Instances Of Model Research Articles

Deepfake forensics: Cross-manipulation robustness of feedforward- and recurrent convolutional forgery detection methods

Recently, deep-learning based video-manipulation techniques, so-called deepfakes, have gained a lot of traction in popular media. Deepfakes are predominantly used to photo-realistically alter the identity of actors recorded on video, and misuse of this new technology has the potential to harm individuals, companies, or to inflict damage in a societal context. The accelerated pace, at which deepfake technology is improved, challenges contemporary research to provide the urgently required detection methods. A major obstacle to reliable deepfake detection in the field is the impaired generalizability across specific instances of deepfake models, i.e. detection performance of previously not encountered manipulation models. The present work aims to establish a better understanding of the critical factors that guide cross-manipulation detection performance. Recent studies have indicated that the properties of video-based modeling may be leveraged to enhance the detection of unknown deepfake manipulations. Therefore, we compare multiple image- and video-based detection models and evaluate their performance on instances both generated by known, and unknown manipulation models. Furthermore we attempt to replicate research that successfully improved cross-manipulation detection by using image-perturbation methods to degrade training data. Using multiple data-sources to emulate in- and out-sample detection performance, we demonstrate that none of the two model types show universally superior performance. Also, our results confirm that cross-manipulation evaluation results in problematic detection accuracy for all models, regardless of whether we apply image-perturbation techniques during training. Additionally, the detection performance on a set of manually selected, high-quality deepfake videos indicates that the current state-of-the-art detection models are not yet fully equipped for real-world applications.

Misinformation and Disinformation in Modern Warfare

Assessing Distorted Information in Modern Warfare Distorted information (misinformation and disinformation) has long been a part of warfare (see the writings of Sun-Tzu). However, the study of the ever-increasing use of distorted information in modern warfare has been rather limited. In “Misinformation and Disinformation in Modern Warfare,” Chang, Keblis, Li, Iakovou, and White model instances of today’s battlespace as a partially observable game with three agents, a leader and two followers, and examine the benefit to the leader (e.g., a military command) of modulating the communication of information between (i) followers who are adversaries and (ii) followers who are allies. Counter to intuition, the study shows that only under certain conditions is it optimal for the leader to degrade (enhance) the quality of the information communicated between adversarial (allied) followers. The developed methodology is applied to warfare instances encountered in the Battle of Mosul.

Numerical Solutions of the Hattendorff Differential Equation for Multi-State Markov Insurance Models

We use the representation of a continuous time Hattendorff differential equation and Matlab to compute 2σt(j), the solution of a backwards in time differential equation that describes the evolution of the variance of Lt(j), the loss at time t random variable for a multi-state Markovian process, given that the state at time t is j. We demonstrate this process by solving examples of several instances of a multi-state model which a practitioner can use as a guide to solve and analyze specific multi-state models. Numerical solutions to compute the variance 2σt(j) enable practitioners and academic researchers to test and simulate various state-space scenarios, with possible transitions to and from temporary disabilities, to permanent disabilities, to and from good health, and eventually to a deceased state. The solution method presented in this paper allows researchers and practitioners to easily compute the evolution of the variance of loss without having to resort to detailed programming.

A Solution Approach to Distributionally Robust Joint-Chance-Constrained Assignment Problems

We study the assignment problem with chance constraints (CAP) and its distributionally robust counterpart DR-CAP. We present a technique for estimating big-M in such a formulation that takes advantage of the ambiguity set. We consider a 0-1 bilinear knapsack set to develop valid inequalities for CAP and DR-CAP. This is generalized to the joint chance constraint problem. A probability cut framework is also developed to solve DR-CAP. A computational study on problem instances obtained from using real hospital surgery data shows that the developed techniques allow us to solve certain model instances and reduce the computational time for others. The use of Wasserstein ambiguity set in the DR-CAP model improves the out-of-sample performance of satisfying the chance constraints more significantly than the one possible by increasing the sample size in the sample average approximation technique. The solution time for DR-CAP model instances is of the same order as that for solving the CAP instances. This finding is important because chance constrained optimization models are very difficult to solve when the coefficients in the constraints are random.

Workflows for Construction of Spatio-Temporal Probabilistic Maps for Volcanic Hazard Assessment

Probabilistic hazard assessments for studying overland pyroclastic flows or atmospheric ash clouds under short timelines of an evolving crisis, require using the best science available unhampered by complicated and slow manual workflows. Although deterministic mathematical models are available, in most cases, parameters and initial conditions for the equations are usually only known within a prescribed range of uncertainty. For the construction of probabilistic hazard assessments, accurate outputs and propagation of the inherent input uncertainty to quantities of interest are needed to estimate necessary probabilities based on numerous runs of the underlying deterministic model. Characterizing the uncertainty in system states due to parametric and input uncertainty, simultaneously, requires using ensemble based methods to explore the full parameter and input spaces. Complex tasks, such as running thousands of instances of a deterministic model with parameter and input uncertainty require a High Performance Computing infrastructure and skilled personnel that may not be readily available to the policy makers responsible for making informed risk mitigation decisions. For efficiency, programming tasks required for executing ensemble simulations need to run in parallel, leading to twin computational challenges of managing large amounts of data and performing CPU intensive processing. The resulting flow of work requires complex sequences of tasks, interactions, and exchanges of data, hence the automatic management of these workflows are essential. Here we discuss a computer infrastructure, methodology and tools which enable scientists and other members of the volcanology research community to develop workflows for construction of probabilistic hazard maps using remotely accessed computing through a web portal.

AVQBits—Adaptive Video Quality Model Based on Bitstream Information for Various Video Applications

The paper presents <i>AV QBits</i>, a versatile, bitstream-based video quality model. It can be applied in several contexts such as video service monitoring, evaluation of video encoding quality, of gaming video QoE, and even of omnidirectional video quality. In the paper, it is shown that <i>AV QBits</i> predictions closely match video quality ratings obained in various subjective tests with human viewers, for videos up to 4K-UHD resolution (Ultra-High Definition, 3840 x 2180 pixels) and framerates up 120 fps. With the different variants of <i>AV QBits</i> presented in the paper, video quality can be monitored either at the client side, in the network or directly after encoding. The no-reference <i>AV QBits</i> model was developed for different video services and types of input data, reflecting the increasing popularity of Video-on-Demand services and widespread use of HTTP-based adaptive streaming. At its core, <i>AV QBits</i> encompasses the standardized ITU-T P.1204.3 model, with further model instances that can either have restricted or extended input information, depending on the application context. Four different instances of <i>AV QBits</i> are presented, that is, a Mode 3 model with full access to the bitstream, a Mode 0 variant using only metadata such as codec type, framerate, resoution and bitrate as input, a Mode 1 model using Mode 0 information and frame-type and -size information, and a Hybrid Mode 0 model that is based on Mode 0 metadata and the decoded video pixel information. The models are trained on the authors&#x2019; own AVT-PNATS-UHD-1 dataset described in the paper. All models show a highly competitive performance by using AVT-VQDB-UHD-1 as validation dataset, e.g., with the Mode 0 variant yielding a value of 0.890 Pearson Correlation, the Mode 1 model of 0.901, the hybrid no-reference mode 0 model of 0.928 and the model with full bitstream access of 0.942. In addition, all four <i>AV QBits</i> variants are evaluated when applying them out-of-the-box to different media formats such as 360&#x00B0; video, high framerate (HFR) content, or gaming videos. The analysis shows that the ITU-T P.1204.3 and Hybrid Mode 0 instances of <i>AV QBits</i> for the considered use-cases either perform on par with or better than even state-of-the-art full reference, pixel-based models. Furthermore, it is shown that the proposed Mode 0 and Mode 1 variants outperform commonly used no-reference models for the different application scopes. Also, a long-term integration model based on the standardized ITU-T P.1203.3 is presented to estimate ratings of overall audiovisual streaming Quality of Experience (QoE) for sessions of 30 s up to 5 min duration. In the paper, the <i>AV QBits</i> instances with their per-1-sec score output are evaluated as the video quality component of the proposed long-term integration model. All <i>AV QBits</i> variants as well as the long-term integration module are made publicly available for the community for further research.

Co-Clustering on Bipartite Graphs for Robust Model Fitting.

Recently, graph-based methods have been widely applied to model fitting. However, in these methods, association information is invariably lost when data points and model hypotheses are mapped to the graph domain. In this paper, we propose a novel model fitting method based on co-clustering on bipartite graphs (CBG) to estimate multiple model instances in data contaminated with outliers and noise. Model fitting is reformulated as a bipartite graph partition behavior. Specifically, we use a bipartite graph reduction technique to eliminate some insignificant vertices (outliers and invalid model hypotheses), thereby improving the reliability of the constructed bipartite graph and reducing the computational complexity. We then use a co-clustering algorithm to learn a structured optimal bipartite graph with exact connected components for partitioning that can directly estimate the model instances (i.e., post-processing steps are not required). The proposed method fully utilizes the duality of data points and model hypotheses on bipartite graphs, leading to superior fitting performance. Exhaustive experiments show that the proposed CBG method performs favorably when compared with several state-of-the-art fitting methods.

Data Boundary and Data Pricing Based on the Shapley Value

Data pricing is to price data as asset to promote the healthy development of data share, exchange, and reuse. However, the uncertain value index and neglection of interactivity lead to information asymmetry in the transaction process. A perfect data pricing system and a well-designed data market can widely promote data transactions. We take a three-agent data market in this paper, the data owner who provides the data record, the model buyer who is interested in buying machine learning (ML) model instances, and the broker who interacts between the data owner and the model buyer. Two interaction scenarios are defined. In scenario 1, we introduce the Shapley value (SV) to measure values of data records fairly and construct a revenue optimization problem based on the sum of the SV for data boundary. The optimal solution is obtained by calculating their derivatives. In scenario 2, we utilize market research and construct a revenue maximization (RM) problem to price the ML models. Further, An Integer Linear Programming for the RM problem (RM-ILP) process is proposed to transform the RM problem into an equivalent integer linear programming (ILP) problem and solve it with the Gurobi solver. Finally, we conduct extensive experiments, which validate that the RM-ILP process can provide high revenue to the broker and high affordability to the model buyers compared to the benchmarks.

Model architecture and tile size selection for convolutional neural network training for non-small cell lung cancer detection on whole slide images

Recent advancements in Artificial-Intelligence-based computer vision systems demonstrated impressive image and pattern recognition capabilities. A special class of neural networks known as Convolutional Neural Networks are used in a wide variety of computer vision tasks such as image classification, object detection and autonomous driving. There is a huge potential to adopt such technology in the domain of pathology. Image data in pathology are considerably larger in size than in typical image recognition problems. Dissecting the image into smaller bits, known as tiling, are often carried out. This paper aims to compare and contrast common model architectures and input tile sizes systematically and find the optimal configuration in the context of lung cancer classification problems.A dataset composed of 87 annotated whole slide images of lung cancer specimens were collected and annotated by two pathologists. Annotated areas were grouped into four classes (Tumor area, Non-tumor area, Necrosis area, and Immune cells). Annotations were converted into labelled tiles at different tile sizes, from 296 to 10000 pixels, (74–2500 μm). The problem was framed as a supervised 4-class classification problem using deep learning.For each tile size, three models, VGG19, InceptionResnetv2 and EfficientNet b3, were trained. Model performances were measured on holdout dataset using standard quantitative metrics including F1-score and AUC-ROC. Our best model instance with tile size at 500 × 500 pixels (125 × 125 μm) achieved an F1-score at 0.9685 and AUC-ROC score at 0.9627.Our results showed that tile size had a significant impact on model performance. The optimal tile size was between 500 and 1000 pixel (125–250 μm) after both quantitative and qualitative assessments. VGG19 marginally outperformed other model architectures.

An integrated open source CAT based on Skin Model Shapes

The control of geometrical deviations and form variations throughout the product life-cycle is a fundamental task in geometric dimensioning and tolerancing. As product complexity increases, it has not only become necessary to rely on computers to process geometrical and non-geometrical information from early design stages but also to come up with more realistic shape representation. Most of computer-aided tolerancing (CAT) packages used nowadays are fully integrated to computer-aided design softwares like CATIA and SolidWorks and they allow to model 2D and 3D tolerances stack-up through worst-case or statistical models. These CAT systems are generally available as proprietary commercial software which can sometimes restrict their domain of application and slow the implementation of new paradigms like the Skin Model. The Skin Model is an abstract surface model that represents the interface between a workpiece and its environment whose implementation involves the modeling of finite instances of the Skin Model called Skin Model Shapes (SMS) that encompass different sources of deviations and constitute a non-ideal geometrical model. The aim of this work is to show the first phase of implementation of an integrated open source environment based on PolitoCAT and Salome to model Skin Model Shapes. An Unified Model Language (UML) based logical data model of the integrated system is presented, it is an extended version of current data models for geometric modeling that includes the objects and relationships to manage form variations at different design stages. The work carried out contributes to the conceptualization of Skin Model Shapes model and it constitutes a support on the implementation of SMS in an open source CAT. An example of this integration involving a Skin Model Shapes implementation is shown as an illustration of the functionality of the platform.

Complex network near-synchronization for non-identical predator-prey systems

&lt;abstract&gt;&lt;p&gt;In this paper, we analyze the properties of a complex network of predator-prey systems, modeling the ecological dynamics of interacting species living in a fragmented environment. We consider non-identical instances of a Lotka-Volterra model with Holling type II functional response, which undergoes a Hopf bifurcation, and focus on the possible synchronization of distinct local behaviours. We prove an original result for the near-synchronization of non-identical systems, which shows how to and to what extent an extinction dynamic can be driven to a persistence equilibrium. Our theoretical statements are illustrated by appropriate numerical simulations.&lt;/p&gt;&lt;/abstract&gt;

Experimental error mitigation using linear rescaling for variational quantum eigensolving with up to 20 qubits

Quantum computers have the potential to help solve a range of physics and chemistry problems, but noise in quantum hardware currently limits our ability to obtain accurate results from the execution of quantum-simulation algorithms. Various methods have been proposed to mitigate the impact of noise on variational algorithms, including several that model the noise as damping expectation values of observables. In this work, we benchmark various methods, including a new method proposed here. We compare their performance in estimating the ground-state energies of several instances of the 1D mixed-field Ising model using the variational-quantum-eigensolver algorithm with up to 20 qubits on two of IBM’s quantum computers. We find that several error-mitigation techniques allow us to recover energies to within 10% of the true values for circuits containing up to about 25 ansatz layers, where each layer consists of CNOT gates between all neighboring qubits and Y-rotations on all qubits.

Developing a multi-label tinyML machine learning model for an active and optimized greenhouse microclimate control from multivariate sensed data

In the uncertainties within which the worldwide food security lies nowadays, the agricultural industry is raising a crucial need for being equipped with the state-of-the-art technologies for a more efficient, climate-resilient and sustainable production. The traditional production methods have to be revisited, and opportunities should be given for the innovative solutions henceforth brought by big data analytics, cloud computing and internet of things (IoT). In this context, we develop an optimized tinyML-oriented model for an active machine learning-based greenhouse microclimate management to be integrated in an on-field microcontroller. We design an experimental strawberry greenhouse from which we collect multivariate climate data through installed sensors. The obtained values' combinations are labeled according to a five-action multi-label control strategy, then used to prepare a machine learning-ready dataset. The dataset is used to train and five-fold cross-validate 90 Multi-Layer Perceptrons (MLPs) with varied hyperparameters to select the most performant –yet optimized– model instance for the addressed task. Our multi-label control approach enables designing highly scalable models with reduced computational complexity, comprising only n control neurons instead of (1 + ∑nk=1Cnk) neurons (usually generated from a classic single-label approach from n input variables). Our final selected model incorporates 2 hidden layers with 7 and 8 neurons respectively and 151 parameters; it scored a mean accuracy of 97% during the cross-validation phase, then 96% on our supplementary test set. The model enables an intelligent and autonomous greenhouse management with the less required computations. It can be efficiently deployed in microcontrollers within real world operating conditions.

A Transformational Approach to Managing Data Model Evolution of Web Services

The communication of web services is typically organized through APIs, which rely on a common data model shared among system components. Over time, this data model must be changed in order to accommodate new or changing requirements, and the components including the data they are operating on must be migrated. In practice however, not all affected components can be migrated instantly and at the same time. A common approach is to plan data model changes in a backward compatible fashion, which causes serious maintenance problems and is a common cause of technical debt. We propose an alternative solution, using a translation layer serving as a round-trip migration service, responsible for the lossless translation of object-oriented data model instances of different versions. We present a framework which offers a version-aware interface definition language (IDL) for APIs, a typed JavaScript-based language using the IDL definition, and a run-time environment. From a methodological point of view, development is supported by a catalog which comprises a set of typical evolution scenarios along with corresponding round-trip migration strategies. We showcase the applicability of our approach via a case study of a real-world e-commerce web application, and evaluate correctness through automated testing.

A spectrum-domain instance segmentation model for casting defects

Accurate segmentation of casting defects plays a positive role in the quality control of casting products, and is of great significance for accurate extraction of the mechanical properties of defects in the casting solidification process. However, as the shape of casting defects is complex and irregular, it is challenging to segment casting defects by existing segmentation methods. To address this, a spectrum domain instance segmentation model (SISN) is proposed for segmenting five types of casting defects with complex shapes accurately. The five defects are inclusion, shrinkage, hot tearing, cold tearing and micro pore. The proposed model consists of three sub-models: the spectrum domain region proposal model (SRPN), spectrum domain region of interest alignment model (SRoIAlign) and spectrum domain instance generation model (SIGN). SRPN uses a multi-scale anchoring mechanism to detect defects of various sizes, where the SSReLU and SCPool functions are used to solve the spectrum domain gradient explosion problem and the spectrum domain over-fitting problem. SRoIAlign uses the floating-point quantization operation and the tri-linear interpolation method to quantize the 3D proposals to the feature values in an accurate manner. SIGN is a full-spectrum domain neural network applied to 3D proposals, generating a segmentation instance of defects in a point-wise manner. In the experiments, we test the effectiveness of the proposed model from three aspects: segmentation accuracy, time performance and mechanical property extraction accuracy.

Predicting Forefoot-Orthosis Interactions in Rheumatoid Arthritis Using Computational Modelling.

Foot orthoses are prescribed to reduce forefoot plantar pressures and pain in people with rheumatoid arthritis. Computational modelling can assess how the orthoses affect internal tissue stresses, but previous studies have focused on a single healthy individual. This study aimed to ascertain whether simplified forefoot models would produce differing biomechanical predictions at the orthotic interface between people with rheumatoid arthritis of varying severity, and in comparison to a healthy control. The forefoot models were developed from magnetic resonance data of 13 participants with rheumatoid arthritis and one healthy individual. Measurements of bony morphology and soft tissue thickness were taken to assess deformity. These were compared to model predictions (99th% shear strain and plantar pressure, max. pressure gradient, volume of soft tissue over 10% shear strain), alongside clinical data including body mass index and Leeds Foot Impact Scale–Impairment/Footwear score (LFIS-IF). The predicted pressure and shear strain for the healthy participant fell at the lower end of the rheumatoid models’ range. Medial first metatarsal head curvature moderately correlated to all model predicted outcomes (0.529 < r < 0.574, 0.040 < p < 0.063). BMI strongly correlated to all model predictions except pressure gradients (0.600 < r < 0.652, p < 0.05). There were no apparent relationships between model predictions and instances of bursae, erosion and synovial hypertrophy or LFIS-IF score. The forefoot models produced differing biomechanical predictions between a healthy individual and participants with rheumatoid arthritis, and between individuals with rheumatoid arthritis. Models capable of predicting subject specific biomechanical orthotic interactions could be used in the future to inform more personalised devices to protect skin and soft tissue health. While the model results did not clearly correlate with all clinical measures, there was a wide range in model predictions and morphological measures across the participants. Thus, the need for assessment of foot orthoses across a population, rather than for one individual, is clear.

Modeling in semiotics: an integrative update

Abstract This paper provides an integrative and updated view of modeling in semiotics. It postulates that the essence of modeling is supersession. In any act or instance of modeling, the model supersedes and is brought to the front for salience, accessibility, and operability, whereas at the same time the modeled recedes and exists in the background, inaccessible and inoperable. The paper goes on to differentiate between two major types of modeling, the underlying “existential modeling,” functioning as the fundamental scaffold and the genuine foundation of all other types of modeling as we know them, and the overlaying “semiotic modeling,” designating the process of creation and use of “forms of meaning,” a process that underlies both cognition and communication. By focusing on semiotic modeling, the paper features an unconventional view that casts a new light on the relation between a model and a sign and thus the relation between semiotic modeling and semiosis. Endorsing an embodied approach to meaning-making as semiotic modeling, the paper finds it important to stress the appropriateness and necessity of understanding the term “model” as a verb rather than as a noun, in that modeling is never static and should be properly regarded in terms of embodied action.

TOrPEDO: witnessing model correctness with topological proofs

Model design is not a linear, one-shot process. It proceeds throughrefinements and revisions. To effectively support developers ingenerating model refinements and revisions, it is desirable to havesome automated support to verify evolvable models. To address thisproblem, we recently proposed to adopt topological proofs,which are slices of the original model that witness propertysatisfaction. We implemented TOrPEDO, a framework that providesautomated support for using topological proofs during model design.Our results showed that topological proofs are significantly smallerthan the original models, and that, in most of the cases, they allowthe property to be re-verified by relying only on a simple syntacticcheck. However, our results also show that the procedure thatcomputes topological proofs, which requires extracting unsatisfiablecores of LTL formulae, is computationally expensive. For thisreason, TOrPEDO currently handles models with a small dimension. Withthe intent of providing practical and efficient support for flexiblemodel design and wider adoption of our framework, in this paper, wepropose an enhanced—re-engineered—version of TOrPEDO. The newversion of TOrPEDO relies on a novel procedure to extracttopological proofs, which has so far represented the bottleneck ofTOrPEDO performances. We implemented our procedure within TOrPEDO byconsidering Partial Kripke Structures (PKSs) and Linear-timeTemporal Logic (LTL): two widely used formalisms to express modelswith uncertain parts and their properties. To extract topologicalproofs, the new version of TOrPEDO converts the LTL formulae into anSMT instance and reuses an existing SMT solver (e.g., MicrosoftZ3) to compute an unsatisfiable core. Then, theunsatisfiable core returned by the SMT solver is automaticallyprocessed to generate the topological proof. We evaluated TOrPEDO byassessing (i) how does the size of the proofs generated by TOrPEDOcompares to the size of the models being analyzed; and (ii) howfrequently the use of the topological proof returned by TOrPEDOavoids re-executing the model checker. Our results show that TOrPEDOprovides proofs that are smaller (approx 60%) than theirrespective initial models effectively supporting designers increating model revisions. In a significant number of cases (approx 79%), the topological proofs returned by TOrPEDO enable assessingthe property satisfaction without re-running the model checker. Weevaluated our new version of TOrPEDO by assessing (i) how it comparesto the previous one; and (ii) how useful it is in supporting theevaluation of alternative design choices of (small) model instancesin applied domains. The results show that the new version of TOrPEDOis significantly more efficient than the previous one and cancompute topological proofs for models with less than 40 stateswithin two hours. The topological proofs and counterexamplesprovided by TOrPEDO are useful to support the development ofalternative design choices of (small) model instances in applieddomains.

Model-based configuration of access protection units for multicore processors in embedded systems

Multiprocessor system-on-chip platforms are becoming increasingly common in embedded systems. To facilitate a logical isolation of physically connected on-chip components, internal communication links of such platforms are often equipped with dedicated access protection units. When performed manually, the configuration of these units is both time-consuming and error-prone. Therefore, we present a model-based design methodology that allows designers to describe envisaged communication link transactions as well as the desired information flow policy in a platform-independent manner. After automatically verifying the consistency between specified transactions and the declared policy, the proposed toolchain delegates model instances to an extensible generation framework. This framework translates the envisaged transactions into configuration code for access protection units of particular platforms. We give a formal description of both the model and the verification procedure, describe the platform-specific generation framework, and evaluate the approach by applying it to an example scenario from the field of autonomous driving.

NeuroGPU: Accelerating multi-compartment, biophysically detailed neuron simulations on GPUs

BackgroundThe membrane potential of individual neurons depends on a large number of interacting biophysical processes operating on spatial-temporal scales spanning several orders of magnitude. The multi-scale nature of these processes dictates that accurate prediction of membrane potentials in specific neurons requires the utilization of detailed simulations. Unfortunately, constraining parameters within biologically detailed neuron models can be difficult, leading to poor model fits. This obstacle can be overcome partially by numerical optimization or detailed exploration of parameter space. However, these processes, which currently rely on central processing unit (CPU) computation, often incur orders of magnitude increases in computing time for marginal improvements in model behavior. As a result, model quality is often compromised to accommodate compute resources. New MethodHere, we present a simulation environment, NeuroGPU, that takes advantage of the inherent parallelized structure of the graphics processing unit (GPU) to accelerate neuronal simulation. Results & comparison with existing methodsNeuroGPU can simulate most biologically detailed models 10–200 times faster than NEURON simulation running on a single core and 5 times faster than GPU simulators (CoreNEURON). NeuroGPU is designed for model parameter tuning and best performs when the GPU is fully utilized by running multiple (> 100) instances of the same model with different parameters. When using multiple GPUs, NeuroGPU can reach to a speed-up of 800 fold compared to single core simulations, especially when simulating the same model morphology with different parameters. We demonstrate the power of NeuoGPU through large-scale parameter exploration to reveal the response landscape of a neuron. Finally, we accelerate numerical optimization of biophysically detailed neuron models to achieve highly accurate fitting of models to simulation and experimental data. ConclusionsThus, NeuroGPU is the fastest available platform that enables rapid simulation of multi-compartment, biophysically detailed neuron models on commonly used computing systems accessible by many scientists.