scholarly journals An Improving Method for Performance of DNN using 2-bit Quantization

2021 ◽  
Vol 12 (6) ◽  
pp. 597-601
Author(s):  
Sang Hyeok Kim Et.al
Keyword(s):  
Matrix Multiplication ◽  
Complex Data ◽  
Mobile Environments ◽  
Data Set ◽  
Additional Function ◽  
Connected Part ◽  
The Difference ◽  
Fully Connected ◽  
New Algorithms ◽  
Additional Hardware

Background/Objectives: Recently, interest in AI(Artificial Intelligence) has increased, and  many studies are being conducted to enable AI to be used in embedded and mobile environments. Among them, quantization is one of the methods to reduce the size of the model, and most quantization of less than 8 bits cannot be implemented without additional hardware such as FPGA. With this in mind, in this paper, we propose two new algorithms that can implement 2bit quantization in software. Methods/Statistical analysis: In this paper, we propose a packing operation that quantizes a weight consisting of 32-bit real values into 2 bits, stores four 2-bit quantization weights in one 8-bit memory, and a Masking Matrix Multiplication function that performs the calculation of the packed weight and input values. These functions operate in parallel in the GPU memory. Findings: The quantization model using the above function showed about 16 times more memory saving and 4 times faster when comparing the operation with the existing 32bit model. Nevertheless, the DNN model showed an error of around 1% in learning using MNIST and HandWritten data, and the CNN model showed an error of around 1% in learning using EEG (Electroencephalograpy) data. Improvements/Applications: The function used in this study is focused on the domain of DNN, and although extended to CNN, quantization could be performed only in the FC (Fully Connected) part. To apply to the convolution layer, an additional function is required, and it is necessary to check whether the difference in accuracy is small even in a more complex data set in the future.

Difference Fourier Analysis of Glucose Embedded and Frozen Hydrated Purple Membrane

1982 ◽  
Vol 40 ◽  
pp. 74-75
Author(s):  
Jules S. Jaffe ◽  
Robert M. Glaeser
Keyword(s):  
Purple Membrane ◽  
Data Set ◽  
High Resolution Data ◽  
X Ray ◽  
X Ray Crystallography ◽  
Fourier Techniques ◽  
Versus Protein ◽  
The Difference ◽  
Difference Fourier ◽  
Ideal Method

Although difference Fourier techniques are standard in X-ray crystallography it has only been very recently that electron crystallographers have been able to take advantage of this method. We have combined a high resolution data set for frozen glucose embedded Purple Membrane (PM) with a data set collected from PM prepared in the frozen hydrated state in order to visualize any differences in structure due to the different methods of preparation. The increased contrast between protein-ice versus protein-glucose may prove to be an advantage of the frozen hydrated technique for visualizing those parts of bacteriorhodopsin that are embedded in glucose. In addition, surface groups of the protein may be disordered in glucose and ordered in the frozen state. The sensitivity of the difference Fourier technique to small changes in structure provides an ideal method for testing this hypothesis.

scholarly journals 136. Don’t Sweat the Small Stuff: Solutions for Large-Scale Stewardship Obstacles

2020 ◽  
Vol 7 (Supplement_1) ◽  
pp. S79-S80
Author(s):  
Joanne Huang ◽  
Zahra Kassamali Escobar ◽  
Rupali Jain ◽  
Jeannie D Chan ◽  
John B Lynch ◽  
...  
Keyword(s):  
Urgent Care ◽  
Complex Data ◽  
Open Communication ◽  
University Of Washington ◽  
Data Set ◽  
Constructive Feedback ◽  
Diagnostic Coding ◽  
Icd 10 ◽  
Tract Infections ◽  
The University

Abstract Background In an effort to support stewardship endeavors, the MITIGATE (a Multifaceted Intervention to Improve Prescribing for Acute Respiratory Infection for Adult and Children in Emergency Department and Urgent Care Settings) Toolkit was published in 2018, aiming to reduce unnecessary antibiotics for viral respiratory tract infections (RTIs). At the University of Washington, we have incorporated strategies from this toolkit at our urgent care clinics. This study aims to address solutions to some of the challenges we experienced. Challenges and Solutions Methods This was a retrospective observational study conducted at Valley Medical Center (Sept 2019-Mar 2020) and the University of Washington (Jan 2019-Feb 2020) urgent care clinics. Patients were identified through ICD-10 diagnosis codes included in the MITIGATE toolkit. The primary outcome was identifying challenges and solutions developed during this process. Results We encountered five challenges during our roll-out of MITIGATE. First, using both ICD-9 and ICD-10 codes can lead to inaccurate data collection. Second, technical support for coding a complex data set is essential and should be accounted for prior to beginning stewardship interventions of this scale. Third, unintentional incorrect diagnosis selection was common and may require reeducation of prescribers on proper selection. Fourth, focusing on singular issues rather than multiple outcomes is more feasible and can offer several opportunities for stewardship interventions. Lastly, changing prescribing behavior can cause unintended tension during implementation. Modifying benchmarks measured, allowing for bi-directional feedback, and identifying provider champions can help maintain open communication. Conclusion Resources such as the MITIGATE toolkit are helpful to implement standardized data driven stewardship interventions. We have experienced some challenges including a complex data build, errors with diagnostic coding, providing constructive feedback while maintaining positive stewardship relationships, and choosing feasible outcomes to measure. We present solutions to these challenges with the aim to provide guidance to those who are considering using this toolkit for outpatient stewardship interventions. Disclosures All Authors: No reported disclosures

Event detection of different English data sources based on transfer learning

2021 ◽  
pp. 1-11
Author(s):  
Yanan Huang ◽  
Yuji Miao ◽  
Zhenjing Da
Keyword(s):  
Transfer Learning ◽  
Event Detection ◽  
Visual Analysis ◽  
Learning Algorithm ◽  
Data Sources ◽  
Data Set ◽  
Data Source ◽  
Single Data Source ◽  
The Difference ◽  
Single Data

The methods of multi-modal English event detection under a single data source and isomorphic event detection of different English data sources based on transfer learning still need to be improved. In order to improve the efficiency of English and data source time detection, based on the transfer learning algorithm, this paper proposes multi-modal event detection under a single data source and isomorphic event detection based on transfer learning for different data sources. Moreover, by stacking multiple classification models, this paper makes each feature merge with each other, and conducts confrontation training through the difference between the two classifiers to further make the distribution of different source data similar. In addition, in order to verify the algorithm proposed in this paper, a multi-source English event detection data set is collected through a data collection method. Finally, this paper uses the data set to verify the method proposed in this paper and compare it with the current most mainstream transfer learning methods. Through experimental analysis, convergence analysis, visual analysis and parameter evaluation, the effectiveness of the algorithm proposed in this paper is demonstrated.

Magnetotelluric static shift: Estimation and removal using the cokriging method

Geophysics ◽  
2007 ◽  
Vol 72 (1) ◽  
pp. F25-F34 ◽  
Author(s):  
Benoit Tournerie ◽  
Michel Chouteau ◽  
Denis Marcotte
Keyword(s):  
Apparent Resistivity ◽  
A Priori ◽  
Shift Factor ◽  
Static Shift ◽  
Geostatistical Model ◽  
Data Set ◽  
Resistivity Sounding ◽  
The Difference ◽  
Cross Variogram

We present and test a new method to correct for the static shift affecting magnetotelluric (MT) apparent resistivity sounding curves. We use geostatistical analysis of apparent resistivity and phase data for selected periods. For each period, we first estimate and model the experimental variograms and cross variogram between phase and apparent resistivity. We then use the geostatistical model to estimate, by cokriging, the corrected apparent resistivities using the measured phases and apparent resistivities. The static shift factor is obtained as the difference between the logarithm of the corrected and measured apparent resistivities. We retain as final static shift estimates the ones for the period displaying the best correlation with the estimates at all periods. We present a 3D synthetic case study showing that the static shift is retrieved quite precisely when the static shift factors are uniformly distributed around zero. If the static shift distribution has a nonzero mean, we obtained best results when an apparent resistivity data subset can be identified a priori as unaffected by static shift and cokriging is done using only this subset. The method has been successfully tested on the synthetic COPROD-2S2 2D MT data set and on a 3D-survey data set from Las Cañadas Caldera (Tenerife, Canary Islands) severely affected by static shift.

scholarly journals What Countries Select More Experienced Leaders? The PolEx Measure of Political Experience

2021 ◽  
pp. 1-10
Author(s):  
Alexander Baturo ◽  
Johan A. Elkink
Keyword(s):  
Career Paths ◽  
Latent Variable ◽  
Empirical Studies ◽  
Future Research ◽  
Variable Model ◽  
Data Set ◽  
Political Experience ◽  
National Leaders ◽  
Regime Types ◽  
The Difference

Abstract How can one assess which countries select more experienced leaders for the highest office? There is wide variation in prior career paths of national leaders within, and even more so between, regime types. It is therefore challenging to obtain a truly comparative measure of political experience; empirical studies have to rely on proxies instead. This article proposes PolEx, a measure of political experience that abstracts away from the details of career paths and generalizes based on the duration, quality and breadth of an individual's experience in politics. The analysis draws on a novel data set of around 2,000 leaders from 1950 to 2017 and uses a Bayesian latent variable model to estimate PolEx. The article illustrates how the new measure can be used comparatively to assess whether democracies select more experienced leaders. The authors find that while on average they do, the difference with non-democracies has declined dramatically since the early 2000s. Future research may leverage PolEx to investigate the role of prior political experience in, for example, policy making and crisis management.

scholarly journals DEELIG: A Deep Learning Approach to Predict Protein-Ligand Binding Affinity

2021 ◽  
Vol 15 ◽  
pp. 117793222110303
Author(s):  
Asad Ahmed ◽  
Bhavika Mam ◽  
Ramanathan Sowdhamini
Keyword(s):  
Deep Learning ◽  
Ligand Binding ◽  
Binding Affinity ◽  
Chemical Space ◽  
Biological Significance ◽  
Protein Crystal ◽  
Ligand Docking ◽  
Complex Data ◽  
Binding Prediction ◽  
Data Set

Protein-ligand binding prediction has extensive biological significance. Binding affinity helps in understanding the degree of protein-ligand interactions and is a useful measure in drug design. Protein-ligand docking using virtual screening and molecular dynamic simulations are required to predict the binding affinity of a ligand to its cognate receptor. Performing such analyses to cover the entire chemical space of small molecules requires intense computational power. Recent developments using deep learning have enabled us to make sense of massive amounts of complex data sets where the ability of the model to “learn” intrinsic patterns in a complex plane of data is the strength of the approach. Here, we have incorporated convolutional neural networks to find spatial relationships among data to help us predict affinity of binding of proteins in whole superfamilies toward a diverse set of ligands without the need of a docked pose or complex as user input. The models were trained and validated using a stringent methodology for feature extraction. Our model performs better in comparison to some existing methods used widely and is suitable for predictions on high-resolution protein crystal (⩽2.5 Å) and nonpeptide ligand as individual inputs. Our approach to network construction and training on protein-ligand data set prepared in-house has yielded significant insights. We have also tested DEELIG on few COVID-19 main protease-inhibitor complexes relevant to the current public health scenario. DEELIG-based predictions can be incorporated in existing databases including RSCB PDB, PDBMoad, and PDBbind in filling missing binding affinity data for protein-ligand complexes.

Investigation of Regression-Based Effect Size Methods Developed in Single-Subject Studies

2021 ◽  
pp. 014544552110540
Author(s):  
Nihal Sen
Keyword(s):  
Effect Size ◽  
Experimental Studies ◽  
Single Subject ◽  
Single Subject Design ◽  
Data Set ◽  
Design Studies ◽  
Sample Data ◽  
The Difference ◽  
D Values ◽  
Effect Size Calculation

The purpose of this study is to provide a brief introduction to effect size calculation in single-subject design studies, including a description of nonparametric and regression-based effect sizes. We then focus the rest of the tutorial on common regression-based methods used to calculate effect size in single-subject experimental studies. We start by first describing the difference between five regression-based methods (Gorsuch, White et al., Center et al., Allison and Gorman, Huitema and McKean). This is followed by an example using the five regression-based effect size methods and a demonstration how these methods can be applied using a sample data set. In this way, the question of how the values obtained from different effect size methods differ was answered. The specific regression models used in these five regression-based methods and how these models can be obtained from the SPSS program were shown. R2 values obtained from these five methods were converted to Cohen’s d value and compared in this study. The d values obtained from the same data set were estimated as 0.003, 0.357, 2.180, 3.470, and 2.108 for the Allison and Gorman, Gorsuch, White et al., Center et al., as well as for Huitema and McKean methods, respectively. A brief description of selected statistical programs available to conduct regression-based methods was given.

scholarly journals Accelerating the HyperLogLog Cardinality Estimation Algorithm

2017 ◽  
Vol 2017 ◽  
pp. 1-8
Author(s):  
Cem Bozkus ◽  
Basilio B. Fraguela
Keyword(s):  
Large Data ◽  
Estimation Algorithm ◽  
Xeon Phi ◽  
Memory Usage ◽  
Intel Xeon Phi ◽  
Data Set ◽  
Internet Routers ◽  
Multicore System ◽  
Sequential Implementation ◽  
New Algorithms

In recent years, vast amounts of data of different kinds, from pictures and videos from our cameras to software logs from sensor networks and Internet routers operating day and night, are being generated. This has led to new big data problems, which require new algorithms to handle these large volumes of data and as a result are very computationally demanding because of the volumes to process. In this paper, we parallelize one of these new algorithms, namely, the HyperLogLog algorithm, which estimates the number of different items in a large data set with minimal memory usage, as it lowers the typical memory usage of this type of calculation from O(n) to O(1). We have implemented parallelizations based on OpenMP and OpenCL and evaluated them in a standard multicore system, an Intel Xeon Phi, and two GPUs from different vendors. The results obtained in our experiments, in which we reach a speedup of 88.6 with respect to an optimized sequential implementation, are very positive, particularly taking into account the need to run this kind of algorithm on large amounts of data.

scholarly journals The Impact of Teachers Unions on Teachers’ Well-Being Under Various Legal Institutions: Evidence From District–Teacher Matched Data

AERA Open ◽  
2019 ◽  
Vol 5 (3) ◽  
pp. 233285841986729 ◽  
Author(s):  
Eunice S. Han
Keyword(s):  
Multilevel Models ◽  
Causal Effect ◽  
Well Being ◽  
Legal Environment ◽  
Legal Institutions ◽  
Data Set ◽  
Teachers Unions ◽  
The Difference ◽  
Matched Data ◽  
The Impact

This article examines how teachers unions affect teachers’ well-being under various legal institutions. Using a district–teacher matched data set, this study identifies the union effects by three approaches. First, I contrast teacher outcomes across different state laws toward unions. Second, I compare the union–nonunion differentials within the same legal environment, using multilevel models and propensity score matching. Finally, unexpected legal changes restricting the collective bargaining of teachers in four states form a natural experiment, allowing me to use the difference-in-difference estimation to identify the causal effect of weakening unionism on teacher outcomes. I find that (a) many teachers join unions even when bargaining is rarely or never available, and meet-and-confer or union membership rate affects teachers’ lives in the absence of a bargaining contract; (b) how unions influence teacher outcomes vary greatly by different legal environment; and (c) the changes in public policy limiting teachers’ bargaining rights significantly decrease teacher compensation.

Sign in / Sign up

Export Citation Format

Share Document