Search
Search Funnelback University
- Refined by:
- Date: 2023
Did you mean economiespast |u:mlg.eng.cam.ac.uk?
61 -
110 of
311
search results for Economics test |u:mlg.eng.cam.ac.uk
where 23
match all words and 288
match some words.
Results that match 1 of 2 words
-
Inferring a measure of physiological age frommultiple ageing related…
https://mlg.eng.cam.ac.uk/pub/pdf/KnoParGlaWin11.pdf13 Feb 2023: Table 2: Hold out test. Values are log10(p) where p is the p-value for the Spearman rank correlationhypothesis test. ... The results are shown in Figure 2.2, where we are also able to include binary variables unlikefor the Spearman test. -
The Random Forest Kerneland creating other kernels for big data from…
https://mlg.eng.cam.ac.uk/pub/pdf/DavGha14a.pdf13 Feb 2023: the test log-likelihood on a symmetric logscale as the discrepancy is very large. ... Test MSE on real-world regression problems (lower isbetter). 5.2. Approximation quality. -
Gaussian Mixture Modeling with Gaussian Process Latent Variable Models
https://mlg.eng.cam.ac.uk/pub/pdf/NicRas10.pdf13 Feb 2023: Secondly, our empirical results (see Section 3)indicate, that the test set performance is simply not good. ... Table 1. Average log test densities over 10 random splits of the data. -
Prediction on Spike DataUsing Kernel Algorithms Jan Eichhorn, Andreas …
https://mlg.eng.cam.ac.uk/pub/pdf/EicTolZieetal04.pdf13 Feb 2023: Finallywe train the best model on these four folds and compute an independent test error on theremaining fold. ... Table 1 Mean test error and standard error on the low contrast dataset. -
4F13 Machine Learning: Coursework #4: Reinforcement Learning Zoubin…
https://mlg.eng.cam.ac.uk/teaching/4f13/1011/cw/coursework4.pdf19 Nov 2023: Test yourvalueIteration algorithm. -
4F13 Machine Learning: Coursework #4: Reinforcement Learning Zoubin…
https://mlg.eng.cam.ac.uk/teaching/4f13/0910/cw/coursework4.pdf19 Nov 2023: Test yourvalueIteration algorithm. -
Adaptive Sequential Bayesian Change Point Detection Ryan…
https://mlg.eng.cam.ac.uk/pub/pdf/TurSaaRas09.pdf13 Feb 2023: We also include the 95% error bars on the NLL and the p-value that the joint model/learned hypers hasa higher NLL using a one sided t-test. ... Industry: We test on the last 8455 points of the portfolio data, 3 July 1975–31 December 2008. -
PROPAGATION OF UNCERTAINTY IN BAYESIAN KERNEL MODELS— APPLICATION TO…
https://mlg.eng.cam.ac.uk/pub/pdf/QuiGirLarRas03.pdf13 Feb 2023: Thiscorresponds to using the model in recall/test phase under uncer-tain input. ... 3. PREDICTION WITH UNCERTAIN INPUT. Assume that the test inputx can not be observed directly and theuncertainty is modeled asx p(x) = N (u, S), with meanuand covariance -
MODEL BASED LEARNING OF SIGMA POINTS IN UNSCENTED KALMAN ...
https://mlg.eng.cam.ac.uk/pub/pdf/TurRas10.pdf13 Feb 2023: 11). 2If we want to integrate the parameters out we must run the UKF witheach sample of θ|y1:T during test and average. ... 6.5. Computational Complexity. The UKF-L, UKF, and EKF have test set computationaltime O(DT(D2 M)). -
Nonparametric Bayesian Sparse Factor Models with application to Gene…
https://mlg.eng.cam.ac.uk/pub/pdf/KnoGha11b.pdf13 Feb 2023: Figure 8(a) shows test set log likelihoods for 10 ran-dom divisions of the data into training and test sets. ... test. dat. a. SFAAFA NS. FA. Fig 9. Test set log likelihoods on Prostate cancer dataset from Yu et al. -
Scaling the Indian Buffet Process via Submodular Maximization
https://mlg.eng.cam.ac.uk/pub/pdf/ReeGha13a.pdf13 Feb 2023: aibp. t-aibp. f-vibpi-vibp. seconds. test. log-likelihood ugibbs. t-ugibbs. t-aibp. aibp. bnmf. ... test. log-likelihood. Piano. meibp. ugibbs. aibp. bnmf. f-vibp. i-vibp. 103 104 1054.95. -
Gaussian Process Change Point Models
https://mlg.eng.cam.ac.uk/pub/pdf/SaaTurRas10.pdf13 Feb 2023: Weevaluated the models’ ability to predict next day snow-fall using 35 years of test data. ... Method Negative Log Likelihood p-value MSE p-valueNile Data (200 Training Points, 462 Test Points). -
Variable noise and dimensionality reduction forsparse Gaussian…
https://mlg.eng.cam.ac.uk/zoubin/papers/snelson_uai.pdf27 Jan 2023: We have triedto implement both versions efficiently. Validation Time /s. Method NLPD MSE Train Test. ... To test this weused PCA to reduce the dimension to 5, before usingthe SPGP. -
G:\bioinformatics\Bioinfo-26(7)issue\btq053.dvi
https://mlg.eng.cam.ac.uk/pub/pdf/LipGhaBor10.pdf13 Feb 2023: A two-sample test tries to decide whether twosamples, in our case x and DC , have been generated by the samedistribution or not. ... edge,and might artificially boost prediction accuracy if they appear in both trainingand test set. -
Relational Learning with Gaussian Processes Wei ChuCCLS Columbia…
https://mlg.eng.cam.ac.uk/pub/pdf/ChuSinGhaetal07.pdf13 Feb 2023: in the input space and provides proba-bilistic induction over unseen test points. ... K̃(zm, zt)]T. One can computethe Bernoulli distribution over the test labelyt by. -
paper8-lect0-13
https://mlg.eng.cam.ac.uk/zoubin/p8-07/lect0.pdf27 Jan 2023: of belonging with the query set. The algorithm is very fast: about 0.2 sec on a laptop to query 22,000 test images. -
4F13 Machine Learning: Coursework #1: Gaussian Processes Carl Edward…
https://mlg.eng.cam.ac.uk/teaching/4f13/1415/cw/coursework1.pdf19 Nov 2023: Show and comment on the fit and the hypers, and the predictions forthe test data. -
Relational Learning with Gaussian Processes Wei ChuCCLS Columbia…
https://mlg.eng.cam.ac.uk/zoubin/papers/relationalgp.pdf27 Jan 2023: in the input space and provides proba-bilistic induction over unseen test points. ... K̃(zm, zt)]T. One can computethe Bernoulli distribution over the test labelyt by. -
544 The Block Diagonal Infinite Hidden Markov Model Thomas ...
https://mlg.eng.cam.ac.uk/pub/pdf/SteGhaGoretal09.pdf13 Feb 2023: Each dataset had 2000 stepsof training data and 2000 steps of test data. ... these were not significantly different (two sam-ple t-test, p = 0.3). -
Determinantal Clustering Process - A Nonparametric BayesianApproach…
https://mlg.eng.cam.ac.uk/pub/pdf/ShaGha13a.pdf13 Feb 2023: Next anyparameters of the density model can be integrated outto produce a predictive clustering of unseen test data. ... lected wheat types) and leave 5 data points from eachwheat type as unobserved test points. -
The Infinite Hidden Markov Model Matthew J. Beal Zoubin ...
https://mlg.eng.cam.ac.uk/zoubin/papers/ihmm.pdf27 Jan 2023: We propose estimating the likelihood of a test sequence given a learned model using particlefiltering. ... 6Different particle initialisations apply if we do not assume that the test sequence immediatelyfollows the training sequence. -
The Supervised IBP: Neighbourhood PreservingInfinite Latent Feature…
https://mlg.eng.cam.ac.uk/pub/pdf/QuaShaKnoGha13.pdf13 Feb 2023: the in-ferred test latent variable z with respect to the train-ing latent variables Z. ... boldface is significant using a one-sided paired t-test with 95% confidence. -
1 Learning the Structure of Deep Sparse Graphical Models ...
https://mlg.eng.cam.ac.uk/pub/pdf/AdaWalGha10.pdf13 Feb 2023: c) (d)Figure 4: Olivetti faces a) Test images on the left, withreconstructed bottom halves on the right. ... Fig 4ashows six bottom-half test set reconstructions on theright, compared to the ground truth on the left. -
Archipelago: Nonparametric Bayesian Semi-Supervised Learning Ryan…
https://mlg.eng.cam.ac.uk/pub/pdf/AdaGha09.pdf13 Feb 2023: In almost all of our tests, Archipelagohad lower test classification error than the NCNM. ... Itimproves over mixture-based Bayesian approaches toSSL while still modeling complex density functions.In empirical tests, our model compares favorably -
Gibbs sampling (an MCMC method) and relations to EM
https://mlg.eng.cam.ac.uk/zoubin/SALD/week7at.pdf27 Jan 2023: Uniform U(0, θ] distribution, θ > 0. We select at random m n disks, having a common θ for failure We select n of these (at random) and test them until -
chu05a.dvi
https://mlg.eng.cam.ac.uk/pub/pdf/ChuGha05a.pdf13 Feb 2023: 0.23370.0072. Table 2: Test results of the three algorithms using a Gaussian kernel. ... Figure 4 presents the test results of the three algorithms for different numbers ofselected genes. -
Dependent Indian Buffet Processes Sinead Williamson Peter Orbanz…
https://mlg.eng.cam.ac.uk/pub/pdf/WilOrbGha10.pdf13 Feb 2023: The data was randomly split into atraining set of 130 countries, and a test set of 14 coun-tries. ... For each test country, one randomly selectedindicator was observed, and the remainder held outfor prediction. -
Learning to Control a Low-Cost Manipulator usingData-Efficient…
https://mlg.eng.cam.ac.uk/pub/pdf/DeiRasFox11.pdf13 Feb 2023: Althoughdeposit failure feedback was not available to the learner, thedeposit success is good across 10 test trials and four differenttraining setups. ... Second, Tab. III reports the block-deposit success rates for10 test runs (and four different -
Probabilistic Modelling, Machine Learning,and the Information…
https://mlg.eng.cam.ac.uk/zoubin/talks/mit12csail.pdf27 Jan 2023: Some Comparisons. Table 1: Test errors and predictive accuracy (smaller is better) for the GP classifier, the supportvector machine, the informative vector machine, and the sparse pseudo-inputGP classifier. ... Data set GPC SVM IVM SPGPC. name train:test -
Approximate inference for the loss-calibrated Bayesian
https://mlg.eng.cam.ac.uk/pub/pdf/LacHusGha11.pdf13 Feb 2023: We also assume. the transductive scenario where we are given a test setS of S points {xs}Ss=1, i.e. ... of the shift between the test and trainingdistributions (columns) and the asymmetry of loss (rows). -
LNAI 3944 - Evaluating Predictive Uncertainty Challenge
https://mlg.eng.cam.ac.uk/pub/pdf/QuiRasSinetal06.pdf13 Feb 2023: The participants could then usethem to train their algorithms before submitting the test predictions. ... The test results were made public on December 11. The website remainsopen for submission. -
Warped Gaussian Processes Edward Snelson∗ Carl Edward Rasmussen†…
https://mlg.eng.cam.ac.uk/pub/pdf/SneRasGha04.pdf13 Feb 2023: the following table which shows the range of the targets(tmin, tmax), the number of input dimensions (D), and the size of the training and test sets(Ntrain, Ntest) that we ... We show three measuresof performance over independent test sets: mean absolute -
Draft version; accepted for NIPS*03 Warped Gaussian Processes Edward…
https://mlg.eng.cam.ac.uk/zoubin/papers/gpwarp.pdf27 Jan 2023: the following table which shows the range of the targets(tmin, tmax), the number of input dimensions (D), and the size of the training and test sets(Ntrain, Ntest) that we ... We show three measuresof performance over independent test sets: mean absolute -
The EM-EP Algorithm forGaussian Process Classification Hyun-Chul Kim� …
https://mlg.eng.cam.ac.uk/zoubin/papers/ecml03.pdf27 Jan 2023: Fig 2 (a) and Fig 2 (b) shows the training set and the test set. ... Each fold were subsequently used as a test set, while the other 9. -
linsys-new.dvi
https://mlg.eng.cam.ac.uk/zoubin/papers/tr-96-2.pdf27 Jan 2023: Ph.D. Thesis, Graduate Group in Managerial Science and Applied Economics,University of Pennsylvania, Philadelphia, PA.Everitt, B. -
Learning to Parse Images
https://mlg.eng.cam.ac.uk/pub/pdf/HinGhaTeh99a.pdf13 Feb 2023: Then the learning. 2,3. 2,4. 2,5. 3,4. 3,5. 4,5. Figure 1: Sample images from the test set. ... tested on the same test set. -
Bayesian Hierarchical Clustering Katherine A. Heller…
https://mlg.eng.cam.ac.uk/zoubin/papers/bhcnew.pdf27 Jan 2023: As we will see, the main difference isthat our algorithm uses a statistical hypothesis test tochoose which clusters to merge. ... Sec-ond our algorithm is derived from Dirichlet processmixtures. Third the hypothesis test at the core ofour algorithm tests -
A Probabilistic Model for Online Document Clustering with Application …
https://mlg.eng.cam.ac.uk/pub/pdf/ZhaGhaYan04a.pdf13 Feb 2023: In addition to the binary decision “novel” or “non-novel”, eachsystem is required to generated a confidence score for each test document. ... References. [1] The 2002 topic detection & tracking task definition and evaluation -
gppl.dvi
https://mlg.eng.cam.ac.uk/zoubin/papers/icml05chuwei-pl.pdf27 Jan 2023: 2.3. Prediction. Now let us take a test pair (r, s) on which the pref-erence relation is unknown. ... our algorithm. The test results of the two algorithmsare presented in the left graph of Figure 2. -
Bayesian Classifier Combination Hyun-Chul Kim Zoubin GhahramaniKorea…
https://mlg.eng.cam.ac.uk/pub/pdf/KimGha12.pdf13 Feb 2023: 6500, 1000, 797 data points were selected from the orig-inal test set as a validation set for DNA data set, Satellitedata set, UCI digit data set, respectively. ... BCCresults are based on comparing the posterior mode ofti for data points in the test set -
3F3: Signal and Pattern Processing Lecture 1: Introduction to ...
https://mlg.eng.cam.ac.uk/teaching/3f3/1011/lect1.pdf19 Nov 2023: Computational Neuroscience: neuronal networks, neural information processing,. • Economics: decision theory, game theory, operational research,. -
Gaussian Processes in Reinforcement Learning Carl Edward Rasmussen…
https://mlg.eng.cam.ac.uk/pub/pdf/RasKus04.pdf13 Feb 2023: The predictive distribution for a novel test input. is Gaussian: ) 2! ... on random examples, the relations can already be approximated to within root meansquared errors (estimated on -& test samples and considering the mean of the predicteddistribution) -
LNAI 8189 - Variational Hidden Conditional Random Fields with Coupled …
https://mlg.eng.cam.ac.uk/pub/pdf/BouZafMor13a.pdf13 Feb 2023: In each case, we evaluated their performance on a test set consist-ing of sequences from 3 debates. ... Since we have introduced parameters θ it is sensible to test our methodologyfor signs of overfitting. -
chu05a.dvi
https://mlg.eng.cam.ac.uk/zoubin/papers/chu05a.pdf27 Jan 2023: 0.23370.0072. Table 2: Test results of the three algorithms using a Gaussian kernel. ... Figure 4 presents the test results of the three algorithms for different numbers ofselected genes. -
Nonparametric Transforms of Graph Kernelsfor Semi-Supervised Learning …
https://mlg.eng.cam.ac.uk/zoubin/papers/ZhuKanGhaLaf04.pdf27 Jan 2023: Morespecifically, we restrict ourselves to thetransductive setting where the unlabeled data alsoserve as the test data. ... All classes must be present in the labeled set. The rest is used asunlabeled (test) set in that trial. -
Computational structure of coordinatetransformations: A…
https://mlg.eng.cam.ac.uk/zoubin/papers/coord.pdf27 Jan 2023: We test this conclusion by mapping out the pattern of generalization inducedby one and two remapped points in two dimensions.In the contextual generalization study we examine the question of whether ... To test learning of the remap-ping and -
The Infinite Hidden Markov Model Matthew J. Beal Zoubin ...
https://mlg.eng.cam.ac.uk/pub/pdf/BeaGhaRas02.pdf13 Feb 2023: We propose estimating the likelihood of a test sequence given a learned model using particlefiltering. ... 6Different particle initialisations apply if we do not assume that the test sequence immediatelyfollows the training sequence. -
Bayesian Exponential Family PCA Shakir Mohamed Katherine Heller…
https://mlg.eng.cam.ac.uk/pub/pdf/MohHelGha08.pdf13 Feb 2023: To evaluate the performance of BXPCA, we define training and test data from the available. ... data. The test data was created by randomly selecting 10% of the data points. -
Learning Multiple Related Tasks using Latent Independent Component…
https://mlg.eng.cam.ac.uk/pub/pdf/ZhaGhaYan05a.pdf13 Feb 2023: For both data sets we use the standardtraining/test split, but for RCV1 since the test part of corpus is huge (around 800k docu-ments) we only randomly sample 10k as ... our test set. -
paper.dvi
https://mlg.eng.cam.ac.uk/zoubin/papers/fhmmML.pdf27 Jan 2023: For each randomly sampled set of parameters, aseparate training set and test set were generated and each algorithm was run once. ... Sixty-six chorales, with 40 or more events each, were divided into atraining set (30 chorales) and a test set (36 chorales
Search history
Recently clicked results
Recently clicked results
Your click history is empty.
Recent searches
Recent searches
Your search history is empty.