SlideShare une entreprise Scribd logo
1  sur  23
Quantum Generalized
Linear Models: A Proof
of Concept
Uchenna Chukwu and Colleen M. Farrelly, Quantopo LLC
Big Picture Summary
Overview and Implications
 Generalized linear models are the simplest instance of link-based statistical
models, which are based on the underlying geometry of an outcome’s underlying
probability distribution (typically from the exponential family).
 Machine learning algorithms provide alternative ways to minimize a model’s sum
of square error (error between predicted values and actual values of a test set).
 However, some deep results regarding the exponential family’s relation to affine
connections in differential geometry provide a possible alternative to link
functions:
1. Algorithms that either continuously deform the outcome distribution from known results
2. Algorithms that superpose all possible distributions and collapse to fit a dataset
 Leveraging the fact that some quantum computer gates, such as the non-Gaussian
transformation gate, essentially perform (1) natively and in a computationally-efficient
way!
 This project provides a proof-of-concept for leveraging specific hardware gates to
solve the affine connection problem, with benchmarking at state-of-the-art levels.
 Results can be extended to many other, more complicated statistical models, such
as generalized estimating equations, hierarchical regression models, and even
homotopy-continuation problems.
Generalized Linear Model
Background and Business Usage
An Introduction to Tweedie Models
Generalized Linear Models
 Generalized linear modeling (GLM) as extension of linear regression to
outcomes with probability distributions that are not Gaussian
 Includes binomial outcomes, Poisson outcomes, gamma outcomes, and many more
 Link functions to transform distribution of these outcomes to a normal distribution
to fit a linear model
 𝐸 𝑌 = µ = 𝑔−1
𝑿𝛽
 Var 𝑌 = 𝑉𝑎𝑟(µ) = 𝑉𝑎𝑟(𝑔−1
𝑿𝛽 )
 Where Y is a vector of outcome values, µ is the mean of Y, X is the matrix of predictor
values, g is a link function (such as the log function), and β is a vector of predictor
weights in the regression equation.
 Many statistical extensions:
 Generalized estimating equations (longitudinal data modeling)
 Generalized linear mixed models (longitudinal data with random effects)
 Generalized additive models (in which the predictor vectors can be transformed within
the model)
 Cox regression and Weibull-based regression (survival data modeling)
 Very high computational cost for many of these extensions
Important Applications of GLMs
 Ubiquitous in part failure modeling, medical research, actuarial science, and many other
problems
 Example problems:
 Modeling likelihood of insurance claims and expected payout (worldwide, a $5 trillion industry)
 Understanding risk behavior in medical research (daily heroin usage, sexual partners within prior month…)
 Modeling expected failure rates and associated conditions for airplane parts or machine parts within a
manufacturing plant (~$4 trillion industry in the USA alone)
 Modeling expected natural disaster impacts and precipitating factors related to impact extent
 Many supervised learning algorithms are extensions of generalized linear models and have link
functions built into the algorithm to model different outcome distributions
 Boosted regression, Morse-Smale regression, dgLARS, Bayesian model averaging…
 Optimization algorithm to find minimum sum of square error differ among machine learning methods and
with respect to GLMs, which use a least square error algorithm
 Methods like deep learning and classical neural networks attempt to solve this problem in a general way
through a series of general mappings leading to a potentially novel link function
 Exploiting the geometric relationships between distributions through a superposition of states collapsed
to the “ideal” link would present an optimal solution to the problem
 Tweedie regression as a general framework that handles many distributions in the exponential
family and the problem of overdispersion of model/outcome variance
 Very nice geometric properties
 Connected to many common exponential family distributions
Details of Tweedie Regression
 Many common distributions of the exponential family converge to Tweedie
distributions and can be formulated through Tweedie distributions, formally
defined as:
 𝐸 𝑌 = µ
 𝑉𝑎𝑟 𝑌 = 𝜑µ 𝜉
 where 𝜑 is the dispersion parameter, and 𝜉is the Tweedie parameter (or shape parameter)
 Tweedie distributions themselves enjoy a variety of useful properties:
 Reproductive properties that allow distributions to be added together to form new
distributions that are themselves Tweedie
 Varying Tweedie parameter and dispersion parameter to recover many exponential
family distributions used in GLMs:
 Tweedie parameter of 0 for normal distribution
 Tweedie parameter of 1 for Poisson distribution
 Tweedie parameter of 2 for gamma distribution
 Dispersion parameter for 0-inflated models and outliers, similar to negative binomial
regression models
The Problem of Overdispersion in
Tweedie Models
 Well-known statistical problem involving dispersion parameters, which relate
to the variance of a outcome
 Many GLMs and their machine learning extensions struggle on problems of
overdispersion
 Simulations show this behavior, particularly as dispersion parameter increases
substantially (values of 4+)
 Empirical datasets with 0-inflation and long tails
 Recent paper exploring bagged KNN models
 Demonstrates problem in simulations
 Demonstrates with open-source datasets, such as UCI’s Forest Fire dataset
 Models that work well, such as the KNN ensemble with varying k parameters, tend
to take a long time to compute
Common Tweedie Models
Family
Distribution
Dispersion
(extra 0’s
and tail
fatness)
Power (variance
proportional to
mean: 1/Power)
Normal 1 0
Poisson 1 1
Compound
Poisson
1 >1 and <2
Gamma 1 2
Inverse-Gaussian 1 3
Stable 1 >2 (Extreme >3)
Negative
Binomial
>1 1
Underdispersion
Poisson
<1 1
Unique Tweedie >=1 >=0
Connection of GLMs to Differential
Geometry
Motivation for Implementation on Xanadu System
Differential Geometry and the
Exponential Family
 Possible to formulate exponential
family distributions and their
parameterizations to form a series of
curves on a 2-dimensional surface
 Each curve defined by 2 points at
either end of the probability function,
0 and 1, connected by a line that
follows a shortest path following
parameterization of the distribution,
called a geodesic
 Because the exponential family can be
generalized into Tweedie distributions
through continuous transformations,
the geodesic connecting 0 and 1 can
flow across distributions defining the
2-dimensional surface in a continuous
manner (much like homotopy
continuation methods).
 This is an affine connection, and the
morphing of the line as it passes
parameters transforms one
distribution to another.
Consequences of Exponential Family
Geometry
 Analytically-
derived
results/equation
for one
distribution
morphed to fit
another
distribution
through
continuous
transformations!
 Limit theorems
derived by
continuous
deformations of
either moment
generating
functions or
characteristic
functions
Xanadu Technology and Suitability to
GLMs
 Xanadu’s qumode formulation makes ideal for implementing quantum GLMs
 Ability to perform linear algebra operations on physical data representations
 GLMs and their extensions all based on simple matrix operations
 𝑀𝑒𝑎𝑛 𝑌 = 𝑔−1
𝑿𝛽 + 𝜀
 Matrix multiplication and addition for the linear model 𝑿𝛽 + 𝜀 coupled with a continuous
transformation of the model results to fit the outcome distribution
 Non-Gaussian transformation gate provides perfect avenue to perform the affine
transformation related to the outcome distribution without a need to specific a
link function to approximate the geometry
 Should be able to approximate any continuous outcome’s distribution, creating potential
new “link functions” through this gate through affine transformation of the wavefunctions
representing the data
 Removes the need for approximations by easy-to-compute link transformations
 In theory, should approximate any continuous distribution, including ones that aren’t
included in common statistical packages implementing GLMs and their
longitudinal/survival data extensions
 Thus, Xanadu’s system provides a general solution to the linear regression equation with
many potential extensions to more sophisticated regression models!
Methods and Results on Example
Cases
Simulated overdispersion dataset and UCI Forest Fire dataset
Methodology
 Simulation
 Similar to simulations used in the KNN ensemble paper
 1000 observations with a 70/30 test/train split
 Tweedie outcome related to 3 predictors (1 interaction term, 1 main effect) with added noise
 Tweedie parameter=1, dispersion parameter=8
 1 noise variable added
 Empirical dataset
 UCI Repository’s Forest Fire dataset
 Notoriously difficult to beat the mean model with machine learning algorithms
 12 predictors (2 spatial coordinates of location, month, day, FFMC index, DMC index, DC index, ISI index,
temperature, relative humidity, wind, and rain) and 517 observations
 t-SNE was used to reduce the dimensionality of the predictor set to 4 components so as to make it compatible
with Xanadu’s capabilities.
 70/30 test/train split
 Comparison methods
 Boosted regression
 Random forest (tree-based bagged ensemble)
 DGLARS (tangent-space-based least angle regression model)
 BART (Bayesian-based tree ensemble)
 HLASSO (homotopy-based LASSO model)
 Poisson regression (GLM without any modifications)
Data Preprossessing
 Dimensionality reduction through t-SNE to create a set of 4 predictors and 1
outcome, such that predictors are uncorrelated when entered into models.
 Easier for systems to calculate with fewer variables.
 Decorrelation helps most regression methods, including linear models and tree
models.
 Other dimensionality reduction methods are possible, including the introduction of
factors from factor analytic models or combinations of linear/nonlinear,
global/local dimensionality reduction algorithms.
 Scaling of outcome to a scale of -3 to 3, such that the Xanadu simulation can
effectively model and process the data in qumodes.
 Slight warping of the most extreme values, but these are generally less than 5
observations per dataset.
 Other types of scaling might be useful to explore.
Qumodes Circuit Details
 GLMs can be embedded within Xanadu’s qumode quantum computer simulation
software (and qumode computer) with a singular value decomposition of the
𝛽 coefficient in the formulation:
 𝑀𝑒𝑎𝑛 𝑌 = 𝑔−1
𝑋𝛽
 This translates to 𝛽 = 𝑂1Σ𝑂2, which can be modeled through a series of quantum
circuit gates:
 Multiplication of X and an orthogonal matrix:
 | 𝑂1 𝑋 ≅ 𝑈1| 𝑋 , which corresponds to a linear interferometer gate (𝑈1) acting on X
 Multiplication of that result by a diagonal matrix:
 |Σ𝑂1 𝑋 ∝ 𝑆 𝑟 | 𝑂1 𝑋 , which corresponds to a squeezing gate that acts on a single qumode
 Multiplication of X and an orthogonal matrix:
 | 𝑂2Σ𝑂1 𝑋 ≅ 𝑈2|Σ 𝑂1 𝑋 , which corresponds to a linear interferometer gate (𝑈2) acting on the result
 Multiplication by a nonlinear function on this result:
 |𝑔−1
( 𝑂2Σ𝑂1 𝑋) ≅ Φ|𝑂2Σ𝑂1 𝑋 , which corresponds to the non-Gaussian gate acting on the result
 This gives a final result of gates acting upon the dataset as:
 Φ ∗ 𝒰2 ∗ 𝒮 ∗ 𝒰1| 𝑋 ∝ | 𝑔−1
𝑋𝛽
Qumodes Parameter Settings
 The algorithm simulation was created through Strawberry Fields.
 The deep learning framework already existed.
 Hidden layers and bias terms were removed to collapse to a generalized linear model
framework.
 The loss function optimized was mean square error, which corresponds to the loss
functions specified in the comparison algorithms.
 Qumode cut-off dimension was set to 10.
 Optimization via least squares was not available, so gradient descent was used with a
learning rate of 0.1 over 80 iterations.
 This gave a qumodes implementation of a quantum generalized linear model with a
boosting feel to it.
 Because the quantum computing component is inherently probabilitistic,
algorithms were run on the same training and test set multiple times to
average out quantum effects.
Results: Simulation of Overdispersion
Algorithm Scaled Model MSE
Random Forest 0.80
BART 0.78
Boosted Regression 0.78
DGLARS 0.81
HLASSO 0.81
GLM 0.81
QGLM 0.82
Mean 0.85
QGLMs yield slightly worse
prediction on the simulated
dataset. However, their
performance is not far off from
state-of-the art algorithms, and
some random error is expected
from the quantum machinery.
Results: Forest Fire Dataset
Algorithm Scaled Model MSE
Random Forest 0.125
BART 0.125
Boosted Regression 0.119
DGLARS 0.114
HLASSO 0.120
GLM 0.119
QGLM 0.106
Mean 0.115
QGLMs emerge as the best-performing algorithm on a difficult, real-
world dataset (Forest Fire dataset in the UCI repository). QGLMs
provide ~10% gain over the next best algorithm on this dataset. This
suggests that they work well on real data and difficult problems.
Conclusions
 This suggests that the qumodes formulation with its unique operators can
eliminate the need for link functions within linear models by exploiting the
geometry of the models and still give good prediction.
 Better than state-of-the-art prediction for a difficult Tweedie regression dataset
(UCI Forest Fire)
 Around state-of-the-art prediction for a simulated dataset
 This has the potential to bring statistical modeling into quantum computing,
by leveraging the underlying geometry and the connection between model
geometry and the geometry of quantum physics.
 Generalized estimating equations/generalized linear mixed models
 Structural equation models/hierarchical regression models
 Also a potential avenue through which to implement the homotopy continuation
method common in dynamic systems research and some machine learning models
(such as homotopy-based LASSO), which take a known problem’s solution and
continuously deform it to fit the problem of interest.
 Currently a computational challenge
 Limited to small datasets
References
 Amari, S. I. (1997). Information geometry. Contemporary Mathematics, 203, 81-96.
 Bulmer, M. G. (1974). On fitting the Poisson lognormal distribution to species-abundance data. Biometrics,
101-110.
 Buscemi, F. (2012). Comparison of quantum statistical models: equivalent conditions for sufficiency.
Communications in Mathematical Physics, 310(3), 625-647.
 Cortez, P., & Morais, A. D. J. R. (2007). A data mining approach to predict forest fires using meteorological
data.
 De Jong, P., & Heller, G. Z. (2008). Generalized linear models for insurance data (Vol. 10). Cambridge:
Cambridge University Press.
 Farrelly, C. M. (2017). KNN Ensembles for Tweedie Regression: The Power of Multiscale Neighborhoods.
arXiv preprint arXiv:1708.02122.
 Farrelly, C. M. (2017). Topology and Geometry in Machine Learning for Logistic Regression.
 Fehm, L., Beesdo, K., Jacobi, F., & Fiedler, A. (2008). Social anxiety disorder above and below the
diagnostic threshold: prevalence, comorbidity and impairment in the general population. Social psychiatry
and psychiatric epidemiology, 43(4), 257-265.
 Fergusson, D. M., Boden, J. M., & Horwood, L. J. (2006). Cannabis use and other illicit drug use: testing
the cannabis gateway hypothesis. Addiction, 101(4), 556-569.
 Frees, E. W., Lee, G., & Yang, L. (2016). Multivariate frequency-severity regression models in insurance.
Risks, 4(1), 4.
 Gardner, W., Mulvey, E. P., & Shaw, E. C. (1995). Regression analyses of counts and rates: Poisson,
overdispersed Poisson, and negative binomial models. Psychological bulletin, 118(3), 392.
 Herings, R., & Erkens, J. A. (2003). Increased suicide attempt rate among patients interrupting use of
atypical antipsychotics. Pharmacoepidemiology and drug safety, 12(5), 423-424.
References
 Jorgensen, B. (1997). The theory of dispersion models. CRC Press.
 Jørgensen, B., Goegebeur, Y., & Martínez, J. R. (2010). Dispersion models for extremes. Extremes, 13(4), 399-437.
 Killoran, N., Bromley, T. R., Arrazola, J. M., Schuld, M., Quesada, N., & Lloyd, S. (2018). Continuous-variable
quantum neural networks. arXiv preprint arXiv:1806.06871.
 Killoran, N., Izaac, J., Quesada, N., Bergholm, V., Amy, M., & Weedbrook, C. (2018). Strawberry Fields: A Software
Platform for Photonic Quantum Computing. arXiv preprint arXiv:1804.03159.
 Luitel, B. N. (2016). Prediction of North Atlantic tropical cyclone activity and rainfall (Doctoral dissertation, The
University of Iowa).
 Marriott, P. (1990). Applications of differential geometry to statistics (Doctoral dissertation, University of
Warwick).
 Mills, K. L., Teesson, M., Ross, J., & Peters, L. (2006). Trauma, PTSD, and substance use disorders: findings from
the Australian National Survey of Mental Health and Well-Being. American Journal of Psychiatry, 163(4), 652-658.
 Nielsen, F., & Garcia, V. (2009). Statistical exponential families: A digest with flash cards. arXiv preprint
arXiv:0911.4863.
 Osborne, M. R., Presnell, B., & Turlach, B. A. (2000). A new approach to variable selection in least squares
problems. IMA journal of numerical analysis, 20(3), 389-403.
 Pistone, G., & Rogantin, M. P. (1999). The exponential statistical manifold: mean parameters, orthogonality and
space transformations. Bernoulli, 5(4), 721-760.
 Sawalha, Z., & Sayed, T. (2006). Traffic accident modeling: some statistical issues. Canadian Journal of Civil
Engineering, 33(9), 1115-1124.
 Tweedie, M. C. K. (1984). An index which distinguishes between some important exponential families. In Statistics:
Applications and new directions: Proc. Indian statistical institute golden Jubilee International conference (Vol.
579, p. 6o4).

Contenu connexe

Tendances

TYBSC IT PGIS Unit II Chapter I Data Management and Processing Systems
TYBSC IT PGIS Unit II Chapter I Data Management and Processing SystemsTYBSC IT PGIS Unit II Chapter I Data Management and Processing Systems
TYBSC IT PGIS Unit II Chapter I Data Management and Processing SystemsArti Parab Academics
 
TYBSC IT PGIS Unit III Chapter II Data Entry and Preparation
TYBSC IT PGIS Unit III Chapter II Data Entry and PreparationTYBSC IT PGIS Unit III Chapter II Data Entry and Preparation
TYBSC IT PGIS Unit III Chapter II Data Entry and PreparationArti Parab Academics
 
Vector and Raster Data data model
Vector and Raster Data data modelVector and Raster Data data model
Vector and Raster Data data modelCalcutta University
 
Introduction to Computational Intelligent
Introduction to Computational IntelligentIntroduction to Computational Intelligent
Introduction to Computational IntelligentKent State University
 
Gis (geographic information system)
Gis (geographic information system)Gis (geographic information system)
Gis (geographic information system)Saad Bare
 
Deep learning health care
Deep learning health care  Deep learning health care
Deep learning health care Meenakshi Sood
 
Remote Sensing Imagery & Artificial Intelligence
Remote Sensing Imagery & Artificial IntelligenceRemote Sensing Imagery & Artificial Intelligence
Remote Sensing Imagery & Artificial IntelligenceEsri Ireland
 
Electronic health records and machine learning
Electronic health records and machine learningElectronic health records and machine learning
Electronic health records and machine learningEman Abdelrazik
 
Overview of big data in cloud computing
Overview of big data in cloud computingOverview of big data in cloud computing
Overview of big data in cloud computingViet-Trung TRAN
 
Geographical information system unit 1
Geographical information  system unit 1Geographical information  system unit 1
Geographical information system unit 1WE-IT TUTORIALS
 
Spatial Analysis and Geomatics
Spatial Analysis and GeomaticsSpatial Analysis and Geomatics
Spatial Analysis and GeomaticsRich Heimann
 
Geographical information systems
Geographical information systemsGeographical information systems
Geographical information systemsGift Musanza
 
Internet of medical things (IOMT)
Internet of medical things (IOMT)Internet of medical things (IOMT)
Internet of medical things (IOMT)K Raman Sethuraman
 
FUTURE OF DATA SCIENCE IN INDIA
FUTURE OF DATA SCIENCE IN INDIAFUTURE OF DATA SCIENCE IN INDIA
FUTURE OF DATA SCIENCE IN INDIAkaranramani4
 
IMED 2018: An intro to Remote Sensing and Machine Learning
IMED 2018: An intro to Remote Sensing and Machine LearningIMED 2018: An intro to Remote Sensing and Machine Learning
IMED 2018: An intro to Remote Sensing and Machine LearningLouisa Diggs
 

Tendances (20)

TYBSC IT PGIS Unit II Chapter I Data Management and Processing Systems
TYBSC IT PGIS Unit II Chapter I Data Management and Processing SystemsTYBSC IT PGIS Unit II Chapter I Data Management and Processing Systems
TYBSC IT PGIS Unit II Chapter I Data Management and Processing Systems
 
Network analysis in gis , part 2 connectivity rules
Network analysis in gis , part 2 connectivity rulesNetwork analysis in gis , part 2 connectivity rules
Network analysis in gis , part 2 connectivity rules
 
TYBSC IT PGIS Unit III Chapter II Data Entry and Preparation
TYBSC IT PGIS Unit III Chapter II Data Entry and PreparationTYBSC IT PGIS Unit III Chapter II Data Entry and Preparation
TYBSC IT PGIS Unit III Chapter II Data Entry and Preparation
 
Spatial Databases
Spatial DatabasesSpatial Databases
Spatial Databases
 
What is GIS?
What is GIS?What is GIS?
What is GIS?
 
Vector and Raster Data data model
Vector and Raster Data data modelVector and Raster Data data model
Vector and Raster Data data model
 
Introduction to Computational Intelligent
Introduction to Computational IntelligentIntroduction to Computational Intelligent
Introduction to Computational Intelligent
 
Gps
GpsGps
Gps
 
Gis (geographic information system)
Gis (geographic information system)Gis (geographic information system)
Gis (geographic information system)
 
Deep learning health care
Deep learning health care  Deep learning health care
Deep learning health care
 
Remote Sensing Imagery & Artificial Intelligence
Remote Sensing Imagery & Artificial IntelligenceRemote Sensing Imagery & Artificial Intelligence
Remote Sensing Imagery & Artificial Intelligence
 
Electronic health records and machine learning
Electronic health records and machine learningElectronic health records and machine learning
Electronic health records and machine learning
 
Overview of big data in cloud computing
Overview of big data in cloud computingOverview of big data in cloud computing
Overview of big data in cloud computing
 
Geographical information system unit 1
Geographical information  system unit 1Geographical information  system unit 1
Geographical information system unit 1
 
Spatial Analysis and Geomatics
Spatial Analysis and GeomaticsSpatial Analysis and Geomatics
Spatial Analysis and Geomatics
 
Geographical information systems
Geographical information systemsGeographical information systems
Geographical information systems
 
AI in Practice for Healthcare
AI in Practice for Healthcare AI in Practice for Healthcare
AI in Practice for Healthcare
 
Internet of medical things (IOMT)
Internet of medical things (IOMT)Internet of medical things (IOMT)
Internet of medical things (IOMT)
 
FUTURE OF DATA SCIENCE IN INDIA
FUTURE OF DATA SCIENCE IN INDIAFUTURE OF DATA SCIENCE IN INDIA
FUTURE OF DATA SCIENCE IN INDIA
 
IMED 2018: An intro to Remote Sensing and Machine Learning
IMED 2018: An intro to Remote Sensing and Machine LearningIMED 2018: An intro to Remote Sensing and Machine Learning
IMED 2018: An intro to Remote Sensing and Machine Learning
 

Similaire à Quantum generalized linear models

PyData Miami 2019, Quantum Generalized Linear Models
PyData Miami 2019, Quantum Generalized Linear ModelsPyData Miami 2019, Quantum Generalized Linear Models
PyData Miami 2019, Quantum Generalized Linear ModelsColleen Farrelly
 
Jgrass-NewAge: Kriging component
Jgrass-NewAge: Kriging componentJgrass-NewAge: Kriging component
Jgrass-NewAge: Kriging componentNiccolò Tubini
 
Data Science Meetup: DGLARS and Homotopy LASSO for Regression Models
Data Science Meetup: DGLARS and Homotopy LASSO for Regression ModelsData Science Meetup: DGLARS and Homotopy LASSO for Regression Models
Data Science Meetup: DGLARS and Homotopy LASSO for Regression ModelsColleen Farrelly
 
ProbabilisticModeling20080411
ProbabilisticModeling20080411ProbabilisticModeling20080411
ProbabilisticModeling20080411Clay Stanek
 
Guide for building GLMS
Guide for building GLMSGuide for building GLMS
Guide for building GLMSAli T. Lotia
 
2007 santiago marchi_cobem_2007
2007 santiago marchi_cobem_20072007 santiago marchi_cobem_2007
2007 santiago marchi_cobem_2007CosmoSantiago
 
Multiple Linear Regression Model with Two Parameter Doubly Truncated New Symm...
Multiple Linear Regression Model with Two Parameter Doubly Truncated New Symm...Multiple Linear Regression Model with Two Parameter Doubly Truncated New Symm...
Multiple Linear Regression Model with Two Parameter Doubly Truncated New Symm...theijes
 
Poor man's missing value imputation
Poor man's missing value imputationPoor man's missing value imputation
Poor man's missing value imputationLeonardo Auslender
 
Abrigo and love_2015_
Abrigo and love_2015_Abrigo and love_2015_
Abrigo and love_2015_Murtaza Khan
 
A General Purpose Exact Solution Method For Mixed Integer Concave Minimizatio...
A General Purpose Exact Solution Method For Mixed Integer Concave Minimizatio...A General Purpose Exact Solution Method For Mixed Integer Concave Minimizatio...
A General Purpose Exact Solution Method For Mixed Integer Concave Minimizatio...Martha Brown
 
ders 6 Panel data analysis.pptx
ders 6 Panel data analysis.pptxders 6 Panel data analysis.pptx
ders 6 Panel data analysis.pptxErgin Akalpler
 
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...Modeling Approaches and Studies of the Impact of Distributed Energy Resources...
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...Power System Operation
 
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...Modeling Approaches and Studies of the Impact of Distributed Energy Resources...
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...Power System Operation
 
Evaluating competing predictive distributions
Evaluating competing predictive distributionsEvaluating competing predictive distributions
Evaluating competing predictive distributionsAndreas Collett
 
Panel data_25412547859_andbcbgajkje852.ppt
Panel data_25412547859_andbcbgajkje852.pptPanel data_25412547859_andbcbgajkje852.ppt
Panel data_25412547859_andbcbgajkje852.pptHinhMo
 

Similaire à Quantum generalized linear models (20)

PyData Miami 2019, Quantum Generalized Linear Models
PyData Miami 2019, Quantum Generalized Linear ModelsPyData Miami 2019, Quantum Generalized Linear Models
PyData Miami 2019, Quantum Generalized Linear Models
 
beven 2001.pdf
beven 2001.pdfbeven 2001.pdf
beven 2001.pdf
 
Jgrass-NewAge: Kriging component
Jgrass-NewAge: Kriging componentJgrass-NewAge: Kriging component
Jgrass-NewAge: Kriging component
 
Data Science Meetup: DGLARS and Homotopy LASSO for Regression Models
Data Science Meetup: DGLARS and Homotopy LASSO for Regression ModelsData Science Meetup: DGLARS and Homotopy LASSO for Regression Models
Data Science Meetup: DGLARS and Homotopy LASSO for Regression Models
 
ProbabilisticModeling20080411
ProbabilisticModeling20080411ProbabilisticModeling20080411
ProbabilisticModeling20080411
 
Guide for building GLMS
Guide for building GLMSGuide for building GLMS
Guide for building GLMS
 
2007 santiago marchi_cobem_2007
2007 santiago marchi_cobem_20072007 santiago marchi_cobem_2007
2007 santiago marchi_cobem_2007
 
Multiple Linear Regression Model with Two Parameter Doubly Truncated New Symm...
Multiple Linear Regression Model with Two Parameter Doubly Truncated New Symm...Multiple Linear Regression Model with Two Parameter Doubly Truncated New Symm...
Multiple Linear Regression Model with Two Parameter Doubly Truncated New Symm...
 
TO_EDIT
TO_EDITTO_EDIT
TO_EDIT
 
Poor man's missing value imputation
Poor man's missing value imputationPoor man's missing value imputation
Poor man's missing value imputation
 
Abrigo and love_2015_
Abrigo and love_2015_Abrigo and love_2015_
Abrigo and love_2015_
 
Cohen
CohenCohen
Cohen
 
A General Purpose Exact Solution Method For Mixed Integer Concave Minimizatio...
A General Purpose Exact Solution Method For Mixed Integer Concave Minimizatio...A General Purpose Exact Solution Method For Mixed Integer Concave Minimizatio...
A General Purpose Exact Solution Method For Mixed Integer Concave Minimizatio...
 
ders 6 Panel data analysis.pptx
ders 6 Panel data analysis.pptxders 6 Panel data analysis.pptx
ders 6 Panel data analysis.pptx
 
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...Modeling Approaches and Studies of the Impact of Distributed Energy Resources...
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...
 
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...Modeling Approaches and Studies of the Impact of Distributed Energy Resources...
Modeling Approaches and Studies of the Impact of Distributed Energy Resources...
 
Evaluating competing predictive distributions
Evaluating competing predictive distributionsEvaluating competing predictive distributions
Evaluating competing predictive distributions
 
panel data.ppt
panel data.pptpanel data.ppt
panel data.ppt
 
Panel data_25412547859_andbcbgajkje852.ppt
Panel data_25412547859_andbcbgajkje852.pptPanel data_25412547859_andbcbgajkje852.ppt
Panel data_25412547859_andbcbgajkje852.ppt
 
gamdependence_revision1
gamdependence_revision1gamdependence_revision1
gamdependence_revision1
 

Plus de Colleen Farrelly

Generative AI for Social Good at Open Data Science East 2024
Generative AI for Social Good at Open Data Science East 2024Generative AI for Social Good at Open Data Science East 2024
Generative AI for Social Good at Open Data Science East 2024Colleen Farrelly
 
Hands-On Network Science, PyData Global 2023
Hands-On Network Science, PyData Global 2023Hands-On Network Science, PyData Global 2023
Hands-On Network Science, PyData Global 2023Colleen Farrelly
 
Modeling Climate Change.pptx
Modeling Climate Change.pptxModeling Climate Change.pptx
Modeling Climate Change.pptxColleen Farrelly
 
Natural Language Processing for Beginners.pptx
Natural Language Processing for Beginners.pptxNatural Language Processing for Beginners.pptx
Natural Language Processing for Beginners.pptxColleen Farrelly
 
The Shape of Data--ODSC.pptx
The Shape of Data--ODSC.pptxThe Shape of Data--ODSC.pptx
The Shape of Data--ODSC.pptxColleen Farrelly
 
Generative AI, WiDS 2023.pptx
Generative AI, WiDS 2023.pptxGenerative AI, WiDS 2023.pptx
Generative AI, WiDS 2023.pptxColleen Farrelly
 
Emerging Technologies for Public Health in Remote Locations.pptx
Emerging Technologies for Public Health in Remote Locations.pptxEmerging Technologies for Public Health in Remote Locations.pptx
Emerging Technologies for Public Health in Remote Locations.pptxColleen Farrelly
 
Applications of Forman-Ricci Curvature.pptx
Applications of Forman-Ricci Curvature.pptxApplications of Forman-Ricci Curvature.pptx
Applications of Forman-Ricci Curvature.pptxColleen Farrelly
 
Geometry for Social Good.pptx
Geometry for Social Good.pptxGeometry for Social Good.pptx
Geometry for Social Good.pptxColleen Farrelly
 
Topology for Time Series.pptx
Topology for Time Series.pptxTopology for Time Series.pptx
Topology for Time Series.pptxColleen Farrelly
 
Time Series Applications AMLD.pptx
Time Series Applications AMLD.pptxTime Series Applications AMLD.pptx
Time Series Applications AMLD.pptxColleen Farrelly
 
An introduction to quantum machine learning.pptx
An introduction to quantum machine learning.pptxAn introduction to quantum machine learning.pptx
An introduction to quantum machine learning.pptxColleen Farrelly
 
An introduction to time series data with R.pptx
An introduction to time series data with R.pptxAn introduction to time series data with R.pptx
An introduction to time series data with R.pptxColleen Farrelly
 
NLP: Challenges and Opportunities in Underserved Areas
NLP: Challenges and Opportunities in Underserved AreasNLP: Challenges and Opportunities in Underserved Areas
NLP: Challenges and Opportunities in Underserved AreasColleen Farrelly
 
Geometry, Data, and One Path Into Data Science.pptx
Geometry, Data, and One Path Into Data Science.pptxGeometry, Data, and One Path Into Data Science.pptx
Geometry, Data, and One Path Into Data Science.pptxColleen Farrelly
 
Topological Data Analysis.pptx
Topological Data Analysis.pptxTopological Data Analysis.pptx
Topological Data Analysis.pptxColleen Farrelly
 
Transforming Text Data to Matrix Data via Embeddings.pptx
Transforming Text Data to Matrix Data via Embeddings.pptxTransforming Text Data to Matrix Data via Embeddings.pptx
Transforming Text Data to Matrix Data via Embeddings.pptxColleen Farrelly
 
Natural Language Processing in the Wild.pptx
Natural Language Processing in the Wild.pptxNatural Language Processing in the Wild.pptx
Natural Language Processing in the Wild.pptxColleen Farrelly
 
SAS Global 2021 Introduction to Natural Language Processing
SAS Global 2021 Introduction to Natural Language Processing SAS Global 2021 Introduction to Natural Language Processing
SAS Global 2021 Introduction to Natural Language Processing Colleen Farrelly
 
2021 American Mathematical Society Data Science Talk
2021 American Mathematical Society Data Science Talk2021 American Mathematical Society Data Science Talk
2021 American Mathematical Society Data Science TalkColleen Farrelly
 

Plus de Colleen Farrelly (20)

Generative AI for Social Good at Open Data Science East 2024
Generative AI for Social Good at Open Data Science East 2024Generative AI for Social Good at Open Data Science East 2024
Generative AI for Social Good at Open Data Science East 2024
 
Hands-On Network Science, PyData Global 2023
Hands-On Network Science, PyData Global 2023Hands-On Network Science, PyData Global 2023
Hands-On Network Science, PyData Global 2023
 
Modeling Climate Change.pptx
Modeling Climate Change.pptxModeling Climate Change.pptx
Modeling Climate Change.pptx
 
Natural Language Processing for Beginners.pptx
Natural Language Processing for Beginners.pptxNatural Language Processing for Beginners.pptx
Natural Language Processing for Beginners.pptx
 
The Shape of Data--ODSC.pptx
The Shape of Data--ODSC.pptxThe Shape of Data--ODSC.pptx
The Shape of Data--ODSC.pptx
 
Generative AI, WiDS 2023.pptx
Generative AI, WiDS 2023.pptxGenerative AI, WiDS 2023.pptx
Generative AI, WiDS 2023.pptx
 
Emerging Technologies for Public Health in Remote Locations.pptx
Emerging Technologies for Public Health in Remote Locations.pptxEmerging Technologies for Public Health in Remote Locations.pptx
Emerging Technologies for Public Health in Remote Locations.pptx
 
Applications of Forman-Ricci Curvature.pptx
Applications of Forman-Ricci Curvature.pptxApplications of Forman-Ricci Curvature.pptx
Applications of Forman-Ricci Curvature.pptx
 
Geometry for Social Good.pptx
Geometry for Social Good.pptxGeometry for Social Good.pptx
Geometry for Social Good.pptx
 
Topology for Time Series.pptx
Topology for Time Series.pptxTopology for Time Series.pptx
Topology for Time Series.pptx
 
Time Series Applications AMLD.pptx
Time Series Applications AMLD.pptxTime Series Applications AMLD.pptx
Time Series Applications AMLD.pptx
 
An introduction to quantum machine learning.pptx
An introduction to quantum machine learning.pptxAn introduction to quantum machine learning.pptx
An introduction to quantum machine learning.pptx
 
An introduction to time series data with R.pptx
An introduction to time series data with R.pptxAn introduction to time series data with R.pptx
An introduction to time series data with R.pptx
 
NLP: Challenges and Opportunities in Underserved Areas
NLP: Challenges and Opportunities in Underserved AreasNLP: Challenges and Opportunities in Underserved Areas
NLP: Challenges and Opportunities in Underserved Areas
 
Geometry, Data, and One Path Into Data Science.pptx
Geometry, Data, and One Path Into Data Science.pptxGeometry, Data, and One Path Into Data Science.pptx
Geometry, Data, and One Path Into Data Science.pptx
 
Topological Data Analysis.pptx
Topological Data Analysis.pptxTopological Data Analysis.pptx
Topological Data Analysis.pptx
 
Transforming Text Data to Matrix Data via Embeddings.pptx
Transforming Text Data to Matrix Data via Embeddings.pptxTransforming Text Data to Matrix Data via Embeddings.pptx
Transforming Text Data to Matrix Data via Embeddings.pptx
 
Natural Language Processing in the Wild.pptx
Natural Language Processing in the Wild.pptxNatural Language Processing in the Wild.pptx
Natural Language Processing in the Wild.pptx
 
SAS Global 2021 Introduction to Natural Language Processing
SAS Global 2021 Introduction to Natural Language Processing SAS Global 2021 Introduction to Natural Language Processing
SAS Global 2021 Introduction to Natural Language Processing
 
2021 American Mathematical Society Data Science Talk
2021 American Mathematical Society Data Science Talk2021 American Mathematical Society Data Science Talk
2021 American Mathematical Society Data Science Talk
 

Dernier

Escorts Service Kumaraswamy Layout ☎ 7737669865☎ Book Your One night Stand (B...
Escorts Service Kumaraswamy Layout ☎ 7737669865☎ Book Your One night Stand (B...Escorts Service Kumaraswamy Layout ☎ 7737669865☎ Book Your One night Stand (B...
Escorts Service Kumaraswamy Layout ☎ 7737669865☎ Book Your One night Stand (B...amitlee9823
 
Cheap Rate Call girls Sarita Vihar Delhi 9205541914 shot 1500 night
Cheap Rate Call girls Sarita Vihar Delhi 9205541914 shot 1500 nightCheap Rate Call girls Sarita Vihar Delhi 9205541914 shot 1500 night
Cheap Rate Call girls Sarita Vihar Delhi 9205541914 shot 1500 nightDelhi Call girls
 
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...ZurliaSoop
 
Capstone Project on IBM Data Analytics Program
Capstone Project on IBM Data Analytics ProgramCapstone Project on IBM Data Analytics Program
Capstone Project on IBM Data Analytics ProgramMoniSankarHazra
 
Call me @ 9892124323 Cheap Rate Call Girls in Vashi with Real Photo 100% Secure
Call me @ 9892124323  Cheap Rate Call Girls in Vashi with Real Photo 100% SecureCall me @ 9892124323  Cheap Rate Call Girls in Vashi with Real Photo 100% Secure
Call me @ 9892124323 Cheap Rate Call Girls in Vashi with Real Photo 100% SecurePooja Nehwal
 
Generative AI on Enterprise Cloud with NiFi and Milvus
Generative AI on Enterprise Cloud with NiFi and MilvusGenerative AI on Enterprise Cloud with NiFi and Milvus
Generative AI on Enterprise Cloud with NiFi and MilvusTimothy Spann
 
Vip Mumbai Call Girls Thane West Call On 9920725232 With Body to body massage...
Vip Mumbai Call Girls Thane West Call On 9920725232 With Body to body massage...Vip Mumbai Call Girls Thane West Call On 9920725232 With Body to body massage...
Vip Mumbai Call Girls Thane West Call On 9920725232 With Body to body massage...amitlee9823
 
FESE Capital Markets Fact Sheet 2024 Q1.pdf
FESE Capital Markets Fact Sheet 2024 Q1.pdfFESE Capital Markets Fact Sheet 2024 Q1.pdf
FESE Capital Markets Fact Sheet 2024 Q1.pdfMarinCaroMartnezBerg
 
ALSO dropshipping via API with DroFx.pptx
ALSO dropshipping via API with DroFx.pptxALSO dropshipping via API with DroFx.pptx
ALSO dropshipping via API with DroFx.pptxolyaivanovalion
 
CebaBaby dropshipping via API with DroFX.pptx
CebaBaby dropshipping via API with DroFX.pptxCebaBaby dropshipping via API with DroFX.pptx
CebaBaby dropshipping via API with DroFX.pptxolyaivanovalion
 
Invezz.com - Grow your wealth with trading signals
Invezz.com - Grow your wealth with trading signalsInvezz.com - Grow your wealth with trading signals
Invezz.com - Grow your wealth with trading signalsInvezz1
 
Discover Why Less is More in B2B Research
Discover Why Less is More in B2B ResearchDiscover Why Less is More in B2B Research
Discover Why Less is More in B2B Researchmichael115558
 
Week-01-2.ppt BBB human Computer interaction
Week-01-2.ppt BBB human Computer interactionWeek-01-2.ppt BBB human Computer interaction
Week-01-2.ppt BBB human Computer interactionfulawalesam
 
Smarteg dropshipping via API with DroFx.pptx
Smarteg dropshipping via API with DroFx.pptxSmarteg dropshipping via API with DroFx.pptx
Smarteg dropshipping via API with DroFx.pptxolyaivanovalion
 
April 2024 - Crypto Market Report's Analysis
April 2024 - Crypto Market Report's AnalysisApril 2024 - Crypto Market Report's Analysis
April 2024 - Crypto Market Report's Analysismanisha194592
 
Call Girls Begur Just Call 👗 7737669865 👗 Top Class Call Girl Service Bangalore
Call Girls Begur Just Call 👗 7737669865 👗 Top Class Call Girl Service BangaloreCall Girls Begur Just Call 👗 7737669865 👗 Top Class Call Girl Service Bangalore
Call Girls Begur Just Call 👗 7737669865 👗 Top Class Call Girl Service Bangaloreamitlee9823
 
Call Girls in Sarai Kale Khan Delhi 💯 Call Us 🔝9205541914 🔝( Delhi) Escorts S...
Call Girls in Sarai Kale Khan Delhi 💯 Call Us 🔝9205541914 🔝( Delhi) Escorts S...Call Girls in Sarai Kale Khan Delhi 💯 Call Us 🔝9205541914 🔝( Delhi) Escorts S...
Call Girls in Sarai Kale Khan Delhi 💯 Call Us 🔝9205541914 🔝( Delhi) Escorts S...Delhi Call girls
 
Junnasandra Call Girls: 🍓 7737669865 🍓 High Profile Model Escorts | Bangalore...
Junnasandra Call Girls: 🍓 7737669865 🍓 High Profile Model Escorts | Bangalore...Junnasandra Call Girls: 🍓 7737669865 🍓 High Profile Model Escorts | Bangalore...
Junnasandra Call Girls: 🍓 7737669865 🍓 High Profile Model Escorts | Bangalore...amitlee9823
 
Mg Road Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Banga...
Mg Road Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Banga...Mg Road Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Banga...
Mg Road Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Banga...amitlee9823
 

Dernier (20)

Escorts Service Kumaraswamy Layout ☎ 7737669865☎ Book Your One night Stand (B...
Escorts Service Kumaraswamy Layout ☎ 7737669865☎ Book Your One night Stand (B...Escorts Service Kumaraswamy Layout ☎ 7737669865☎ Book Your One night Stand (B...
Escorts Service Kumaraswamy Layout ☎ 7737669865☎ Book Your One night Stand (B...
 
Cheap Rate Call girls Sarita Vihar Delhi 9205541914 shot 1500 night
Cheap Rate Call girls Sarita Vihar Delhi 9205541914 shot 1500 nightCheap Rate Call girls Sarita Vihar Delhi 9205541914 shot 1500 night
Cheap Rate Call girls Sarita Vihar Delhi 9205541914 shot 1500 night
 
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
 
Capstone Project on IBM Data Analytics Program
Capstone Project on IBM Data Analytics ProgramCapstone Project on IBM Data Analytics Program
Capstone Project on IBM Data Analytics Program
 
Call me @ 9892124323 Cheap Rate Call Girls in Vashi with Real Photo 100% Secure
Call me @ 9892124323  Cheap Rate Call Girls in Vashi with Real Photo 100% SecureCall me @ 9892124323  Cheap Rate Call Girls in Vashi with Real Photo 100% Secure
Call me @ 9892124323 Cheap Rate Call Girls in Vashi with Real Photo 100% Secure
 
Generative AI on Enterprise Cloud with NiFi and Milvus
Generative AI on Enterprise Cloud with NiFi and MilvusGenerative AI on Enterprise Cloud with NiFi and Milvus
Generative AI on Enterprise Cloud with NiFi and Milvus
 
Vip Mumbai Call Girls Thane West Call On 9920725232 With Body to body massage...
Vip Mumbai Call Girls Thane West Call On 9920725232 With Body to body massage...Vip Mumbai Call Girls Thane West Call On 9920725232 With Body to body massage...
Vip Mumbai Call Girls Thane West Call On 9920725232 With Body to body massage...
 
FESE Capital Markets Fact Sheet 2024 Q1.pdf
FESE Capital Markets Fact Sheet 2024 Q1.pdfFESE Capital Markets Fact Sheet 2024 Q1.pdf
FESE Capital Markets Fact Sheet 2024 Q1.pdf
 
ALSO dropshipping via API with DroFx.pptx
ALSO dropshipping via API with DroFx.pptxALSO dropshipping via API with DroFx.pptx
ALSO dropshipping via API with DroFx.pptx
 
Predicting Loan Approval: A Data Science Project
Predicting Loan Approval: A Data Science ProjectPredicting Loan Approval: A Data Science Project
Predicting Loan Approval: A Data Science Project
 
CebaBaby dropshipping via API with DroFX.pptx
CebaBaby dropshipping via API with DroFX.pptxCebaBaby dropshipping via API with DroFX.pptx
CebaBaby dropshipping via API with DroFX.pptx
 
Invezz.com - Grow your wealth with trading signals
Invezz.com - Grow your wealth with trading signalsInvezz.com - Grow your wealth with trading signals
Invezz.com - Grow your wealth with trading signals
 
Discover Why Less is More in B2B Research
Discover Why Less is More in B2B ResearchDiscover Why Less is More in B2B Research
Discover Why Less is More in B2B Research
 
Week-01-2.ppt BBB human Computer interaction
Week-01-2.ppt BBB human Computer interactionWeek-01-2.ppt BBB human Computer interaction
Week-01-2.ppt BBB human Computer interaction
 
Smarteg dropshipping via API with DroFx.pptx
Smarteg dropshipping via API with DroFx.pptxSmarteg dropshipping via API with DroFx.pptx
Smarteg dropshipping via API with DroFx.pptx
 
April 2024 - Crypto Market Report's Analysis
April 2024 - Crypto Market Report's AnalysisApril 2024 - Crypto Market Report's Analysis
April 2024 - Crypto Market Report's Analysis
 
Call Girls Begur Just Call 👗 7737669865 👗 Top Class Call Girl Service Bangalore
Call Girls Begur Just Call 👗 7737669865 👗 Top Class Call Girl Service BangaloreCall Girls Begur Just Call 👗 7737669865 👗 Top Class Call Girl Service Bangalore
Call Girls Begur Just Call 👗 7737669865 👗 Top Class Call Girl Service Bangalore
 
Call Girls in Sarai Kale Khan Delhi 💯 Call Us 🔝9205541914 🔝( Delhi) Escorts S...
Call Girls in Sarai Kale Khan Delhi 💯 Call Us 🔝9205541914 🔝( Delhi) Escorts S...Call Girls in Sarai Kale Khan Delhi 💯 Call Us 🔝9205541914 🔝( Delhi) Escorts S...
Call Girls in Sarai Kale Khan Delhi 💯 Call Us 🔝9205541914 🔝( Delhi) Escorts S...
 
Junnasandra Call Girls: 🍓 7737669865 🍓 High Profile Model Escorts | Bangalore...
Junnasandra Call Girls: 🍓 7737669865 🍓 High Profile Model Escorts | Bangalore...Junnasandra Call Girls: 🍓 7737669865 🍓 High Profile Model Escorts | Bangalore...
Junnasandra Call Girls: 🍓 7737669865 🍓 High Profile Model Escorts | Bangalore...
 
Mg Road Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Banga...
Mg Road Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Banga...Mg Road Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Banga...
Mg Road Call Girls Service: 🍓 7737669865 🍓 High Profile Model Escorts | Banga...
 

Quantum generalized linear models

  • 1. Quantum Generalized Linear Models: A Proof of Concept Uchenna Chukwu and Colleen M. Farrelly, Quantopo LLC
  • 3. Overview and Implications  Generalized linear models are the simplest instance of link-based statistical models, which are based on the underlying geometry of an outcome’s underlying probability distribution (typically from the exponential family).  Machine learning algorithms provide alternative ways to minimize a model’s sum of square error (error between predicted values and actual values of a test set).  However, some deep results regarding the exponential family’s relation to affine connections in differential geometry provide a possible alternative to link functions: 1. Algorithms that either continuously deform the outcome distribution from known results 2. Algorithms that superpose all possible distributions and collapse to fit a dataset  Leveraging the fact that some quantum computer gates, such as the non-Gaussian transformation gate, essentially perform (1) natively and in a computationally-efficient way!  This project provides a proof-of-concept for leveraging specific hardware gates to solve the affine connection problem, with benchmarking at state-of-the-art levels.  Results can be extended to many other, more complicated statistical models, such as generalized estimating equations, hierarchical regression models, and even homotopy-continuation problems.
  • 4. Generalized Linear Model Background and Business Usage An Introduction to Tweedie Models
  • 5. Generalized Linear Models  Generalized linear modeling (GLM) as extension of linear regression to outcomes with probability distributions that are not Gaussian  Includes binomial outcomes, Poisson outcomes, gamma outcomes, and many more  Link functions to transform distribution of these outcomes to a normal distribution to fit a linear model  𝐸 𝑌 = µ = 𝑔−1 𝑿𝛽  Var 𝑌 = 𝑉𝑎𝑟(µ) = 𝑉𝑎𝑟(𝑔−1 𝑿𝛽 )  Where Y is a vector of outcome values, µ is the mean of Y, X is the matrix of predictor values, g is a link function (such as the log function), and β is a vector of predictor weights in the regression equation.  Many statistical extensions:  Generalized estimating equations (longitudinal data modeling)  Generalized linear mixed models (longitudinal data with random effects)  Generalized additive models (in which the predictor vectors can be transformed within the model)  Cox regression and Weibull-based regression (survival data modeling)  Very high computational cost for many of these extensions
  • 6. Important Applications of GLMs  Ubiquitous in part failure modeling, medical research, actuarial science, and many other problems  Example problems:  Modeling likelihood of insurance claims and expected payout (worldwide, a $5 trillion industry)  Understanding risk behavior in medical research (daily heroin usage, sexual partners within prior month…)  Modeling expected failure rates and associated conditions for airplane parts or machine parts within a manufacturing plant (~$4 trillion industry in the USA alone)  Modeling expected natural disaster impacts and precipitating factors related to impact extent  Many supervised learning algorithms are extensions of generalized linear models and have link functions built into the algorithm to model different outcome distributions  Boosted regression, Morse-Smale regression, dgLARS, Bayesian model averaging…  Optimization algorithm to find minimum sum of square error differ among machine learning methods and with respect to GLMs, which use a least square error algorithm  Methods like deep learning and classical neural networks attempt to solve this problem in a general way through a series of general mappings leading to a potentially novel link function  Exploiting the geometric relationships between distributions through a superposition of states collapsed to the “ideal” link would present an optimal solution to the problem  Tweedie regression as a general framework that handles many distributions in the exponential family and the problem of overdispersion of model/outcome variance  Very nice geometric properties  Connected to many common exponential family distributions
  • 7. Details of Tweedie Regression  Many common distributions of the exponential family converge to Tweedie distributions and can be formulated through Tweedie distributions, formally defined as:  𝐸 𝑌 = µ  𝑉𝑎𝑟 𝑌 = 𝜑µ 𝜉  where 𝜑 is the dispersion parameter, and 𝜉is the Tweedie parameter (or shape parameter)  Tweedie distributions themselves enjoy a variety of useful properties:  Reproductive properties that allow distributions to be added together to form new distributions that are themselves Tweedie  Varying Tweedie parameter and dispersion parameter to recover many exponential family distributions used in GLMs:  Tweedie parameter of 0 for normal distribution  Tweedie parameter of 1 for Poisson distribution  Tweedie parameter of 2 for gamma distribution  Dispersion parameter for 0-inflated models and outliers, similar to negative binomial regression models
  • 8. The Problem of Overdispersion in Tweedie Models  Well-known statistical problem involving dispersion parameters, which relate to the variance of a outcome  Many GLMs and their machine learning extensions struggle on problems of overdispersion  Simulations show this behavior, particularly as dispersion parameter increases substantially (values of 4+)  Empirical datasets with 0-inflation and long tails  Recent paper exploring bagged KNN models  Demonstrates problem in simulations  Demonstrates with open-source datasets, such as UCI’s Forest Fire dataset  Models that work well, such as the KNN ensemble with varying k parameters, tend to take a long time to compute
  • 9. Common Tweedie Models Family Distribution Dispersion (extra 0’s and tail fatness) Power (variance proportional to mean: 1/Power) Normal 1 0 Poisson 1 1 Compound Poisson 1 >1 and <2 Gamma 1 2 Inverse-Gaussian 1 3 Stable 1 >2 (Extreme >3) Negative Binomial >1 1 Underdispersion Poisson <1 1 Unique Tweedie >=1 >=0
  • 10. Connection of GLMs to Differential Geometry Motivation for Implementation on Xanadu System
  • 11. Differential Geometry and the Exponential Family  Possible to formulate exponential family distributions and their parameterizations to form a series of curves on a 2-dimensional surface  Each curve defined by 2 points at either end of the probability function, 0 and 1, connected by a line that follows a shortest path following parameterization of the distribution, called a geodesic  Because the exponential family can be generalized into Tweedie distributions through continuous transformations, the geodesic connecting 0 and 1 can flow across distributions defining the 2-dimensional surface in a continuous manner (much like homotopy continuation methods).  This is an affine connection, and the morphing of the line as it passes parameters transforms one distribution to another.
  • 12. Consequences of Exponential Family Geometry  Analytically- derived results/equation for one distribution morphed to fit another distribution through continuous transformations!  Limit theorems derived by continuous deformations of either moment generating functions or characteristic functions
  • 13. Xanadu Technology and Suitability to GLMs  Xanadu’s qumode formulation makes ideal for implementing quantum GLMs  Ability to perform linear algebra operations on physical data representations  GLMs and their extensions all based on simple matrix operations  𝑀𝑒𝑎𝑛 𝑌 = 𝑔−1 𝑿𝛽 + 𝜀  Matrix multiplication and addition for the linear model 𝑿𝛽 + 𝜀 coupled with a continuous transformation of the model results to fit the outcome distribution  Non-Gaussian transformation gate provides perfect avenue to perform the affine transformation related to the outcome distribution without a need to specific a link function to approximate the geometry  Should be able to approximate any continuous outcome’s distribution, creating potential new “link functions” through this gate through affine transformation of the wavefunctions representing the data  Removes the need for approximations by easy-to-compute link transformations  In theory, should approximate any continuous distribution, including ones that aren’t included in common statistical packages implementing GLMs and their longitudinal/survival data extensions  Thus, Xanadu’s system provides a general solution to the linear regression equation with many potential extensions to more sophisticated regression models!
  • 14. Methods and Results on Example Cases Simulated overdispersion dataset and UCI Forest Fire dataset
  • 15. Methodology  Simulation  Similar to simulations used in the KNN ensemble paper  1000 observations with a 70/30 test/train split  Tweedie outcome related to 3 predictors (1 interaction term, 1 main effect) with added noise  Tweedie parameter=1, dispersion parameter=8  1 noise variable added  Empirical dataset  UCI Repository’s Forest Fire dataset  Notoriously difficult to beat the mean model with machine learning algorithms  12 predictors (2 spatial coordinates of location, month, day, FFMC index, DMC index, DC index, ISI index, temperature, relative humidity, wind, and rain) and 517 observations  t-SNE was used to reduce the dimensionality of the predictor set to 4 components so as to make it compatible with Xanadu’s capabilities.  70/30 test/train split  Comparison methods  Boosted regression  Random forest (tree-based bagged ensemble)  DGLARS (tangent-space-based least angle regression model)  BART (Bayesian-based tree ensemble)  HLASSO (homotopy-based LASSO model)  Poisson regression (GLM without any modifications)
  • 16. Data Preprossessing  Dimensionality reduction through t-SNE to create a set of 4 predictors and 1 outcome, such that predictors are uncorrelated when entered into models.  Easier for systems to calculate with fewer variables.  Decorrelation helps most regression methods, including linear models and tree models.  Other dimensionality reduction methods are possible, including the introduction of factors from factor analytic models or combinations of linear/nonlinear, global/local dimensionality reduction algorithms.  Scaling of outcome to a scale of -3 to 3, such that the Xanadu simulation can effectively model and process the data in qumodes.  Slight warping of the most extreme values, but these are generally less than 5 observations per dataset.  Other types of scaling might be useful to explore.
  • 17. Qumodes Circuit Details  GLMs can be embedded within Xanadu’s qumode quantum computer simulation software (and qumode computer) with a singular value decomposition of the 𝛽 coefficient in the formulation:  𝑀𝑒𝑎𝑛 𝑌 = 𝑔−1 𝑋𝛽  This translates to 𝛽 = 𝑂1Σ𝑂2, which can be modeled through a series of quantum circuit gates:  Multiplication of X and an orthogonal matrix:  | 𝑂1 𝑋 ≅ 𝑈1| 𝑋 , which corresponds to a linear interferometer gate (𝑈1) acting on X  Multiplication of that result by a diagonal matrix:  |Σ𝑂1 𝑋 ∝ 𝑆 𝑟 | 𝑂1 𝑋 , which corresponds to a squeezing gate that acts on a single qumode  Multiplication of X and an orthogonal matrix:  | 𝑂2Σ𝑂1 𝑋 ≅ 𝑈2|Σ 𝑂1 𝑋 , which corresponds to a linear interferometer gate (𝑈2) acting on the result  Multiplication by a nonlinear function on this result:  |𝑔−1 ( 𝑂2Σ𝑂1 𝑋) ≅ Φ|𝑂2Σ𝑂1 𝑋 , which corresponds to the non-Gaussian gate acting on the result  This gives a final result of gates acting upon the dataset as:  Φ ∗ 𝒰2 ∗ 𝒮 ∗ 𝒰1| 𝑋 ∝ | 𝑔−1 𝑋𝛽
  • 18. Qumodes Parameter Settings  The algorithm simulation was created through Strawberry Fields.  The deep learning framework already existed.  Hidden layers and bias terms were removed to collapse to a generalized linear model framework.  The loss function optimized was mean square error, which corresponds to the loss functions specified in the comparison algorithms.  Qumode cut-off dimension was set to 10.  Optimization via least squares was not available, so gradient descent was used with a learning rate of 0.1 over 80 iterations.  This gave a qumodes implementation of a quantum generalized linear model with a boosting feel to it.  Because the quantum computing component is inherently probabilitistic, algorithms were run on the same training and test set multiple times to average out quantum effects.
  • 19. Results: Simulation of Overdispersion Algorithm Scaled Model MSE Random Forest 0.80 BART 0.78 Boosted Regression 0.78 DGLARS 0.81 HLASSO 0.81 GLM 0.81 QGLM 0.82 Mean 0.85 QGLMs yield slightly worse prediction on the simulated dataset. However, their performance is not far off from state-of-the art algorithms, and some random error is expected from the quantum machinery.
  • 20. Results: Forest Fire Dataset Algorithm Scaled Model MSE Random Forest 0.125 BART 0.125 Boosted Regression 0.119 DGLARS 0.114 HLASSO 0.120 GLM 0.119 QGLM 0.106 Mean 0.115 QGLMs emerge as the best-performing algorithm on a difficult, real- world dataset (Forest Fire dataset in the UCI repository). QGLMs provide ~10% gain over the next best algorithm on this dataset. This suggests that they work well on real data and difficult problems.
  • 21. Conclusions  This suggests that the qumodes formulation with its unique operators can eliminate the need for link functions within linear models by exploiting the geometry of the models and still give good prediction.  Better than state-of-the-art prediction for a difficult Tweedie regression dataset (UCI Forest Fire)  Around state-of-the-art prediction for a simulated dataset  This has the potential to bring statistical modeling into quantum computing, by leveraging the underlying geometry and the connection between model geometry and the geometry of quantum physics.  Generalized estimating equations/generalized linear mixed models  Structural equation models/hierarchical regression models  Also a potential avenue through which to implement the homotopy continuation method common in dynamic systems research and some machine learning models (such as homotopy-based LASSO), which take a known problem’s solution and continuously deform it to fit the problem of interest.  Currently a computational challenge  Limited to small datasets
  • 22. References  Amari, S. I. (1997). Information geometry. Contemporary Mathematics, 203, 81-96.  Bulmer, M. G. (1974). On fitting the Poisson lognormal distribution to species-abundance data. Biometrics, 101-110.  Buscemi, F. (2012). Comparison of quantum statistical models: equivalent conditions for sufficiency. Communications in Mathematical Physics, 310(3), 625-647.  Cortez, P., & Morais, A. D. J. R. (2007). A data mining approach to predict forest fires using meteorological data.  De Jong, P., & Heller, G. Z. (2008). Generalized linear models for insurance data (Vol. 10). Cambridge: Cambridge University Press.  Farrelly, C. M. (2017). KNN Ensembles for Tweedie Regression: The Power of Multiscale Neighborhoods. arXiv preprint arXiv:1708.02122.  Farrelly, C. M. (2017). Topology and Geometry in Machine Learning for Logistic Regression.  Fehm, L., Beesdo, K., Jacobi, F., & Fiedler, A. (2008). Social anxiety disorder above and below the diagnostic threshold: prevalence, comorbidity and impairment in the general population. Social psychiatry and psychiatric epidemiology, 43(4), 257-265.  Fergusson, D. M., Boden, J. M., & Horwood, L. J. (2006). Cannabis use and other illicit drug use: testing the cannabis gateway hypothesis. Addiction, 101(4), 556-569.  Frees, E. W., Lee, G., & Yang, L. (2016). Multivariate frequency-severity regression models in insurance. Risks, 4(1), 4.  Gardner, W., Mulvey, E. P., & Shaw, E. C. (1995). Regression analyses of counts and rates: Poisson, overdispersed Poisson, and negative binomial models. Psychological bulletin, 118(3), 392.  Herings, R., & Erkens, J. A. (2003). Increased suicide attempt rate among patients interrupting use of atypical antipsychotics. Pharmacoepidemiology and drug safety, 12(5), 423-424.
  • 23. References  Jorgensen, B. (1997). The theory of dispersion models. CRC Press.  Jørgensen, B., Goegebeur, Y., & Martínez, J. R. (2010). Dispersion models for extremes. Extremes, 13(4), 399-437.  Killoran, N., Bromley, T. R., Arrazola, J. M., Schuld, M., Quesada, N., & Lloyd, S. (2018). Continuous-variable quantum neural networks. arXiv preprint arXiv:1806.06871.  Killoran, N., Izaac, J., Quesada, N., Bergholm, V., Amy, M., & Weedbrook, C. (2018). Strawberry Fields: A Software Platform for Photonic Quantum Computing. arXiv preprint arXiv:1804.03159.  Luitel, B. N. (2016). Prediction of North Atlantic tropical cyclone activity and rainfall (Doctoral dissertation, The University of Iowa).  Marriott, P. (1990). Applications of differential geometry to statistics (Doctoral dissertation, University of Warwick).  Mills, K. L., Teesson, M., Ross, J., & Peters, L. (2006). Trauma, PTSD, and substance use disorders: findings from the Australian National Survey of Mental Health and Well-Being. American Journal of Psychiatry, 163(4), 652-658.  Nielsen, F., & Garcia, V. (2009). Statistical exponential families: A digest with flash cards. arXiv preprint arXiv:0911.4863.  Osborne, M. R., Presnell, B., & Turlach, B. A. (2000). A new approach to variable selection in least squares problems. IMA journal of numerical analysis, 20(3), 389-403.  Pistone, G., & Rogantin, M. P. (1999). The exponential statistical manifold: mean parameters, orthogonality and space transformations. Bernoulli, 5(4), 721-760.  Sawalha, Z., & Sayed, T. (2006). Traffic accident modeling: some statistical issues. Canadian Journal of Civil Engineering, 33(9), 1115-1124.  Tweedie, M. C. K. (1984). An index which distinguishes between some important exponential families. In Statistics: Applications and new directions: Proc. Indian statistical institute golden Jubilee International conference (Vol. 579, p. 6o4).

Notes de l'éditeur

  1. Jorgensen, B. (1997). The theory of dispersion models. CRC Press. Kendal, W. S., & Jørgensen, B. (2011). Tweedie convergence: A mathematical basis for Taylor's power law, 1/f noise, and multifractality. Physical review E, 84(6), 066120.
  2. Marriott, P. (1990). Applications of differential geometry to statistics (Doctoral dissertation, University of Warwick). Nielsen, F., & Garcia, V. (2009). Statistical exponential families: A digest with flash cards. arXiv preprint arXiv:0911.4863. Amari, S. I. (1997). Information geometry. Contemporary Mathematics, 203, 81-96. Pistone, G., & Rogantin, M. P. (1999). The exponential statistical manifold: mean parameters, orthogonality and space transformations. Bernoulli, 5(4), 721-760.
  3. Jørgensen, B., & Martínez, J. R. (2013, March). Multivariate exponential dispersion models. In Multivariate Statistics: Theory and Applications. Proceedings of the IX Tartu Conference on Multivariate Statistics & XX International Workshop on Matrices and Statistics (pp. 73-98). Jorgensen, B. (1997). The theory of dispersion models. CRC Press. Sawalha, Z., & Sayed, T. (2006). Traffic accident modeling: some statistical issues. Canadian Journal of Civil Engineering, 33(9), 1115-1124. Jørgensen, B., Goegebeur, Y., & Martínez, J. R. (2010). Dispersion models for extremes. Extremes, 13(4), 399-437. Marriott, P. (1990). Applications of differential geometry to statistics (Doctoral dissertation, University of Warwick).