Merge pull request #1 from mlatcl/filter-target

Combine files from lawrennd/talks

_gpss/_config.yml

@@ -0,0 +1,28 @@
+author:
+- given: Neil D.
+  family: Lawrence
+  institution: University of Cambridge
+  gscholar: r3SJcvoAAAAJ
+  twitter: lawrennd
+  orcid: 0000-0001-9258-1030
+  url: http://inverseprobability.com
+layout: lecture
+venue: Gaussian Process Summer School
+talkcss: https://inverseprobability.com/assets/css/talks.css
+postsdir: ../../../mlatcl/gpss/_lectures/
+slidesdir: ../../../mlatcl/gpss/slides/
+notesdir: ../../../mlatcl/gpss/_notes/
+notebooksdir: ../../../mlatcl/gpss/_notebooks/
+writediagramsdir: .
+scriptsdir: ../scripts/
+diagramsdir: ./slides/diagrams/
+potx: ../_includes/custom-reference.potx 
+dotx: ../_includes/custom-reference.dotx
+transition: None
+baseurl: "gpss/" # the subpath of your site, e.g. /blog/
+url: "https://mlatcl.github.io/" # the base hostname & protocol for your site
+ghub:
+- organization: lawrennd
+  repository: talks
+  branch: gh-pages
+  directory: _gpss

_gpss/approximate-gps.md

@@ -0,0 +1,15 @@
+---
+title: Approximate Gaussian Processes
+featured_image: slides/diagrams/gp/sparse-demo-unconstrained-inducing-6-gp.svg
+week: 6
+youtube: b635kuSqLww
+abstract: 
+---
+
+\include{_gpss/includes/gpss-notebook-setup.md}
+\include{_gp/includes/approximate-gps.md}
+
+\thanks
+
+\references
+

_gpss/bayesian-learning-gplvm.md

@@ -0,0 +1,18 @@
+---
+title: Bayesian Learning of GP-LVM
+featured_image: slides/diagrams/gplvm/singlecell-bayes-gplvm.svg
+week: 10
+abstract:
+---
+
+\include{_gpss/includes/gpss-notebook-setup.md}
+
+\include{_gplvm/includes/bayes-gplvm-intro.md}
+\include{_gplvm/includes/variational-bayes-gplvm-long.md}
+\include{_gplvm/includes/mrd-gplvm.md}
+\include{_gplvm/includes/bayes-gplvm-tutorial.md}
+\include{_gplvm/includes/singlecell-bayes-gplvm.md}
+
+\thanks
+
+\references

_gpss/compile.sh

@@ -0,0 +1,11 @@
+#/bin/bash
+
+FILES=""
+SKIP=true
+while read stub; do
+    if $SKIP; then
+	SKIP=false
+    else
+	maketalk $stub
+    fi
+done < lectures.csv

_gpss/covariance-functions.md

@@ -0,0 +1,71 @@
+---
+title: Covariance Functions and Hyperparameter Optimization
+week: 3
+featured_image: slides/diagrams/kern/sinc_covariance.gif
+abstract: >
+  In this talk we review covariance functions and optimization of the GP log likelihoood.
+author: 
+- given: Neil D. 
+  family: Lawrence
+affiliation: University of Cambridge
+transition: None
+---
+
+<!-- To compile -->
+
+
+\include{_gpss/includes/gpss-notebook-setup.md}
+
+\include{_gp/includes/gp-covariance-function-importance.md}
+\include{_gp/includes/gp-numerics-and-optimization.md}
+\include{_gp/includes/gp-optimize.md}
+
+\include{_kern/includes/eq-covariance.md}
+\include{_kern/includes/computing-rbf-covariance.md}
+
+\comment{Markov property}
+\include{_kern/includes/brownian-covariance.md}
+\include{_kern/includes/precision-matrices.md}
+\include{_kern/includes/ou-covariance.md}
+
+\comment{Basis functions}
+\include{_kern/includes/basis-covariance.md}
+\include{_kern/includes/rbf-basis-covariance.md}
+
+\comment{Fourier space}
+\include{_kern/includes/boechners-theorem.md}
+\include{_kern/includes/sinc-covariance.md}
+\include{_kern/includes/matern32-covariance.md}
+\include{_kern/includes/matern52-covariance.md}
+
+\comment{Scale mixture}
+\include{_kern/includes/ratquad-covariance.md}
+
+\comment{Polynomial}
+\include{_kern/includes/poly-covariance.md}
+
+\comment{Periodic}
+\include{_kern/includes/periodic-covariance.md}
+
+\comment{Infinite Neural Networks}
+\include{_kern/includes/mlp-covariance.md}
+\include{_kern/includes/relu-covariance.md}
+
+\comment{Combining Covariances}
+\include{_kern/includes/add-covariance.md}
+\include{_kern/includes/prod-covariance.md}
+
+\comment{Examples of Deploying Kernels}
+\include{_gp/includes/mauna-loa-gp.md}
+\include{_gp/includes/box-jenkins-airline-gp.md}
+
+
+\comment{Spectral Mixture Kernel}
+\include{_kern/includes/spectral-mixture-kernel.md}
+
+\thanks
+
+\references
+
+
+

_gpss/deep-gps-I.md

@@ -0,0 +1,21 @@
+---
+title: Deep Gaussian Processes I
+featured_image: slides/diagrams//deepgp/deep-nn-bottleneck2.svg
+week: 11
+abstract: 
+---
+
+\include{_gp/includes/mackay-bathwater.md}
+\include{_deepgp/includes/deep-nn-gp.md}
+\include{_ml/includes/deep-learning-overview.md}
+\include{_deepgp/includes/deep-theory.md}
+\include{_deepgp/includes/deep-gp-setup-code.md}
+\include{_deepgp/includes/olympic-marathon-deep-gp.md}
+\include{_deepgp/includes/della-gatta-deep-gp.md}
+\include{_deepgp/includes/step-function-deep-gp.md}
+\include{_deepgp/includes/motorcycle-helmet-deep-gp.md}
+
+\thanks
+
+\references
+

_gpss/deep-gps-II.md

@@ -0,0 +1,14 @@
+---
+title: Deep Gaussian Processes II
+featured_image: slides/diagrams/deepgp/olympic-marathon-deep-gp-pinball.svg
+week: 12
+abstract:
+---
+
+\include{_deepgp/includes/robot-wireless-deep-gp.md}
+\include{_deepgp/includes/deep-results.md}
+\include{_health/includes/deep-health-model.md}
+
+\thanks
+
+\references

_gpss/emulation.md

@@ -0,0 +1,64 @@
+---
+title: Emulation
+week: 13
+featured_image: slides/diagrams/uq/statistical-emulation001.svg
+abstract: In this session we introduce the notion of emulation and systems modeling with Gaussian processes.
+date: 2021-09-15
+venue: Virtual GPSS
+author:
+- family: Lawrence
+  given: Neil D.
+  gscholar: r3SJcvoAAAAJ
+  institute: University of Cambridge
+  twitter: lawrennd
+  orcid: 0000-0001-9258-1030
+  url: http://inverseprobability.com
+---
+
+
+\newslide{}
+
+> We may regard the present state of the universe as the effect of its
+> past and the cause of its future. An intellect which at a certain
+> moment would know all forces that set nature in motion, and all
+> positions of all items of which nature is composed, ...
+\newslide{}
+> ... if this intellect
+> were also vast enough to submit these data to analysis, it would
+> embrace in a single formula the movements of the greatest bodies of
+> the universe and those of the tiniest atom; for such an intellect
+> nothing would be uncertain and the future just like the past would be
+> present before its eyes.
+>
+> --- Pierre Simon Laplace [@Laplace-essai14]
+
+
+\include{_simulation/includes/game-of-life.md}
+
+\speakernotes{Laplace's demon requires us to also know positions of all items and to submit the data to analysis.}
+
+\newslide{}
+
+\notes{We summarize this notion as}
+$$
+\text{data} + \text{model} \stackrel{\text{compute}}{\rightarrow} \text{prediction}
+$$
+\notes{As we pointed out, there is an irony in Laplace's demon forming the cornerstone of a movement known as 'determinism', because Laplace wrote about this idea in an essay on probabilities. The more important quote in the essay was }
+
+\include{_physics/includes/laplaces-gremlin.md}
+
+\include{_simulation/includes/simulation-system.md}
+\include{_data-science/includes/experiment-analyze-design.md}
+\include{_uq/includes/emulation.md}
+\notes{\include{_gp/includes/gpy-emulation.md}}
+\include{_software/includes/emukit-software.md}
+\include{_uq/includes/emukit-vision.md}
+\include{_uq/includes/emukit-playground.md}
+\notes{\include{_uq/includes/emukit-tutorial.md}}
+\notes{\include{_uq/includes/emukit-sensitivity-analysis.md}}
+\notes{\include{_uq/includes/catapult-sensitivity-analysis.md}}
+
+\thanks
+
+\references
+

_gpss/gaussian-distributions-to-processes.md

@@ -0,0 +1,43 @@
+---
+title: Gaussian Distributions to Processes
+abstract: >
+  In this sesson we go from the Gaussian distribution to the Gaussian process and in doing so we move from a finite system to an infinite system.
+featured_image: slides/diagrams/gp/two_point_sample008.svg
+week: 2
+---
+
+\include{_gpss/includes/gpss-notebook-setup.md}
+
+\newslide{Two Dimensional Gaussian Distribution}
+
+\include{_ml/includes/two-d-gaussian.md}
+
+\newslide{Multivariate Gaussian Properties}
+
+\include{_ml/includes/multivariate-gaussian-properties-summary.md}
+
+\newslide{Linear Gaussian Models}
+\slides{
+Gaussian processes are initially of interest because
+1. linear Gaussian models are easier to deal with 
+2. Even the parameters *within* the process can be handled, by considering a particular limit.
+}
+
+\include{_ml/includes/multivariate-gaussian-properties.md}
+\include{_ml/includes/linear-model-overview.md}
+
+\newslide{Distributions over Functions}
+
+\include{_gp/includes/gp-intro-very-short.md}
+
+\include{_gp/includes/gpdistfunc.md}
+
+\include{_kern/includes/computing-rbf-covariance.md}
+
+\include{_kern/includes/poly-covariance.md}
+\include{_kern/includes/rbf-basis-covariance.md}
+\include{_gp/includes/infinite-basis.md}
+
+\thanks
+
+\references

_gpss/latent-force-models.md

@@ -0,0 +1,12 @@
+---
+title: Latent Force Models
+week: 9
+abstract: 
+---
+
+
+
+\thanks
+
+\references
+

_gpss/lectures.csv

@@ -0,0 +1,14 @@
+lectureId
+uncertainty-and-modelling
+gaussian-distributions-to-processes
+covariance-functions
+optimizing-parameters
+multi-output-gps
+approximate-gps
+non-gaussian-likelihoods
+unsupervised-learning-with-gps
+latent-force-models
+bayesian-learning-gplvm
+deep-gps-I
+deep-gps-II
+emulation

_gpss/makefile

@@ -0,0 +1,3 @@
+BASE=emulation
+include ../make-talk-flags.mk
+include ../make-talk.mk

_gpss/multi-output-gps.md

@@ -0,0 +1,18 @@
+---
+title: Multi-output Gaussian Processes
+featured_image: slides/diagrams//kern/kronecker_KI.svg
+week: 5
+youtube: MkkeBmEZ8LE
+abstract: >
+  In this lecture we review multi-output Gaussian processes. Introducing them initially through a Kalman filter representation of a GP.
+---
+
+\include{_gpss/includes/gpss-notebook-setup.md}
+\include{_gp/includes/multi-output-gaussian-process.md}
+
+\thanks
+
+\references
+
+
+

_gpss/non-gaussian-likelihoods.md

@@ -0,0 +1,13 @@
+---
+title: Non Gaussian Likelihoods
+week: 7
+abstract:
+---
+
+\include{_gpss/includes/gpss-notebook-setup.md}
+
+\include{_gp/includes/non-gaussian-gps.md}
+
+\thanks
+
+\references