30 research outputs found
A computational study on the hydration-shell properties of antifreeze and non-antifreeze proteins
No full textHere we present a computational approach based on molecular dynamics (MD)
simulation to study the hydration-shell density of several proteins which include a
special group of proteins, namely antifreeze proteins, AFPs. AFPs have the ability
to inhibit ice growth by binding to ice nuclei. Their ice-binding mechanism is still
unclear, yet the hydration layer is thought to play a fundamental role. In particular,
the hydration-shell density of eighteen dierent proteins comprising eight AFPs is
calculated. The results obtained show that an increase in the hydration-shell density,
relative to that of the bulk, is observed (in the range of 4{14%) for all studied proteins
and that this increment strongly correlates with the protein size, while it does not
depend on whether the protein is an AFP or not. In particular, a decrease in the
density increment is observed for decreasing protein size. A simple model is proposed
according to which almost all of the hydration-density increase is located in pockets
within, or at the surface of, the protein molecule. We then further investigated the
local properties of the hydration shell around the ice-binding surface (IBS) of the
AFPs. We found that the hydration shell density of the ice-binding surfaces is always
higher than the bulk density and, thus, no ice-like (i.e. with a density lower than the
bulk) layer is detected at the IBS. However, the local water-density around the IBS
is found to be lower than that around the non-ice-binding surfaces and this dierence
correlates to the higher hydrophobic character of the IBS with respect to the non-IBS.
We hypothesize that the lower solvent density at the ice-binding site can pave the way
to the protein binding to ice nuclei, while the higher solvent density at the non-icebinding
surfaces might provide protection against ice growth. Finally, we tested our
hypothesis by studying the dependence of the antifreeze activity of seven AFPs on
various structural and chemical properties of the IBS and non-IBS and found that the
activity strongly correlates with the dierence in the local hydration-shell properties
of the non-ice-binding surfaces, rather than of the IBSs
Surface assembly of nano-metalorganic framework on amine functionalized indium tin oxide substrate for impedimetric sensing of parathion
No full textThe present paper reports the assembly and pesticide sensing application of a nanometal organic framework [Cd(atc)(H2O)2]n (‘atc’=2-aminoterephthalic acid). The assembly of the NMOF film has been achieved by sequential dipping of a 2-aminobenzylamine (2-ABA) modified indium tin oxide (ITO) slide in organic linker ‘atc’ and metal ion ‘Cd2+’ solutions. The different structural and morphological characteristics of the NMOF thin film have been characterized. The availability of pendent –COOH functional groups on the assembled NMOF film is exploited to synthesize a pesticide immunosensor by conjugating the NMOF film with anti-parathion antibody. This immunosensor has been explored for the electrochemical impedance spectroscopy (EIS) based analysis of parathion in the concentration range of 0.1–20 ng/mL. The proposed detection is specific with respect to other organophosphate compounds, e.g. malathion, paraoxon, fenitrothion, monochrotophos and dichlorovos. The proposed sensor shows the detection limit of 0.1 ng/mL and it is applicable for analysis of parathion in a rice sample. The sensor's performance is validated by comparting the obtained results with gas chromatographic data.Authors gratefully acknowledge the financial grant from CSIR India through project OMEGA/PSC0202/2.2.5. We are thankful to the Director, CSIR-CSIO, Chandigarh, India. The fourth author acknowledges partial financial support from the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (MEST) (No. 2009-0093848)
Polycyclic aromatic hydrocarbons and volatile organic compounds in biochar and biochar-amended soil:a review
No full textResidual pollutants including polycyclic aromatic hydrocarbons (PAHs), volatile organic compounds (VOCs), and carbon (aceous) nanoparticles are inevitably generated during the pyrolysis of waste biomass and remain on the solid coproduct called biochar. Such pollutants could have adverse effects on the plant growth as well as microbial community in soil. Although biochar has been proposed as a ‘carbon negative strategy’ to mitigate the greenhouse gas emissions, the impacts of its application with respect to long-term persistence and bioavailability of hazardous components are not clear. Moreover, the co-occurrence of low molecular weight VOCs with PAHs in biochar may exert further phytotoxic effects. This review describes the basic need to unravel key mechanisms driving the storage vs. emission of these organics and the dynamics between the sorbent (biochar) and soil microbes. Moreover, there is an urgent need for standardized methods for quantitative analysis of PAHs and VOCs in biochar under environmentally relevant conditions. This review is also extended to cover current research gaps including the influence of biochar application on the short- and long-term fate of PAHs and VOCs and the proper control tactics for biochar quality and associated risk.This study was supported by a grant from the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (MEST) (No. 2009-0093848). This work was also carried out with the support of the 'Cooperative Research Program for Agriculture Science & Technology Development' (Project title: Study on model development to control odor from hogpens, Project No. PJ01052101) Rural Development Administration, Republic of Korea. The second author also acknowledges the support made by a National Research Foundation of Korea (NRF) Grant funded by the Korean Government (MSIP) (No. 2014RA1A004893)
Target Prediction by Multiple Virtual Screenings: Analyzing the SARS-CoV-2 Phenotypic Screening by the Docking Simulations Submitted to the MEDIATE Initiative
No full textPhenotypic screenings are usually combined with deconvolution techniques to characterize the mechanism of action for the retrieved hits. These studies can be supported by various computational analyses, although docking simulations are rarely employed. The present study aims to assess if multiple docking calculations can prove successful in target prediction. In detail, the docking simulations submitted to the MEDIATE initiative are utilized to predict the viral targets involved in the hits retrieved by a recently published cytopathic screening. Multiple docking results are combined by the EFO approach to develop target-specific consensus models. The combination of multiple docking simulations enhances the performances of the developed consensus models (average increases in EF1% value of 40% and 25% when combining three and two docking runs, respectively). These models are able to propose reliable targets for about half of the retrieved hits (31 out of 59). Thus, the study emphasizes that docking simulations might be effective in target identification and provide a convincing validation for the collaborative strategies that inspire the MEDIATE initiative. Disappointingly, cross-target and cross-program correlations suggest that common scoring functions are not specific enough for the simulated target
Length-scale dependence of protein hydration-shell density
No full textHere we present a computational approach based on molecular dynamics (MD) simulation to study the dependence of the protein hydration-shell density on the size of the protein molecule. The hydration-shell density of eighteen different proteins, differing in size, shape and function (eight of them are antifreeze proteins), is calculated. The results obtained show that an increase in the hydration-shell density, relative to that of the bulk, is observed (in the range of 4-14%) for all studied proteins and that this increment strongly correlates with the protein size. In particular, a decrease in the density increment is observed for decreasing protein size. A simple model is proposed in which the basic idea is to approximate the protein molecule as an effective ellipsoid and to partition the relevant parameters, i.e. the solvent-accessible volume and the corresponding solvent density, into two regions: inside and outside the effective protein ellipsoid. It is found that, within the model developed here, almost all of the hydration-density increase is located inside the protein ellipsoid, basically corresponding to pockets within, or at the surface of the protein molecule. The observed decrease in the density increment is caused by the protein size only and no difference is found between antifreeze and non-antifreeze proteins
Advanced financial market forecasting: integrating Monte Carlo simulations with ensemble Machine Learning models
No full text© 2024 the Author(s), licensee AIMS Press. cc-byThis paper presents a novel integration of Machine Learning (ML) models with Monte Carlo simulations to enhance financial forecasting and risk assessments in dynamic market environments. Traditional financial forecasting methods, which primarily rely on linear statistical and econometric models, face limitations in addressing the complexities of modern financial datasets. To overcome these challenges, we explore the evolution of financial forecasting, transitioning from time-series analyses to sophisticated ML techniques such as Random Forest, Support Vector Machines, and Long Short-Term Memory (LSTM) networks. Our methodology combines an ensemble of these ML models, each providing unique insights into market dynamics, with the probabilistic scenario analysis of Monte Carlo simulations. This integration aims to improve the predictive accuracy and risk evaluation in financial markets. We apply this integrated approach to a quantitative analysis of the SPY Exchange-Traded Fund (ETF) and selected major stocks, focusing on various risk-reward ratios including Sharpe, Sortino, and Treynor. The results demonstrate the potential of our approach in providing a comprehensive view of risks and rewards, highlighting the advantages of combining traditional risk assessment methods with advanced predictive models. This research contributes to the field of applied mathematical finance by offering a more nuanced, adaptive tool for financial market analyses and decision-making
Post-Training Optimization of Cross-layer Approximate Computing for Edge Inference of Deep Learning Applications
No full textOver the past decade, the rapid development of deep learning (DL) algorithms has enabled extraordinary advances in perception tasks throughout different fields, from computer vision to audio signal processing. Additionally, increasing computational resources available in supercomputers and graphic processor clusters have provided a suitable environment to train larger and deeper deep neural network (DNN) models for improved performances. However, the resulting memory bandwidth and computational requirements of such DNN models restricts their deployment in embedded systems with constrained hardware resources.
To overcome this challenge, it is important to establish new paradigms to reduce the computational workload of such DL algorithms while maintaining their original accuracy. A key observation of previous research is that DL models are resilient to input noise and computational errors; therefore, a reasonable approach to decreasing such hardware requirements is to embrace DNN resiliency and utilize approximate computing techniques at different system design layers. This approach requires, however, constant monitoring as well as a careful combination of approximation techniques to avoid performance degradation while maximizing computational savings. Within this context, the focus of this thesis is the simulation of cross-layer approximate computing (AC) methods for DNN computation and the development of optimization methods to compensate AC errors in approximated DNNs.
The first part of this thesis proposes the simulation framework ProxSim. This framework enables accelerated approximate computational unit (ACU) simulation for evaluation and training of approximated DNNs. ProxSim supports quantization and approximation of common neural layers such as fully connected (FC), convolutional, and recurrent layers. A performance evaluation using a variety of DNN architectures, as well as a comparison with the state of the art is also presented. The author used ProxSim to implement and evaluate the following methods presented in this work.
The second part of this thesis introduces an approach to model the approximation error in DNN computation. First, the author thoroughly anaylzes the error caused by approximate multipliers to compute the multiply and accumulate (MAC) operations in DNN models. From this analysis, a statistical model of the approximation error is obtained. Through various experiments with DNNs for image classification, the proposed model is verified and compared with other methods from the literature. The results demonstrate the validity of the approximation error model and reinforce a general understanding of approximate computing in DNNs.
In the third part of this thesis, the author presents a methodology for uniform systematic approximation of DNNs. This methodology focuses on the optimization of full DNN approximation with a single type of ACU to minimize power consumption without accuracy loss. The backbone of this methodology is the custom fine-tuning methods the author proposes to compensate for the approximation error. These methods enable the use of ACUs with large approximation errors, which results in significant power savings and negligible accuracy losses. This process is corroborated by extensive experiments, where the estimated savings and the accuracy achieved after approximation are thoroughly examined using ProxSim.
In the last part of this thesis, the author proposes two different methodologies to further boost energy savings after applying uniform approximation. This increment in energy savings is achieved by computing more resilient DNN elements (neurons or layers) with increased approximation levels. The first methodology focuses on iterative kernel-wise approximation and quantization enabled by a custom approximate MAC unit. The second method is based on flexible layer-wise approximation, and applied to bit-decomposed in-memory computing (IMC) architectures as a case study to demonstrate the effectiveness of the proposed approach
Post-Training Optimization of Cross-layer Approximate Computing for Edge Inference of Deep Learning Applications
No full textOver the past decade, the rapid development of deep learning (DL) algorithms has enabled extraordinary advances in perception tasks throughout different fields, from computer vision to audio signal processing. Additionally, increasing computational resources available in supercomputers and graphic processor clusters have provided a suitable environment to train larger and deeper deep neural network (DNN) models for improved performances. However, the resulting memory bandwidth and computational requirements of such DNN models restricts their deployment in embedded systems with constrained hardware resources.
To overcome this challenge, it is important to establish new paradigms to reduce the computational workload of such DL algorithms while maintaining their original accuracy. A key observation of previous research is that DL models are resilient to input noise and computational errors; therefore, a reasonable approach to decreasing such hardware requirements is to embrace DNN resiliency and utilize approximate computing techniques at different system design layers. This approach requires, however, constant monitoring as well as a careful combination of approximation techniques to avoid performance degradation while maximizing computational savings. Within this context, the focus of this thesis is the simulation of cross-layer approximate computing (AC) methods for DNN computation and the development of optimization methods to compensate AC errors in approximated DNNs.
The first part of this thesis proposes the simulation framework ProxSim. This framework enables accelerated approximate computational unit (ACU) simulation for evaluation and training of approximated DNNs. ProxSim supports quantization and approximation of common neural layers such as fully connected (FC), convolutional, and recurrent layers. A performance evaluation using a variety of DNN architectures, as well as a comparison with the state of the art is also presented. The author used ProxSim to implement and evaluate the following methods presented in this work.
The second part of this thesis introduces an approach to model the approximation error in DNN computation. First, the author thoroughly anaylzes the error caused by approximate multipliers to compute the multiply and accumulate (MAC) operations in DNN models. From this analysis, a statistical model of the approximation error is obtained. Through various experiments with DNNs for image classification, the proposed model is verified and compared with other methods from the literature. The results demonstrate the validity of the approximation error model and reinforce a general understanding of approximate computing in DNNs.
In the third part of this thesis, the author presents a methodology for uniform systematic approximation of DNNs. This methodology focuses on the optimization of full DNN approximation with a single type of ACU to minimize power consumption without accuracy loss. The backbone of this methodology is the custom fine-tuning methods the author proposes to compensate for the approximation error. These methods enable the use of ACUs with large approximation errors, which results in significant power savings and negligible accuracy losses. This process is corroborated by extensive experiments, where the estimated savings and the accuracy achieved after approximation are thoroughly examined using ProxSim.
In the last part of this thesis, the author proposes two different methodologies to further boost energy savings after applying uniform approximation. This increment in energy savings is achieved by computing more resilient DNN elements (neurons or layers) with increased approximation levels. The first methodology focuses on iterative kernel-wise approximation and quantization enabled by a custom approximate MAC unit. The second method is based on flexible layer-wise approximation, and applied to bit-decomposed in-memory computing (IMC) architectures as a case study to demonstrate the effectiveness of the proposed approach
Post-Training Optimization of Cross-layer Approximate Computing for Edge Inference of Deep Learning Applications
No full textOver the past decade, the rapid development of deep learning (DL) algorithms has enabled extraordinary advances in perception tasks throughout different fields, from computer vision to audio signal processing. Additionally, increasing computational resources available in supercomputers and graphic processor clusters have provided a suitable environment to train larger and deeper deep neural network (DNN) models for improved performances. However, the resulting memory bandwidth and computational requirements of such DNN models restricts their deployment in embedded systems with constrained hardware resources.
To overcome this challenge, it is important to establish new paradigms to reduce the computational workload of such DL algorithms while maintaining their original accuracy. A key observation of previous research is that DL models are resilient to input noise and computational errors; therefore, a reasonable approach to decreasing such hardware requirements is to embrace DNN resiliency and utilize approximate computing techniques at different system design layers. This approach requires, however, constant monitoring as well as a careful combination of approximation techniques to avoid performance degradation while maximizing computational savings. Within this context, the focus of this thesis is the simulation of cross-layer approximate computing (AC) methods for DNN computation and the development of optimization methods to compensate AC errors in approximated DNNs.
The first part of this thesis proposes the simulation framework ProxSim. This framework enables accelerated approximate computational unit (ACU) simulation for evaluation and training of approximated DNNs. ProxSim supports quantization and approximation of common neural layers such as fully connected (FC), convolutional, and recurrent layers. A performance evaluation using a variety of DNN architectures, as well as a comparison with the state of the art is also presented. The author used ProxSim to implement and evaluate the following methods presented in this work.
The second part of this thesis introduces an approach to model the approximation error in DNN computation. First, the author thoroughly anaylzes the error caused by approximate multipliers to compute the multiply and accumulate (MAC) operations in DNN models. From this analysis, a statistical model of the approximation error is obtained. Through various experiments with DNNs for image classification, the proposed model is verified and compared with other methods from the literature. The results demonstrate the validity of the approximation error model and reinforce a general understanding of approximate computing in DNNs.
In the third part of this thesis, the author presents a methodology for uniform systematic approximation of DNNs. This methodology focuses on the optimization of full DNN approximation with a single type of ACU to minimize power consumption without accuracy loss. The backbone of this methodology is the custom fine-tuning methods the author proposes to compensate for the approximation error. These methods enable the use of ACUs with large approximation errors, which results in significant power savings and negligible accuracy losses. This process is corroborated by extensive experiments, where the estimated savings and the accuracy achieved after approximation are thoroughly examined using ProxSim.
In the last part of this thesis, the author proposes two different methodologies to further boost energy savings after applying uniform approximation. This increment in energy savings is achieved by computing more resilient DNN elements (neurons or layers) with increased approximation levels. The first methodology focuses on iterative kernel-wise approximation and quantization enabled by a custom approximate MAC unit. The second method is based on flexible layer-wise approximation, and applied to bit-decomposed in-memory computing (IMC) architectures as a case study to demonstrate the effectiveness of the proposed approach
