Scene flow tracks the three-dimensional (3D) motion of each point in adjacent point clouds. It provides fundamental 3D motion perception for autonomous driving and server robot. Although the Red Green Blue Depth (RGBD) camera or Light Detection and Ranging (LiDAR) capture discrete 3D points in space, the objects and motions usually are continuous in the macro world. That is, the objects keep themselves consistent as they flow from the current frame to the next frame. Based on this insight, the Generative Adversarial Networks (GAN) is utilized to self-learn 3D scene flow with no need for ground truth. The fake point cloud of the second frame is synthesized from the predicted scene flow and the point cloud of the first frame. The adversarial training of the generator and discriminator is realized through synthesizing indistinguishable fake point cloud and discriminating the real point cloud and the synthesized fake point cloud. The experiments on Karlsruhe Institute of Technology and Toyota Technological Institute (KITTI) scene flow dataset show that our method realizes promising results without ground truth. Just as human, the proposed method can identify the similar local structures of two adjacent frames even without knowing the ground truth scene flow. Then, the local correspondence can be correctly estimated, and further the scene flow is correctly estimated. Corresponding author(s) Email: [email protected]
Quantum chemistry must evolve if it wants to fully leverage the benefits of the internet age, where the world wide web offers a vast tapestry of tools that enable users to communicate and interact with complex data at the speed and convenience of a button press. The Open Chemistry project has developed an open source framework that offers an end-to-end solution for producing, sharing, and visualizing quantum chemical data interactively on the web using an array of modern tools and approaches. These tools build on some of the best open source community projects such as Jupyter for interactive online notebooks, coupled with 3D accelerated visualization, state-of-the-art computational chemistry codes including NWChem and Psi4 and emerging machine learning and data mining tools such as ChemML and ANI. They offer flexible formats to import and export data, along with approaches to compare computational and experimental data.
Leukocyte differential test is a widely performed clinical procedure for screening infectious diseases. Existing hematology analyzers require labor-intensive work and a panel of expensive reagents. Here we report an artificial-intelligence enabled reagent-free imaging hematology analyzer (AIRFIHA) modality that can accurately classify subpopulations of leukocytes with minimal sample preparation. AIRFIHA is realized through training a two-step residual neural network using label-free images of isolated leukocytes acquired from a custom-built quantitative phase microscope. By leveraging the rich information contained in quantitative phase images, we not only achieved high accuracy in differentiating B and T lymphocytes, but also classified CD4 and CD8 cells, therefore outperforming the classification accuracy of most current hematology analyzers. We validated the performance of AIRFIHA in a randomly selected test set and cross-validated it across all blood donors. Owing to its easy operation, low cost, and accurate discerning capability of complex leukocyte subpopulations, we envision AIRFIHA is clinically translatable and can also be deployed in resource-limited settings, e.g., during pandemic situations for the rapid screening of infectious diseases. Corresponding author(s) Email: [email protected], [email protected]
We have performed a large-scale evaluation of current computational methods, including conventional small-molecule force fields, semiempirical, density functional, ab initio electronic structure methods, and current machine learning (ML) techniques to evaluate relative single-point energies. Using up to 10 local minima geometries across ~700 molecules, each optimized by B3LYP-D3BJ with single-point DLPNO-CCSD(T) triple-zeta energies, we consider over 6,500 single points to compare the correlation between different methods for both relative energies and ordered rankings of minima. We find promise from current ML methods and recommend methods at each tier of the accuracy-time tradeoff, particularly the recent GFN2 semiempirical method, the B97-3c density functional approximation, and RI-MP2 for accurate conformer energies. The ANI family of ML methods shows promise, particularly the ANI-1ccx variant trained in part on coupled-cluster energies. Multiple methods suggest continued improvements should be expected in both performance and accuracy.
Counting parameters has become customary in the density functional theory community as a way to infer the transferability of popular approximations to the exchange–correlation functionals. Recent work in data science, however, has demonstrated that the number of parameters of a fitted model is not related to the complexity of the model itself, nor to its eventual overfitting. Using similar arguments, we show here that it is possible to represent every modern exchange–correlation functional approximation using just one single parameter. This procedure proves the futility of the number of parameters as a measure of transferability. To counteract this shortcoming, we introduce and analyze the performance of three statistical criteria for the evaluation of the transferability of exchange–correlation functionals. The three criteria are called Akaike information criterion (AIC), Vapnik–Chervonenkis criterion (VCC), and cross-validation criterion (CVC) and are used in a preliminary assessment to rank 60 exchange–correlation functional approximations using the ASCDB database of chemical data.
The expense of quantum chemistry calculations significantly hinders the search for novel catalysts. Here, we provide a tutorial for using an easy and highly cost-efficient calculation scheme called alchemical perturbation density functional theory (APDFT) for rapid predictions of binding energies of reaction intermediates and reaction barrier heights based on Kohn-Sham density functional theory reference data. We outline standard procedures used in computational catalysis applications, explain how computational alchemy calculations can be carried out for those applications, and then present bench marking studies of binding energy and barrier height predictions. Using a single OH binding energy on the Pt(111) surface as a reference case, we use computational alchemy to predict binding energies of 32 variations of this system with a mean unsigned error of less than 0.05 eV relative to single-point DFT calculations. Using a single nudged elastic band calculation for CH4 dehydrogenation on Pt(111) as a reference case, we generate 32 new pathways with barrier heights having mean unsigned errors of less than 0.3 eV relative to single-point DFT calculations. Notably, this easy APDFT scheme brings no appreciable computational cost once reference calculations are done, and this shows that simple applications of computational alchemy can significantly impact DFT-driven explorations for catalysts. To accelerate computational catalysis discovery and ensure computational reproducibility, we also include Python modules that allow users to perform their own computational alchemy calculations.Keywords --- Computational catalysis, density functional theory (DFT), adsorption energies, nudged elastic band calculations, binding energies, barrier heights
The growing generation of data and their wide availability has led to the development of tools to produce, analyze and store this information. Computational chemistry studies and especially catalytic applications often yield a vast amount of chemical information that can be analyzed and stored using these tools. In this manuscript we present a framework that automatically performs a full automated procedure consisting in the transfer of an adsorbate from a known metal slab to a new metal slab with similar packing. Our method generates the new geometry and also performs the required calculations and analysis to finally upload the processed data to an online database (ioChem-BD). Two different implementations have been built, one to relocate minimum energy point structures and the second to transfer transition states. Our framework shows good performance for the minimum point location and a decent performance for the transition state identification. Most of the failures occurred during the transition state searches needed additional steps to fully complete the process. Further improvements of our framework are required to increase the performance of both implementations. These results point to the _avoidhuman_ path as a feasible solution for studies on very large systems that require a significant amount of human resources and in consequence are prone to human errors.
How to prepare and submit your paper1) AISY Interactive Papers are prepared and submitted via Authorea. You can get started by using the submission template. (Click Use Template to get started). If you prefer, you can also start a document from scratch or upload an existing manuscript. If you want to do so, from your Authorea account, in the top right click the + button to see available options.2) We recommend to organize your article by including dedicated (sub)sections; please refer to the more general Advanced Intelligent Systems guideline for authors for details on document structure. The Submission Template that we provide has the required document structure.3) Once you have prepared your document and are ready to submit it for consideration, submit it to the AISY Interactive Papers Authorea Collection. (Click Submit and select your manuscript). You can also submit directly from the document by selecting "Submit for review" and finding the "AISY Interactive Papers" portal from the list.4) Make your document public, by publishing it as a preprint. Click Publish and follow the prompts. Choose a few suitable keywords; also select CC 4.0 is the copyright license.5) Submit a static version of your document through AISY's journal's editorial platform. We recommend that you download/export the interactive article in Word, PDF, or LaTeX format and submit your manuscript files.
Engineering microscopic collectives of cells or microrobots is challenging due to the often-limited capabilities of the individual agents, our inability to reliably program their motion and local interactions, and difficulties visualising their behaviours. Here, we present a low-cost, modular and open-source Dynamic Optical MicroEnvironment (DOME) and demonstrate its ability to augment microagent capabilities and control collective behaviours using light. The DOME offers an accessible means to study complex multicellular phenomena and implement de-novo microswarms with desired functionalities. Corresponding author(s) Email: [email protected] [email protected]
Soft fluidic actuators produce continuous and life-like motions that are intrinsically safe, but current designs are not yet mature enough to enable large deployment with high force and low-cost fabrication methods. Here, soft fluidic actuators with two superimposed origami architectures are reported. Driven by a fluid input, the presented dual-origami soft actuators produce quasi-sequential deployment and bending motion that is guided by unsymmetric unfolding of low-stretchable origami components. The dominance between the deployment and bending can be shifted by varying the unfolding behavior, enabling pre-programming of the motion. The proposed origami-inspired soft actuators are directly fabricated by low-cost fused deposition modeling 3D-printing, and subjected to a heat treatment post-processing to enhance the fluid sealing performance. Finally, soft gripper applications are presented and they successfully demonstrate gripping tasks that each requires strength, delicacy, precision and dexterity. The dual-origami approach offers a design guidance for soft robots to embody grow-and-retract motion with a small initial form factor, promising for applications in next-generation soft robotic systems.
Identifying Clostridioides difficile infection (CDI) patients at risk of mortality or recurrence will facilitate prevention, timely treatment and improve clinical outcomes. We aim to establish an open-access web-based prediction system, which estimates CDI patients’ mortality and recurrence outcomes, and explains the machine learning prediction with patients’ characteristics. Prognostic models were developed using four various types of machine learning algorithms and statistical logistics regression model utilizing over 15,000 CDI patients from 41 hospitals in Hong Kong. The boosting-based machine learning algorithm Gradient Boosting Machine (Mortality AUC: 0.7878; Recurrence AUC: 0.7076) outperformed statistical models (Mortality AUC: 0.7573; Recurrence AUC: 0.6927) and other machine learning algorithms. The open-access prediction system for clinicians to assess and interpret the risk factors of CDI patients is now available at https://www.cdiml.care/. In this article, we explain the development of machine learning models and illustrate how to apply hyperparameter tuning with cross-validation to optimize the model accuracy.
Bacteria-mediated drug delivery systems comprising nanotherapeutics conjugated onto bacteria synergistically augment the efficacy of both therapeutic modalities in cancer therapy. Nanocarriers preserve therapeutics' bioavailability and reduce systemic toxicity, while bacteria selectively colonize the cancerous tissue, impart intrinsic and immune-mediated antitumor effects, and propel nanotherapeutics interstitially. The optimal bacteria-nanoparticle (NP) conjugates would carry the maximal NP load with minimal motility speed hindrance for effective interstitial distribution. Furthermore, a well-defined and repeatable NP attachment density distribution is crucial to determining these biohybrid systems' efficacious dosage and robust performance. Herein, we utilized our Nanoscale Bacteria-Enabled Autonomous Delivery System (NanoBEADS) platform to investigate the effects of assembly process parameters of mixing method, volume, and duration on NP attachment density and repeatability. We also evaluated the effect of linkage chemistry and NP size on NP attachment density, viability, growth rate, and motility of NanoBEADS. We show that the linkage chemistry impacts NP attachment density while the self-assembly process parameters affect the repeatability and, to a lesser extent, attachment density. Lastly, the attachment density affects NanoBEADS' growth rate and motility in an NP size-dependent manner. These findings will contribute to the development of scalable and repeatable bacteria-nanoparticle biohybrids for applications in drug delivery and beyond. Corresponding author(s) Email: [email protected]