Felix Agakov - Academia.edu (original) (raw)
Papers by Felix Agakov
Bookmarks Related papers MentionsView impact
Neural Information Processing Systems, Dec 5, 2005
Bookmarks Related papers MentionsView impact
Developing an optimizing compiler for a newly proposed architecture is ex-tremely difficult when ... more Developing an optimizing compiler for a newly proposed architecture is ex-tremely difficult when there is only a simulator of the machine available. Designing such a compiler requires running many experiments in order to understand how different optimizations interact. Given that simulators are orders of magnitude slower than real processors, such experiments are highly restricted. This paper develops a technique to automatically build a performance model for predicting the impact of program transformations on any architecture, based on a limited number of automatically selected runs. As a result, the time for evaluating the impact of any compiler optimiza-tion in early design stages can be drastically reduced such that all selected potential compiler optimizations can be evaluated. This is achieved by first evaluating a small set of sample compiler optimizations on a prior set of benchmarks in order to train a model, followed by a very small number of evaluations, or probes, of the...
Bookmarks Related papers MentionsView impact
Variational methods have proved popular and e#ective for inference and learning in intractable gr... more Variational methods have proved popular and e#ective for inference and learning in intractable graphical models. An attractive feature of the approaches based on the Kullback-Leibler divergence is a rigorous lower bound on the normalization constants in undirected models. In the suggested work we explore the idea of using auxiliary variables to improve on the lower bound of standard mean field methods. Our approach forms a more powerful class of approximations than any structured mean field technique. Furthermore, the existing lower bounds of the variational mixture models could be seen as computationally expensive special cases of our method. A byproduct of our work is an e#cient way to calculate a set of mixture coe#cients for any set of tractable distributions that principally improves on a flat combination.
Bookmarks Related papers MentionsView impact
Recently Hinton (1999) has introduced the Products of Experts (PoE) model in which several indivi... more Recently Hinton (1999) has introduced the Products of Experts (PoE) model in which several individual probabilistic models for data are combined to provide an overall model of the data. Below we consider PoE models in which each expert is a Gaussian.
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Depression is a common psychiatric disorder with substantial recurrence risk. Accurate prediction... more Depression is a common psychiatric disorder with substantial recurrence risk. Accurate prediction from easily collected data would aid in diagnosis, treatment and prevention. We used machine learning in the Generation Scotland cohort to predict lifetime risk of depression and, among cases, recurrent depression. Rank aggregation was used to combine results across ten different algorithms and identify highly predictive variables. The model containing all but the cardiometabolic predictors had the highest predictive ability on independent data. Rank aggregation produced a reduced set of predictors without decreasing predictive performance (lifetime: 20 out of 154 predictors and Receiver Operating Characteristic area under the curve (AUC)=0·84, recurrent: 10 out of 180 predictors and AUC=0·76). Here we develop a pipeline which leads to a small set of highly predictive variables. This information can be easily collected with a smartphone ‘application’ to help diagnosis and treatment, whi...
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Diabetologia, 2015
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Rheumatology, 2019
Objectives Imaging of joint inflammation provides a standard against which to derive an updated D... more Objectives Imaging of joint inflammation provides a standard against which to derive an updated DAS for RA. Our objectives were to develop and validate a DAS based on reweighting the DAS28 components to maximize association with US-assessed synovitis. Methods Early RA patients from two observational cohorts (n = 434 and n = 117) and a clinical trial (n = 59) were assessed at intervals up to 104 weeks from baseline; all US scans were within 1 week of clinical exam. There were 899, 163 and 183 visits in each cohort. Associations of combined US grey scale and power Doppler scores (GSPD) with 28 tender joint count and 28 swollen joint count (SJC28), CRP, ESR and general health visual analogue scale were examined in linear mixed model regressions. Cross-validation evaluated model predictive ability. Coefficients learned from training data defined a re-weighted DAS28 that was validated against radiographic progression in independent data (3037 observations; 717 patients). Results Of the c...
Bookmarks Related papers MentionsView impact
Diabetologia, Jan 5, 2018
As part of the Surrogate Markers for Micro- and Macrovascular Hard Endpoints for Innovative Diabe... more As part of the Surrogate Markers for Micro- and Macrovascular Hard Endpoints for Innovative Diabetes Tools (SUMMIT) programme we previously reported that large panels of biomarkers derived from three analytical platforms maximised prediction of progression of renal decline in type 2 diabetes. Here, we hypothesised that smaller (n ≤ 5), platform-specific combinations of biomarkers selected from these larger panels might achieve similar prediction performance when tested in three additional type 2 diabetes cohorts. We used 657 serum samples, held under differing storage conditions, from the Scania Diabetes Registry (SDR) and Genetics of Diabetes Audit and Research Tayside (GoDARTS), and a further 183 nested case-control sample set from the Collaborative Atorvastatin in Diabetes Study (CARDS). We analysed 42 biomarkers measured on the SDR and GoDARTS samples by a variety of methods including standard ELISA, multiplexed ELISA (Luminex) and mass spectrometry. The subset of 21 Luminex bio...
Bookmarks Related papers MentionsView impact
Biochimica et biophysica acta, Sep 28, 2017
Type 2 diabetes results from interplay between genetic and acquired factors. Glycans on proteins ... more Type 2 diabetes results from interplay between genetic and acquired factors. Glycans on proteins reflect genetic, metabolic and environmental factors. However, associations of IgG glycans with type 2 diabetes have not been described. We compared IgG N-glycan patterns in type 2 diabetes with healthy subjects. In the DiaGene study, a population-based case-control study, (1886 cases and 854 controls) 58 IgG glycan traits were analyzed. Findings were replicated in the population-based CROATIA-Korcula-CROATIA-Vis-ORCADES studies (162 cases and 3162 controls), and meta-analyzed. AUCs of ROC-curves were calculated using 10-fold cross-validation for clinical characteristics, IgG glycans and their combination. After correction for extensive clinical covariates, 5 IgG glycans and 13 derived traits significantly associated with type 2 diabetes in meta-analysis (after Bonferroni correction). Adding IgG glycans to age and sex increased the AUC from 0.542 to 0.734. Adding them to the extensive mo...
Bookmarks Related papers MentionsView impact
Journal of chromatography. B, Analytical technologies in the biomedical and life sciences, Jan 4, 2017
Identification of metabolites in non-targeted metabolomics continues to be a bottleneck in metabo... more Identification of metabolites in non-targeted metabolomics continues to be a bottleneck in metabolomics studies in large human cohorts. Unidentified metabolites frequently emerge in the results of association studies linking metabolite levels to, for example, clinical phenotypes. For further analyses these unknown metabolites must be identified. Current approaches utilize chemical information, such as spectral details and fragmentation characteristics to determine components of unknown metabolites. Here, we propose a systems biology model exploiting the internal correlation structure of metabolite levels in combination with existing biochemical and genetic information to characterize properties of unknown molecules. Levels of 758 metabolites (439 known, 319 unknown) in human blood samples of 2279 subjects were measured using a non-targeted metabolomics platform (LC-MS and GC-MS). We reconstructed the structure of biochemical pathways that are imprinted in these metabolomics data by ...
Bookmarks Related papers MentionsView impact
Kidney international, Jan 22, 2015
Here we evaluated the performance of a large set of serum biomarkers for the prediction of rapid ... more Here we evaluated the performance of a large set of serum biomarkers for the prediction of rapid progression of chronic kidney disease (CKD) in patients with type 2 diabetes. We used a case-control design nested within a prospective cohort of patients with baseline eGFR 30-60 ml/min per 1.73 m(2). Within a 3.5-year period of Go-DARTS study patients, 154 had over a 40% eGFR decline and 153 controls maintained over 95% of baseline eGFR. A total of 207 serum biomarkers were measured and logistic regression was used with forward selection to choose a subset that were maximized on top of clinical variables including age, gender, hemoglobin A1c, eGFR, and albuminuria. Nested cross-validation determined the best number of biomarkers to retain and evaluate for predictive performance. Ultimately, 30 biomarkers showed significant associations with rapid progression and adjusted for clinical characteristics. A panel of 14 biomarkers increased the area under the ROC curve from 0.706 (clinical d...
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Diabetologia, 2013
Bookmarks Related papers MentionsView impact
Information maximization is a common framework of unsupervised learning, which may be used for ex... more Information maximization is a common framework of unsupervised learning, which may be used for extracting informative representations y of the observed patterns x. The key idea there is to maximize mutual information (MI), which is a formal measure of coding efficiency. Unfortunately, exact maximization of MI is computationally tractable only in a few special cases; more generally, approximations need to be considered. Here we describe a family of variational lower bounds on mutual information which gives rise to a formal and theoretically rigorous approach to information maximization in large-scale stochastic channels. We hope that the results presented in this work are potentially interesting for maximizing mutual information from several perspectives. First of all, our method optimizes a proper lower bound, rather than a surrogate objective criterion or an approximation of MI (which may only be accurate under specific asymptotic assumptions, and weak or even undefined when the as...
Bookmarks Related papers MentionsView impact
Human Heredity, 2016
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Neural Information Processing Systems, Dec 5, 2005
Bookmarks Related papers MentionsView impact
Developing an optimizing compiler for a newly proposed architecture is ex-tremely difficult when ... more Developing an optimizing compiler for a newly proposed architecture is ex-tremely difficult when there is only a simulator of the machine available. Designing such a compiler requires running many experiments in order to understand how different optimizations interact. Given that simulators are orders of magnitude slower than real processors, such experiments are highly restricted. This paper develops a technique to automatically build a performance model for predicting the impact of program transformations on any architecture, based on a limited number of automatically selected runs. As a result, the time for evaluating the impact of any compiler optimiza-tion in early design stages can be drastically reduced such that all selected potential compiler optimizations can be evaluated. This is achieved by first evaluating a small set of sample compiler optimizations on a prior set of benchmarks in order to train a model, followed by a very small number of evaluations, or probes, of the...
Bookmarks Related papers MentionsView impact
Variational methods have proved popular and e#ective for inference and learning in intractable gr... more Variational methods have proved popular and e#ective for inference and learning in intractable graphical models. An attractive feature of the approaches based on the Kullback-Leibler divergence is a rigorous lower bound on the normalization constants in undirected models. In the suggested work we explore the idea of using auxiliary variables to improve on the lower bound of standard mean field methods. Our approach forms a more powerful class of approximations than any structured mean field technique. Furthermore, the existing lower bounds of the variational mixture models could be seen as computationally expensive special cases of our method. A byproduct of our work is an e#cient way to calculate a set of mixture coe#cients for any set of tractable distributions that principally improves on a flat combination.
Bookmarks Related papers MentionsView impact
Recently Hinton (1999) has introduced the Products of Experts (PoE) model in which several indivi... more Recently Hinton (1999) has introduced the Products of Experts (PoE) model in which several individual probabilistic models for data are combined to provide an overall model of the data. Below we consider PoE models in which each expert is a Gaussian.
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Depression is a common psychiatric disorder with substantial recurrence risk. Accurate prediction... more Depression is a common psychiatric disorder with substantial recurrence risk. Accurate prediction from easily collected data would aid in diagnosis, treatment and prevention. We used machine learning in the Generation Scotland cohort to predict lifetime risk of depression and, among cases, recurrent depression. Rank aggregation was used to combine results across ten different algorithms and identify highly predictive variables. The model containing all but the cardiometabolic predictors had the highest predictive ability on independent data. Rank aggregation produced a reduced set of predictors without decreasing predictive performance (lifetime: 20 out of 154 predictors and Receiver Operating Characteristic area under the curve (AUC)=0·84, recurrent: 10 out of 180 predictors and AUC=0·76). Here we develop a pipeline which leads to a small set of highly predictive variables. This information can be easily collected with a smartphone ‘application’ to help diagnosis and treatment, whi...
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Diabetologia, 2015
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Rheumatology, 2019
Objectives Imaging of joint inflammation provides a standard against which to derive an updated D... more Objectives Imaging of joint inflammation provides a standard against which to derive an updated DAS for RA. Our objectives were to develop and validate a DAS based on reweighting the DAS28 components to maximize association with US-assessed synovitis. Methods Early RA patients from two observational cohorts (n = 434 and n = 117) and a clinical trial (n = 59) were assessed at intervals up to 104 weeks from baseline; all US scans were within 1 week of clinical exam. There were 899, 163 and 183 visits in each cohort. Associations of combined US grey scale and power Doppler scores (GSPD) with 28 tender joint count and 28 swollen joint count (SJC28), CRP, ESR and general health visual analogue scale were examined in linear mixed model regressions. Cross-validation evaluated model predictive ability. Coefficients learned from training data defined a re-weighted DAS28 that was validated against radiographic progression in independent data (3037 observations; 717 patients). Results Of the c...
Bookmarks Related papers MentionsView impact
Diabetologia, Jan 5, 2018
As part of the Surrogate Markers for Micro- and Macrovascular Hard Endpoints for Innovative Diabe... more As part of the Surrogate Markers for Micro- and Macrovascular Hard Endpoints for Innovative Diabetes Tools (SUMMIT) programme we previously reported that large panels of biomarkers derived from three analytical platforms maximised prediction of progression of renal decline in type 2 diabetes. Here, we hypothesised that smaller (n ≤ 5), platform-specific combinations of biomarkers selected from these larger panels might achieve similar prediction performance when tested in three additional type 2 diabetes cohorts. We used 657 serum samples, held under differing storage conditions, from the Scania Diabetes Registry (SDR) and Genetics of Diabetes Audit and Research Tayside (GoDARTS), and a further 183 nested case-control sample set from the Collaborative Atorvastatin in Diabetes Study (CARDS). We analysed 42 biomarkers measured on the SDR and GoDARTS samples by a variety of methods including standard ELISA, multiplexed ELISA (Luminex) and mass spectrometry. The subset of 21 Luminex bio...
Bookmarks Related papers MentionsView impact
Biochimica et biophysica acta, Sep 28, 2017
Type 2 diabetes results from interplay between genetic and acquired factors. Glycans on proteins ... more Type 2 diabetes results from interplay between genetic and acquired factors. Glycans on proteins reflect genetic, metabolic and environmental factors. However, associations of IgG glycans with type 2 diabetes have not been described. We compared IgG N-glycan patterns in type 2 diabetes with healthy subjects. In the DiaGene study, a population-based case-control study, (1886 cases and 854 controls) 58 IgG glycan traits were analyzed. Findings were replicated in the population-based CROATIA-Korcula-CROATIA-Vis-ORCADES studies (162 cases and 3162 controls), and meta-analyzed. AUCs of ROC-curves were calculated using 10-fold cross-validation for clinical characteristics, IgG glycans and their combination. After correction for extensive clinical covariates, 5 IgG glycans and 13 derived traits significantly associated with type 2 diabetes in meta-analysis (after Bonferroni correction). Adding IgG glycans to age and sex increased the AUC from 0.542 to 0.734. Adding them to the extensive mo...
Bookmarks Related papers MentionsView impact
Journal of chromatography. B, Analytical technologies in the biomedical and life sciences, Jan 4, 2017
Identification of metabolites in non-targeted metabolomics continues to be a bottleneck in metabo... more Identification of metabolites in non-targeted metabolomics continues to be a bottleneck in metabolomics studies in large human cohorts. Unidentified metabolites frequently emerge in the results of association studies linking metabolite levels to, for example, clinical phenotypes. For further analyses these unknown metabolites must be identified. Current approaches utilize chemical information, such as spectral details and fragmentation characteristics to determine components of unknown metabolites. Here, we propose a systems biology model exploiting the internal correlation structure of metabolite levels in combination with existing biochemical and genetic information to characterize properties of unknown molecules. Levels of 758 metabolites (439 known, 319 unknown) in human blood samples of 2279 subjects were measured using a non-targeted metabolomics platform (LC-MS and GC-MS). We reconstructed the structure of biochemical pathways that are imprinted in these metabolomics data by ...
Bookmarks Related papers MentionsView impact
Kidney international, Jan 22, 2015
Here we evaluated the performance of a large set of serum biomarkers for the prediction of rapid ... more Here we evaluated the performance of a large set of serum biomarkers for the prediction of rapid progression of chronic kidney disease (CKD) in patients with type 2 diabetes. We used a case-control design nested within a prospective cohort of patients with baseline eGFR 30-60 ml/min per 1.73 m(2). Within a 3.5-year period of Go-DARTS study patients, 154 had over a 40% eGFR decline and 153 controls maintained over 95% of baseline eGFR. A total of 207 serum biomarkers were measured and logistic regression was used with forward selection to choose a subset that were maximized on top of clinical variables including age, gender, hemoglobin A1c, eGFR, and albuminuria. Nested cross-validation determined the best number of biomarkers to retain and evaluate for predictive performance. Ultimately, 30 biomarkers showed significant associations with rapid progression and adjusted for clinical characteristics. A panel of 14 biomarkers increased the area under the ROC curve from 0.706 (clinical d...
Bookmarks Related papers MentionsView impact
Bookmarks Related papers MentionsView impact
Diabetologia, 2013
Bookmarks Related papers MentionsView impact
Information maximization is a common framework of unsupervised learning, which may be used for ex... more Information maximization is a common framework of unsupervised learning, which may be used for extracting informative representations y of the observed patterns x. The key idea there is to maximize mutual information (MI), which is a formal measure of coding efficiency. Unfortunately, exact maximization of MI is computationally tractable only in a few special cases; more generally, approximations need to be considered. Here we describe a family of variational lower bounds on mutual information which gives rise to a formal and theoretically rigorous approach to information maximization in large-scale stochastic channels. We hope that the results presented in this work are potentially interesting for maximizing mutual information from several perspectives. First of all, our method optimizes a proper lower bound, rather than a surrogate objective criterion or an approximation of MI (which may only be accurate under specific asymptotic assumptions, and weak or even undefined when the as...
Bookmarks Related papers MentionsView impact
Human Heredity, 2016
Bookmarks Related papers MentionsView impact