Free Energy Calculations using Lambda Dynamics

10.26434/chemrxiv.12781310.v1 ◽

2020 ◽

Author(s):

E. Prabhu Raman ◽

Thomas J. Paul ◽

Ryan L. Hayes ◽

Charles L. Brooks III

Keyword(s):

Free Energy ◽

Ligand Binding ◽

Binding Affinity ◽

Binding Free Energy ◽

Computational Cost ◽

Combinatorial Libraries ◽

Free Energy Calculations ◽

Lead Optimization ◽

Efficient Estimation ◽

Lead Compound

<p>Accurate predictions of changes to protein-ligand binding affinity in response to chemical modifications are of utility in small molecule lead optimization. Relative free energy perturbation (FEP) approaches are one of the most widely utilized for this goal, but involve significant computational cost, thus limiting their application to small sets of compounds. Lambda dynamics, also rigorously based on the principles of statistical mechanics, provides a more efficient alternative. In this paper, we describe the development of a workflow to setup, execute, and analyze Multi-Site Lambda Dynamics (MSLD) calculations run on GPUs with CHARMm implemented in BIOVIA Discovery Studio and Pipeline Pilot. The workflow establishes a framework for setting up simulation systems for exploratory screening of modifications to a lead compound, enabling the calculation of relative binding affinities of combinatorial libraries. To validate the workflow, a diverse dataset of congeneric ligands for seven proteins with experimental binding affinity data is examined. A protocol to automatically tailor fit biasing potentials iteratively to flatten the free energy landscape of any MSLD system is developed that enhances sampling and allows for efficient estimation of free energy differences. The protocol is first validated on a large number of ligand subsets that model diverse substituents, which shows accurate and reliable performance. The scalability of the workflow is also tested to screen more than a hundred ligands modeled in a single system, which also resulted in accurate predictions. With a cumulative sampling time of 150ns or less, the method results in average unsigned errors of under 1 kcal/mol in most cases for both small and large combinatorial libraries. For the multi-site systems examined, the method is estimated to be more than an order of magnitude more efficient than contemporary FEP applications. The results thus demonstrate the utility of the presented MSLD workflow to efficiently screen combinatorial libraries and explore chemical space around a lead compound, and thus are of utility in lead optimization.</p>

Download Full-text

Quantitative Ranking of Ligand Binding Kinetics with a Multiscale Milestoning Simulation Approach

10.26434/chemrxiv.6726977.v1 ◽

2018 ◽

Author(s):

Benjamin R. Jagger ◽

Christoper T. Lee ◽

Rommie Amaro

Keyword(s):

Molecular Dynamics ◽

Drug Discovery ◽

Small Molecule ◽

Brownian Dynamics ◽

Model Simulation ◽

Computational Cost ◽

Lead Selection ◽

Delta G ◽

Dynamics Simulations

<p>The ranking of small molecule binders by their kinetic (kon and koff) and thermodynamic (delta G) properties can be a valuable metric for lead selection and optimization in a drug discovery campaign, as these quantities are often indicators of in vivo efficacy. Efficient and accurate predictions of these quantities can aid the in drug discovery effort, acting as a screening step. We have previously described a hybrid molecular dynamics, Brownian dynamics, and milestoning model, Simulation Enabled Estimation of Kinetic Rates (SEEKR), that can predict kon’s, koff’s, and G’s. Here we demonstrate the effectiveness of this approach for ranking a series of seven small molecule compounds for the model system, -cyclodextrin, based on predicted kon’s and koff’s. We compare our results using SEEKR to experimentally determined rates as well as rates calculated using long-timescale molecular dynamics simulations and show that SEEKR can effectively rank the compounds by koff and G with reduced computational cost. We also provide a discussion of convergence properties and sensitivities of calculations with SEEKR to establish “best practices” for its future use.</p>

Download Full-text