Search
Search Funnelback University
- Refined by:
- Date: 2023
Did you mean economiespast |u:mlg.eng.cam.ac.uk?
101 -
150 of
314
search results for Economics test |u:mlg.eng.cam.ac.uk
where 23
match all words and 291
match some words.
Results that match 1 of 2 words
-
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. -
erice.dvi
https://mlg.eng.cam.ac.uk/zoubin/papers/erice.pdf27 Jan 2023: The 2000 digits were divided into a training set of 1400 digits,and a test set of 600 digits, with twos and threes being equally representedin both sets. ... a) Average activity fortwos in the test set. b) Average activity for threes 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,. -
Bayesian Learning of Model Structure Zoubin GhahramaniGatsb y…
https://mlg.eng.cam.ac.uk/zoubin/talks/cmu-talk.pdf27 Jan 2023: 8 8. 3. 9 9. 614 176. Training data Test data. ... Tru. e. Tru. e. Classified Classified. $ Each image is classified using hard assignment$ Unsupervised classif: 8.8% train, 7.9% test error.$ K-means (same # of clusters): 12.2%, -
A Probabilistic Model for Online DocumentClustering with Application…
https://mlg.eng.cam.ac.uk/zoubin/papers/zgy-nips04.pdf27 Jan 2023: In addition to the binary decision “novel” or “non-novel”, each system is required to gener-ated a confidence score for each test document. ... References. [1] The 2002 topic detection & tracking task definition and evaluation plan.http://www.nist -
Graphical Models Zoubin Ghahramani Department of…
https://mlg.eng.cam.ac.uk/zoubin/talks/lect2gm.pdf27 Jan 2023: Constraint-Based Learning: Use statistical tests of marginal and conditionalindependence. Find the set of DAGs whose d-separation relations match theresults of conditional independence tests. -
Scaling Multidimensional Gaussian Processes using ProjectedAdditive…
https://mlg.eng.cam.ac.uk/pub/pdf/GilSaaCun13.pdf13 Feb 2023: This paper introduces and tests a novelmethod of projected additive approximationto multidimensional GPs. ... Scaling Multidimensional Gaussian Processes. Algorithm 1 Gaussian Process Regression using SSMsInput: Jointly sorted training and test input -
Policy Search for Learning Robot Control Using Sparse Data
https://mlg.eng.cam.ac.uk/pub/pdf/BisNguHooetal14.pdf13 Feb 2023: For an uncertain test input x N(µ, Σ), the predictivemean µ is given by. ... concludes the derivationsfor the GP prediction with a linear prior mean function atuncertain test inputs. -
Bayesian Gaussian Process Classificationwith the EM-EP…
https://mlg.eng.cam.ac.uk/pub/pdf/KimGha06a.pdf13 Feb 2023: Fig. 2. Graphical model for GPCs with n training data points and one test. ... 0B@. 1CA; ð25Þ. if we assume the test data point does not have labelingerrors and. -
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 -
0000010020030040050060070080090100110120130140150160170180190200210220…
https://mlg.eng.cam.ac.uk/pub/pdf/KnoParGlaWin10.pdf13 Feb 2023: Due to spacelimitations we cannot document all of these tests here, but give one example. ... EP Ordinal Probit. VMP Ordinal Logistic. EP Linear. Figure 4: Synthetic data test. -
Reinforcement Learning with Reference Tracking Controlin Continuous…
https://mlg.eng.cam.ac.uk/pub/pdf/HalRasMac11.pdf13 Feb 2023: The joint distribution of the observed target valuesand the function value at a single deterministic test input x. ... with the new policy and this sampled reference. The datagenerated from this test run is then used to update the learnedGP model and the -
Bayesian Hierarchical Clustering Katherine A. Heller…
https://mlg.eng.cam.ac.uk/zoubin/papers/icml05heller.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 -
Data-Efficient Reinforcement Learning inContinuous State-Action…
https://mlg.eng.cam.ac.uk/pub/pdf/McaRas17.pdf13 Feb 2023: We test ourmethod on the cartpole swing-up task, which involves nonlinear dynamics andrequires nonlinear control. ... We test our algorithm on the cartpole swing-up problem (Figure sec:app-cartpole). -
1 Graph Kernels by Spectral Transforms Xiaojin Zhu Jaz ...
https://mlg.eng.cam.ac.uk/zoubin/papers/ssl-book.pdf27 Jan 2023: We restrict. ourselves to transduction, i.e., the unlabeled data Xu is also the test data. ... the results, we perform paired t-test on test accuracy. We highlight the best accu-. -
Bayesian Learning in Undirected Graphical Models:Approximate MCMC…
https://mlg.eng.cam.ac.uk/pub/pdf/MurGha04a.pdf13 Feb 2023: Our other two test systems are 100-node Boltzmannmachines and demonstrate learning where exact com-putation of Z(W ) is intractable3. ... 3These test sets are available online:http://www.gatsby.ucl.ac.uk/iam23/04blug/. 396 MURRAY & GHAHRAMANI UAI 2004. -
Message Passing Algorithms for Dirichlet Diffusion Trees
https://mlg.eng.cam.ac.uk/pub/pdf/KnoGaeGha11.pdf13 Feb 2023: time (min) time (min) time (min). Figure 6. Per instance test set performance on the macaque skull measurement data Adams et al. ... Score is the per test point, per dimensionlog predictive likelihood. Time is the average computationtime on the system -
Mind Reading by Machine Learning
https://mlg.eng.cam.ac.uk/pub/pdf/HusNopLen10.pdf13 Feb 2023: In trial i of the experiment, thesubject is presented a set of test stimuli Si and gives a re-sponse ri. ... Rows correspondto different conditions, using different training distributions, and different task types to test subjects’ -
Minor typos in eq. (11) and (12) corrected, 2004-12-08.in ...
https://mlg.eng.cam.ac.uk/pub/pdf/Ras00.pdf13 Feb 2023: Further tests on a variety of problems reveals that the infinite mixture model producesdensities whose generalisation is highly competitive with other commonly used methods.Current work is undertaken to explore performance -
btc654.tex
https://mlg.eng.cam.ac.uk/zoubin/papers/Bioinformatics02raval.pdf27 Jan 2023: The RFP for a protein is defined as the fraction ofnegative test proteins (i.e. ... All of the models considered dramaticallyoutperform BLAST2 in this test of remote homologrecognition. -
Learning with Multiple Labels
https://mlg.eng.cam.ac.uk/pub/pdf/JinGha02a.pdf13 Feb 2023: However, since sometimes the prior knowledge on the class label can be misleading, we need to test the robustness of the 'EMPrior Model' to such noisy prior knowledge. ... We varied the number of added classes to test reliability of our algorithm. • -
BIOINFORMATICS ORIGINAL PAPER Vol. 21 no. 3 2005, pages ...
https://mlg.eng.cam.ac.uk/pub/pdf/BeaFalGha05a.pdf13 Feb 2023: The classical approach used in our previous work tests each‘gene–gene’ interaction by doing a hypothesis test compar-ing the bootstrap confidence interval of each parameter to. ... ALGORITHMVariational Bayesian learningThe classical approach tests -
rottpap.dvi
https://mlg.eng.cam.ac.uk/pub/pdf/MurSbaRasGir03.pdf13 Feb 2023: posteriordensity function conditional on the training datax, yand the test pointsx. ... 2.2.1. Prediction at a random input If we now as-sume that the test inputx has a Gaussian distribution,x N (µµµx , ΣΣΣx ), the predictive distribution is -
Learning Multiple Related Tasks using LatentIndependent Component…
https://mlg.eng.cam.ac.uk/zoubin/papers/zgy-nips05.pdf27 Jan 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. -
Occam’s Razor Carl Edward RasmussenDepartment of Mathematical…
https://mlg.eng.cam.ac.uk/pub/pdf/RasGha01.pdf13 Feb 2023: If the modelcomplexity is either too low or too high performance on an independent test set will suffer,giving rise to a characteristic Occam’s Hill. -
1 Robust Filtering and Smoothing with Gaussian Processes Marc ...
https://mlg.eng.cam.ac.uk/pub/pdf/DeiTurHubetal12.pdf13 Feb 2023: Note that xt1serves as a test input from the perspective of the GP regression model. ... HYPOTHESIS THAT THE OTHER FILTERS ARE BETTER THAN THE GP-ADF USING A ONE-SIDED T-TEST. -
manual.dvi
https://mlg.eng.cam.ac.uk/pub/pdf/RasNeaHinetal96.pdf13 Feb 2023: 558.2 Analysis of experiments with common test sets. 588.3 Obtaining performance statistics: The mstats command. ... for the test cases, andthe normalize.n les the normalization constants used in encoding the data. -
Continuous Relaxations for Discrete Hamiltonian Monte Carlo
https://mlg.eng.cam.ac.uk/pub/pdf/ZhaSutSto12a.pdf13 Feb 2023: We test all three samplers on 36 random graphs from each of the two generating processes, usingdifferent values of c1 and c2 for each random graph. ... of messages. Our test set uses a subset of the messages which are small enough that we can run -
Analytic Moment-based Gaussian Process Filtering Marc Peter…
https://mlg.eng.cam.ac.uk/pub/pdf/DeiHubHan09.pdf13 Feb 2023: The posterior predictive dis-tribution of the function value h = h(x) for an arbi-trary test input x is Gaussian with mean and variance. ... Consider the problem of predicting a function valueh(x) for an uncertain test input x N(µ, Σ), whereh GP with -
book
https://mlg.eng.cam.ac.uk/zoubin/papers/CGM.pdf27 Jan 2023: We test these claims experimentally inthe next section. 1.4 Experiments. We test the CGM with a handwritten-word recognition task. ... We test the CGM using the 3 graphical models shown in Figure 1.2. -
Nonlinear Set Membership Regression with Adaptive…
https://mlg.eng.cam.ac.uk/pub/pdf/CalRobRasMac18.pdf13 Feb 2023: We created 555 randomised test runs of the wing rocktracking problems and tested each control algorithm on eachone of them. ... The noise bound would then becomputed as a function of the worst-case error of the POKI-LC predictor on a test sample. -
A Simple Bayesian Framework for Content-Based Image Retrieval…
https://mlg.eng.cam.ac.uk/pub/pdf/HelGha06a.pdf13 Feb 2023: ages were used with their labels as a training set, D, whilethe rest comprised the unlabelled test set, Du. ... at least one image in the Corel test set which is basically. -
Tree-Based Inference for Dirichlet Process Mixtures Yang Xu Machine…
https://mlg.eng.cam.ac.uk/pub/pdf/XuHelGha09.pdf13 Feb 2023: It then iteratively mergespairs of clusters to construct the hierarchy. The maindifference between BHC and traditional hierarchicalclustering methods is that BHC uses a statistical hy-pothesis test to choose which clusters ... For the synthetic datasets, -
ency02.dvi
https://mlg.eng.cam.ac.uk/zoubin/course04/hbtnn2e-III.pdf27 Jan 2023: statistical tests are performed on the data to determine independence and dependence re-. ... be considered for a xed amount computation, because the results of some statistical tests. -
4F13 Machine Learning: Coursework #3: Latent Dirichlet Allocation…
https://mlg.eng.cam.ac.uk/teaching/4f13/1819/cw/coursework3.pdf19 Nov 2023: c) For the Bayesian model, what is the log probability for the test document with ID 2001? ... Explainwhether, when computing the log probability of a test document, you would use the multinomial withor without the “combinatorial factor”. -
Modelling data
https://mlg.eng.cam.ac.uk/teaching/4f13/1819/modelling%20data.pdf19 Nov 2023: generalize from observations in the training set to new test cases(interpolation and extrapolation). • ... make predictions on test cases• interpret the trained model, what insights is the model providing?• evaluate the accuracy of model. • -
Infinite Hidden Markov Models and extensions
https://mlg.eng.cam.ac.uk/zoubin/talks/BayesHMMs10.pdf27 Jan 2023: IHMM evaluation in (Beal et al.,2002) is more elaborate: it allows the IHMM to con-tinue learning about new data encountered during test-ing. ... Subjecting theseresults to the same analysis as the artificial data re-veals similar compared test-set -
- 4F13: Machine Learning
https://mlg.eng.cam.ac.uk/teaching/4f13/0809/lect06.pdf19 Nov 2023: Constraint-Based Learning: Use statistical tests of marginal and conditionalindependence. Find the set of DAGs whose d-separation relations match theresults of conditional independence tests. -
standalone.dvi
https://mlg.eng.cam.ac.uk/pub/pdf/QuiRasWil07.pdf13 Feb 2023: test conditional: p(f|u) = N (K,uK1. u,uu, K, Q,) , (10b). ... Lower graph: assumption of conditional independencebetween training and test function values given u. -
Variational Inference for the Indian Buffet Process Finale…
https://mlg.eng.cam.ac.uk/pub/pdf/DosMilVanTeh09.pdf13 Feb 2023: Table 1: Running times in seconds and test log-likelihoods for the Yale Faces dataset. ... Algorithm K Time Test Log-Likelihood. 2 56 -0.7444Finite Gibbs 5 120 -0.4220. -
A Kernel Approach to Tractable Bayesian Nonparametrics
https://mlg.eng.cam.ac.uk/pub/pdf/HusLac11.pdf13 Feb 2023: We foundthat kst outperformed both competing methods sig-nificantly (p < 0.01, two sample T-test). ... 3. As measureof performance, we used average confusion, i. e. thefraction of misclassified test images. -
Recent Advanced in Causal Modelling Using Directed Graphs
https://mlg.eng.cam.ac.uk/zoubin/SALD/scheines.ppt27 Jan 2023: Can be applied to distributions where tests of conditional independence are known, but scores aren’t. ... 2) X1 - X2 into X2. Test. Test Conditions. X3. X2. -
PILCO: A Model-Based and Data-Efficient Approach to Policy Search
https://mlg.eng.cam.ac.uk/pub/pdf/DeiRas11.pdf13 Feb 2023: 3)–(5). Doing this properlyrequires mapping uncertain test inputs through theGP dynamics model. ... b) Histogram (after 1,000 test runs)of the distances of the flywheel frombeing upright. -
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. -
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. -
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. -
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. -
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)).
Search history
Recently clicked results
Recently clicked results
Your click history is empty.
Recent searches
Recent searches
Your search history is empty.