Last updated: 2021-04-09

Checks: 7 0

Knit directory: fastTopics-experiments/analysis/

This reproducible R Markdown analysis was created with workflowr (version 1.6.2.9000). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.


Great! Since the R Markdown file has been committed to the Git repository, you know the exact version of the code that produced these results.

Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.

The command set.seed(1) was run prior to running the code in the R Markdown file. Setting a seed ensures that any results that rely on randomness, e.g. subsampling or permutations, are reproducible.

Great job! Recording the operating system, R version, and package versions is critical for reproducibility.

Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.

Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.

Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility.

The results in this page were generated with repository version 0ab6c04. See the Past versions tab to see a history of the changes made to the R Markdown and HTML files.

Note that you need to be careful to ensure that all relevant files for the analysis have been committed to Git prior to generating the results (you can use wflow_publish or wflow_git_commit). workflowr only checks the R Markdown file, but you know if there are other scripts or data files that it depends on. Below is the status of the Git repository when the results were generated:


Ignored files:
    Ignored:    data/20news-bydate/
    Ignored:    data/droplet.RData
    Ignored:    data/nips_1-17.mat
    Ignored:    data/pbmc_68k.RData
    Ignored:    output/droplet/fits-droplet.RData
    Ignored:    output/newsgroups/fits-newsgroups.RData
    Ignored:    output/nips/fits-nips.RData
    Ignored:    output/pbmc68k/fits-pbmc68k.RData

Untracked files:
    Untracked:  .DS_Store
    Untracked:  plots/

Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.


These are the previous versions of the repository in which changes were made to the R Markdown (analysis/assess_fits_newsgroups.Rmd) and HTML (docs/assess_fits_newsgroups.html) files. If you’ve configured a remote Git repository (see ?wflow_git_remote), click on the hyperlinks in the table below to view the files as they were in that past version.

File Version Author Date Message
Rmd 0ab6c04 Peter Carbonetto 2021-04-09 workflowr::wflow_publish(“assess_fits_newsgroups.Rmd”)
html 42cb17e Peter Carbonetto 2021-04-09 Adjusted scatterplots in assess_fits_newsgroups analysis.
Rmd 8f157b4 Peter Carbonetto 2021-04-09 workflowr::wflow_publish(“assess_fits_newsgroups.Rmd”)
html d1dadf8 Peter Carbonetto 2021-04-09 Added loadings scatterplots to aassess_fits_newsgroups analysis.
Rmd b37146b Peter Carbonetto 2021-04-09 workflowr::wflow_publish(“assess_fits_newsgroups.Rmd”)
html c679bdc Peter Carbonetto 2021-04-07 Small fix to assess_fits_newsgroups output.
Rmd db4870b Peter Carbonetto 2021-04-07 workflowr::wflow_publish(“assess_fits_newsgroups.Rmd”)
html ca4f229 Peter Carbonetto 2021-04-07 Added max KKT residual progress plots to assess_fits_newsgroups
Rmd b916251 Peter Carbonetto 2021-04-07 workflowr::wflow_publish(“assess_fits_newsgroups.Rmd”)
Rmd 7827775 Peter Carbonetto 2021-04-06 Added plots-for-paper-1 code chunk to assess_fits_newsgroups.Rmd.
Rmd 0f88216 Peter Carbonetto 2021-04-06 A couple of small changes.
html cb0de01 Peter Carbonetto 2021-04-06 Added loglik progress plots to assess_fits_newsgroups analysis.
Rmd 720e225 Peter Carbonetto 2021-04-06 workflowr::wflow_publish(“assess_fits_newsgroups.Rmd”)
html 51a6dd3 Peter Carbonetto 2021-04-06 First build of the assess_fits_newsgroups analysis.
Rmd b57553c Peter Carbonetto 2021-04-06 workflowr::wflow_publish(“assess_fits_newsgroups.Rmd”)

Here we compare the quality of the fits obtained from the different updates (EM and SCD, with and without extrapolation), and with different numbers of topics, \(K\).

Load the packages used in the analysis below, as well as some additional functions for creating the plots.

library(fastTopics)
library(ggplot2)
library(cowplot)
set.seed(1)

Load the results of running fit_poisson_nmf on the “20 newsgroups” data, with different algorithms, and for various settings of \(K\).

load("../output/newsgroups/fits-newsgroups.RData")
fits <- lapply(fits,poisson2multinom)

This plot shows the improvement in the log-likelihood as the rank, \(K\), is increased. The log-likelihoods are shown relative to the log-likelihood at \(K = 2\).

plot_loglik_vs_rank(fits) +
  theme_cowplot(font_size = 12)

Version Author Date
cb0de01 Peter Carbonetto 2021-04-06
51a6dd3 Peter Carbonetto 2021-04-06

The next set of plots shows the improvement in the fit over time, for \(K\) from 2 to 12, using EM or SCD, with and without extrapolation. The quality of the fit is measured by the log-likelihood relative to the best log-likelihood that was identified among all methods compared.

prune_prefit_iters <- function (fit) {
  n <- nrow(fit$progress)
  fit$progress <- fit$progress[1000:n,]
  fit$progress <- transform(fit$progress,timing = timing/60^2)
  return(fit)
}
create_progress_plot <- function (fits, k, y = "loglik")
  plot_progress(fits,y = y,add.point.every = 100,shapes = 21,
                colors = c("dodgerblue","red","dodgerblue","red"),
                fills = c("dodgerblue","red","white","white")) +
  scale_y_continuous(trans = "log10",breaks = 10^seq(-8,8)) +
  guides(color = "none",fill = "none",size = "none",
         shape = "none",linetype = "none") +
  labs(x = "runtime (h)",title = paste("K =",k)) +
  theme_cowplot(font_size = 10) +
  theme(plot.title = element_text(size = 10,face = "plain"))
fits <- lapply(fits,prune_prefit_iters)
p <- vector("list",12)
for (i in 2:12)
  p[[i]] <- create_progress_plot(fits[dat$k == i],i)
plot_grid(p[[2]],p[[3]],p[[4]],p[[5]],
          p[[6]],p[[7]],p[[8]],p[[9]],
          p[[10]],p[[11]],p[[12]],
          nrow = 3,ncol = 4)

Version Author Date
cb0de01 Peter Carbonetto 2021-04-06

These plots shows the evolution of the KKT residuals over time.

for (i in 2:12)
  p[[i]] <- create_progress_plot(fits[dat$k == i],i,y = "res")
plot_grid(p[[2]],p[[3]],p[[4]],p[[5]],
          p[[6]],p[[7]],p[[8]],p[[9]],
          p[[10]],p[[11]],p[[12]],
          nrow = 3,ncol = 4)

Version Author Date
ca4f229 Peter Carbonetto 2021-04-07

An example in which the EM and (extrapolated) CD estimates largely agree:

topic_colors <- c("#a6cee3","#1f78b4","#b2df8a","#33a02c","#fb9a99",
                  "#e31a1c","#fdbf6f","#ff7f00","#cab2d6","#6a3d9a")
fit1 <- fits[["fit-newsgroups-em-k=9"]]
fit2 <- fits[["fit-newsgroups-scd-ex-k=9"]]
n    <- nrow(fit1$L)
pdat <- data.frame(x = as.vector(fit1$L),
                   y = as.vector(fit2$L),
                   k = factor(rep(1:9,each = n)))
pdat <- pdat[sample(9*n),]
p1 <- ggplot(pdat,aes(x = x,y = y,fill = k)) +
  geom_point(color = "white",shape = 21,size = 2) +
  geom_abline(color = "black",linetype = "dotted") +
  scale_fill_manual(values = topic_colors) +
  labs(x = "EM estimate",y = "extrapolated CD estimate") +
  theme_cowplot(font_size = 12)
print(p1)

Version Author Date
42cb17e Peter Carbonetto 2021-04-09
d1dadf8 Peter Carbonetto 2021-04-09

An example in which the EM and (extrapolated) CD estimates greatly differ:

fit1 <- fits[["fit-newsgroups-em-k=10"]]
fit2 <- fits[["fit-newsgroups-scd-ex-k=10"]]
n    <- nrow(fit1$L)
pdat <- data.frame(x = as.vector(fit1$L),
                   y = as.vector(fit2$L),
                   k = factor(rep(1:10,each = n)))
pdat <- pdat[sample(10*n),]
p2 <- ggplot(pdat,aes(x = x,y = y,fill = k)) +
  geom_point(color = "white",shape = 21,size = 2) +
  geom_abline(color = "black",linetype = "dotted") +
  scale_fill_manual(values = topic_colors) +
  labs(x = "EM estimate",y = "extrapolated CD estimate") +
  theme_cowplot(font_size = 12)
print(p2)

Version Author Date
42cb17e Peter Carbonetto 2021-04-09
d1dadf8 Peter Carbonetto 2021-04-09

sessionInfo()
# R version 3.6.2 (2019-12-12)
# Platform: x86_64-apple-darwin15.6.0 (64-bit)
# Running under: macOS Catalina 10.15.7
# 
# Matrix products: default
# BLAS:   /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRblas.0.dylib
# LAPACK: /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRlapack.dylib
# 
# locale:
# [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
# 
# attached base packages:
# [1] stats     graphics  grDevices utils     datasets  methods   base     
# 
# other attached packages:
# [1] cowplot_1.0.0     ggplot2_3.3.0     fastTopics_0.5-24
# 
# loaded via a namespace (and not attached):
#  [1] ggrepel_0.9.0        Rcpp_1.0.5           invgamma_1.1        
#  [4] lattice_0.20-38      tidyr_1.0.0          prettyunits_1.1.1   
#  [7] assertthat_0.2.1     zeallot_0.1.0        rprojroot_1.3-2     
# [10] digest_0.6.23        truncnorm_1.0-8      R6_2.4.1            
# [13] backports_1.1.5      MatrixModels_0.4-1   evaluate_0.14       
# [16] coda_0.19-3          httr_1.4.2           pillar_1.4.3        
# [19] progress_1.2.2       rlang_0.4.5          lazyeval_0.2.2      
# [22] data.table_1.12.8    irlba_2.3.3          SparseM_1.78        
# [25] whisker_0.4          Matrix_1.2-18        rmarkdown_2.3       
# [28] labeling_0.3         Rtsne_0.15           stringr_1.4.0       
# [31] htmlwidgets_1.5.1    munsell_0.5.0        mixsqp_0.3-44       
# [34] compiler_3.6.2       httpuv_1.5.2         xfun_0.11           
# [37] pkgconfig_2.0.3      SQUAREM_2017.10-1    mcmc_0.9-6          
# [40] htmltools_0.4.0      tidyselect_0.2.5     tibble_2.1.3        
# [43] workflowr_1.6.2.9000 quadprog_1.5-8       viridisLite_0.3.0   
# [46] withr_2.1.2          crayon_1.3.4         dplyr_0.8.3         
# [49] later_1.0.0          MASS_7.3-51.4        grid_3.6.2          
# [52] jsonlite_1.6         gtable_0.3.0         lifecycle_0.1.0     
# [55] git2r_0.26.1         magrittr_1.5         scales_1.1.0        
# [58] RcppParallel_4.4.2   stringi_1.4.3        farver_2.0.1        
# [61] fs_1.3.1             promises_1.1.0       vctrs_0.2.1         
# [64] tools_3.6.2          glue_1.3.1           purrr_0.3.3         
# [67] hms_0.5.2            yaml_2.2.0           colorspace_1.4-1    
# [70] ashr_2.2-51          plotly_4.9.2         knitr_1.26          
# [73] quantreg_5.54        MCMCpack_1.4-5