{ "_id": "6ad0b683b0a1b4b4099d735d", "Package": "RGAN", "Title": "Generative Adversarial Nets (GAN) in R", "Version": "0.2.2", "Authors@R": "person(given = \"Marcel\",\nfamily = \"Neunhoeffer\",\nrole = c(\"aut\", \"cre\"),\nemail = \"marcel.neunhoeffer@gmail.com\",\ncomment = c(ORCID = \"0000-0002-9137-5785\"))", "Description": "An easy way to get started with Generative Adversarial\nNets (GAN) in R. The GAN algorithm was initially described by\nGoodfellow et al. 2014\n.\nA GAN can be used to learn the joint distribution of complex\ndata by comparison. A GAN consists of two neural networks a\nGenerator and a Discriminator, where the two neural networks\nplay an adversarial minimax game. Built-in GAN models make the\ntraining of GANs in R possible in one line and make it easy to\nexperiment with different design choices (e.g. different\nnetwork architectures, value functions, optimizers). The\nbuilt-in GAN models work with tabular data (e.g. to produce\nsynthetic data) and image data. Methods to post-process the\noutput of GAN models to enhance the quality of samples are\navailable.", "License": "MIT + file LICENSE", "URL": "https://github.com/mneunhoe/RGAN", "BugReports": "https://github.com/mneunhoe/RGAN/issues", "Encoding": "UTF-8", "Roxygen": "list(markdown = TRUE)", "RoxygenNote": "7.3.3", "Config/testthat/edition": "3", "VignetteBuilder": "knitr", "Repository": "https://mneunhoe.r-universe.dev", "Date/Publication": "2026-09-15 13:27:53 UTC", "RemoteUrl": "https://github.com/mneunhoe/rgan", "RemoteRef": "HEAD", "RemoteSha": "fe64f21fd23b52de03b718277de8136adcaf2616", "NeedsCompilation": "no", "Packaged": { "Date": "2026-10-15 11:14:40 UTC", "User": "root" }, "Author": "Marcel Neunhoeffer [aut, cre] (ORCID:\n)", "Maintainer": "Marcel Neunhoeffer ", "_user": "mneunhoe", "_type": "src", "_file": "RGAN_0.2.2.tar.gz", "_fileid": "https://r2.ropensci.org/be2c138a1db3d2731e938fa7f7e171283262896141de147056788ff033762f7e", "_filesize": 450762, "_sha256": "be2c138a1db3d2731e938fa7f7e171283262896141de147056788ff033762f7e", "_expires": "2027-01-23T11:18:25.000Z", "_created": "2026-10-15T11:14:40.000Z", "_published": "2026-10-15T11:18:27.848Z", "_buildurl": "https://github.com/r-universe/mneunhoe/actions/runs/38725475208", "_status": "failure", "_host": "GitHub-Actions", "_jobs": [ { "job": 116391854583, "time": 184, "config": "linux-devel-x86_64", "r": "4.7.0", "check": "ERROR", "artifact": "11967907803" }, { "job": 116391854513, "time": 151, "config": "linux-release-x86_64", "r": "4.6.1", "check": "ERROR", "artifact": "11968342648" }, { "job": 116391854433, "time": 167, "config": "macos-oldrel-arm64", "r": "4.5.3", "check": "ERROR", "artifact": "11967802833" }, { "job": 116391854545, "time": 163, "config": "macos-release-arm64", "r": "4.6.1", "check": "ERROR", "artifact": "11967828011" }, { "job": 116390362340, "time": 221, "config": "source", "r": "4.6.1", "check": "ERROR", "artifact": "11967567949" }, { "job": 116391854481, "time": 95, "config": "wasm-release", "r": "4.6.0", "check": "OK", "artifact": "11967538269" }, { "job": 116391854741, "time": 153, "config": "windows-devel-x86_64", "r": "4.7.0", "check": "ERROR", "artifact": "11967238774" }, { "job": 116391854432, "time": 127, "config": "windows-oldrel-x86_64", "r": "4.5.3", "check": "ERROR", "artifact": "11967737991" }, { "job": 116391854443, "time": 163, "config": "windows-release-x86_64", "r": "4.6.1", "check": "ERROR", "artifact": "11967932757" } ], "_upstream": "https://github.com/mneunhoe/rgan", "_commit": { "id": "fe64f21fd23b52de03b718277de8136adcaf2616", "author": "Marcel Neunhoeffer ", "committer": "Marcel Neunhoeffer ", "message": "Fix the RDP accountant and the mechanism/accountant mismatch\n\nTwo independent defects, both affecting any 0.2.0 or 0.2.1 run.\n\nThe mechanism released a quantity the privacy analysis does not cover.\ndp_discriminator_step() divided the noisy gradient sum by the realised\nPoisson batch size. The Sampled Gaussian Mechanism bound covers the noisy\nsum, and dividing it by a constant is free post-processing, but the realised\nbatch size is a data-dependent count -- under add/remove-one adjacency a\ndataset with one more record can yield one more sampled record -- and it set\nboth the location and the scale of what was released. The divisor is now the\nexpected batch size, sampling_rate * n_samples, which is what Opacus and\nTensorFlow Privacy use. An empty Poisson draw also returned early without\ntaking an optimizer step, making \"no step\" an observable event of probability\n(1 - q)^n; the mechanism now releases noise and steps, via the new\ndp_noise_only_step().\n\nThe accountant understated epsilon by up to four orders of magnitude in the\nregime DP-SGD normally runs in: for a sampling rate below 0.01 the previous\nhand-written two-branch formula returned q^2 * alpha / (2 sigma^2), and the\nbound jumped by a factor of 10,000 across the branch boundary at q = 0.01.\nThe RDP of the Sampled Gaussian Mechanism is now the bound of Mironov,\nTalwar and Zhang (2019), evaluated in log space by the internal rdp_sgm().\nget_epsilon() uses the Canonne, Kamath and Steinke (2020) conversion, the\norder grid is fine between 1 and 11 and configurable through `orders`, and\nfractional orders use a series expansion (log_moment_sgm_frac()) that stops\non a bound for its whole remaining tail rather than on the size of the last\nterm. compute_max_steps() inherits the corrected accountant, so training now\nstops at the right step. calibrate_noise_multiplier() is one-sided: the\nmultiplier it returns always satisfies epsilon <= target_epsilon, and it\nerrors rather than silently returning max_noise when the target ca", "time": 1789478873 }, "_maintainer": { "name": "Marcel Neunhoeffer", "email": "marcel.neunhoeffer@gmail.com", "orcid": "0000-0002-9137-5785" }, "_distro": "resolute", "_registered": true, "_dependencies": [ { "package": "cli", "role": "Imports" }, { "package": "R6", "role": "Imports" }, { "package": "torch", "role": "Imports" }, { "package": "viridis", "role": "Imports" }, { "package": "opendp", "role": "Suggests" }, { "package": "rmarkdown", "role": "Suggests" }, { "package": "knitr", "role": "Suggests" }, { "package": "testthat", "version": ">= 3.0.0", "role": "Suggests" } ], "_owner": "mneunhoe", "_selfowned": true, "_usedby": 0, "_updates": [ { "week": "2026-02", "n": 40 }, { "week": "2026-38", "n": 2 } ], "_tags": [ { "name": "v0.2.0", "date": "2026-01-09" }, { "name": "v0.2.1", "date": "2026-09-15" }, { "name": "v0.2.2", "date": "2026-09-15" } ], "_stars": 19, "_userbio": { "uuid": 21174641, "type": "user", "name": "Marcel Neunhoeffer", "followers": 35 }, "_downloads": { "count": 230, "source": "https://cranlogs.r-pkg.org/downloads/total/last-month/RGAN" }, "_devurl": "https://github.com/mneunhoe/rgan", "_searchresults": 15, "_rbuild": "4.6.1", "_assets": [ "extra/citation.cff", "extra/citation.html", "extra/citation.json", "extra/citation.txt", "extra/contents.json", "extra/manual.pdf", "extra/NEWS.html", "extra/NEWS.txt", "extra/readme.html", "extra/readme.md", "extra/RGAN.html", "LICENSE" ], "_homeurl": "https://github.com/mneunhoe/rgan", "_realowner": "mneunhoe", "_cranurl": true, "_releases": [ { "version": "0.1.1", "date": "2022-03-29" } ], "_exports": [ "apply_post_gan_boosting", "compute_discriminator_scores", "data_transformer", "DCGAN_Discriminator", "DCGAN_Generator", "Discriminator", "dp_gan_trainer", "expert_sample_synthetic_data", "gan_trainer", "GAN_update_plot", "GAN_update_plot_image", "gan_update_step", "GAN_value_fct", "Generator", "gradient_penalty", "gumbel_softmax", "KLWGAN_value_fct", "load_gan", "plot_losses", "post_gan_boosting", "sample_synthetic_data", "sample_toydata", "save_gan", "TabularGenerator", "torch_rand_ab", "WGAN_value_fct", "WGAN_weight_clipper" ], "_help": [ { "page": "apply_post_gan_boosting", "title": "Apply Post-GAN Boosting to a Trained GAN", "topics": [ "apply_post_gan_boosting" ] }, { "page": "compute_discriminator_scores", "title": "Compute Discriminator Scores from Checkpointed Discriminators", "topics": [ "compute_discriminator_scores" ] }, { "page": "data_transformer", "title": "Data Transformer", "topics": [ "data_transformer" ] }, { "page": "DCGAN_Discriminator", "title": "DCGAN Discriminator", "topics": [ "DCGAN_Discriminator" ] }, { "page": "DCGAN_Generator", "title": "DCGAN Generator", "topics": [ "DCGAN_Generator" ] }, { "page": "Discriminator", "title": "Discriminator", "topics": [ "Discriminator" ] }, { "page": "dp_gan_trainer", "title": "dp_gan_trainer", "topics": [ "dp_gan_trainer" ] }, { "page": "expert_sample_synthetic_data", "title": "Sample Synthetic Data with explicit noise input", "topics": [ "expert_sample_synthetic_data" ] }, { "page": "gan_trainer", "title": "gan_trainer", "topics": [ "gan_trainer" ] }, { "page": "GAN_update_plot", "title": "GAN_update_plot", "topics": [ "GAN_update_plot" ] }, { "page": "GAN_update_plot_image", "title": "GAN_update_plot_image", "topics": [ "GAN_update_plot_image" ] }, { "page": "GAN_update_step", "title": "gan_update_step", "topics": [ "gan_update_step" ] }, { "page": "GAN_value_fct", "title": "GAN Value Function", "topics": [ "GAN_value_fct" ] }, { "page": "Generator", "title": "Generator", "topics": [ "Generator" ] }, { "page": "gradient_penalty", "title": "Gradient Penalty for WGAN-GP", "topics": [ "gradient_penalty" ] }, { "page": "gumbel_softmax", "title": "Gumbel-Softmax Sampling", "topics": [ "gumbel_softmax" ] }, { "page": "kl_fake", "title": "KL WGAN loss on fake examples", "topics": [ "kl_fake" ] }, { "page": "kl_gen", "title": "KL WGAN loss for Generator training", "topics": [ "kl_gen" ] }, { "page": "kl_real", "title": "KL WGAN loss on real examples", "topics": [ "kl_real" ] }, { "page": "KLWGAN_value_fct", "title": "KLWGAN Value Function", "topics": [ "KLWGAN_value_fct" ] }, { "page": "load_gan", "title": "Load a Trained GAN", "topics": [ "load_gan" ] }, { "page": "plot_losses", "title": "Plot GAN Training Losses", "topics": [ "plot_losses" ] }, { "page": "post_gan_boosting", "title": "Post-GAN Boosting", "topics": [ "post_gan_boosting" ] }, { "page": "print.trained_RGAN", "title": "Print Method for Trained RGAN Objects", "topics": [ "print.trained_RGAN" ] }, { "page": "sample_synthetic_data", "title": "Sample Synthetic Data from a trained RGAN", "topics": [ "sample_synthetic_data" ] }, { "page": "sample_toydata", "title": "Sample Toydata", "topics": [ "sample_toydata" ] }, { "page": "save_gan", "title": "Save a Trained GAN", "topics": [ "save_gan" ] }, { "page": "TabularGenerator", "title": "Tabular Generator with Gumbel-Softmax", "topics": [ "TabularGenerator" ] }, { "page": "torch_rand_ab", "title": "Uniform Random numbers between values a and b", "topics": [ "torch_rand_ab" ] }, { "page": "WGAN_value_fct", "title": "WGAN Value Function", "topics": [ "WGAN_value_fct" ] }, { "page": "WGAN_weight_clipper", "title": "WGAN Weight Clipper", "topics": [ "WGAN_weight_clipper" ] } ], "_readme": "https://github.com/mneunhoe/rgan/raw/HEAD/README.md", "_rundeps": [ "bit", "bit64", "callr", "cli", "coro", "cpp11", "desc", "farver", "ggplot2", "glue", "gridExtra", "gtable", "isoband", "jsonlite", "labeling", "lifecycle", "magrittr", "otel", "processx", "ps", "R6", "RColorBrewer", "Rcpp", "rlang", "S7", "safetensors", "scales", "torch", "vctrs", "viridis", "viridisLite", "withr" ], "_score": 4.45484486000851, "_indexed": true, "_nocasepkg": "rgan", "_universes": [ "mneunhoe" ], "_binaries": [ { "r": "4.7.0", "os": "linux", "version": "0.2.2", "date": "2026-10-15T11:17:15.000Z", "distro": "resolute", "commit": "fe64f21fd23b52de03b718277de8136adcaf2616", "fileid": "https://r2.ropensci.org/4d51e5e43028a40217bd350186b2664dfcc18ee8e412acc4469057c3052023b7", "status": "failure", "check": "ERROR", "buildurl": "https://github.com/r-universe/mneunhoe/actions/runs/38725475208" }, { "r": "4.6.1", "os": "linux", "version": "0.2.2", "date": "2026-10-15T11:16:55.000Z", "distro": "resolute", "commit": "fe64f21fd23b52de03b718277de8136adcaf2616", "fileid": "https://r2.ropensci.org/f11aa965055fe239e2e8a948a8d4b2ac30f050e9bfac8a9c7c7a28aa305c8d72", "status": "failure", "check": "ERROR", "buildurl": "https://github.com/r-universe/mneunhoe/actions/runs/38725475208" }, { "r": "4.5.3", "os": "mac", "version": "0.2.2", "date": "2026-10-15T11:17:17.000Z", "commit": "fe64f21fd23b52de03b718277de8136adcaf2616", "fileid": "https://r2.ropensci.org/75aef93787edda076a9cd89069df1d5849ae937cb1cbd014c2ec9e720654d300", "status": "failure", "check": "ERROR", "buildurl": "https://github.com/r-universe/mneunhoe/actions/runs/38725475208" }, { "r": "4.6.1", "os": "mac", "version": "0.2.2", "date": "2026-10-15T11:17:02.000Z", "commit": "fe64f21fd23b52de03b718277de8136adcaf2616", "fileid": "https://r2.ropensci.org/ed224527cd4ba662c1d67166ad940666d8c35e50b8379cfefcfda8f1a4ac9aee", "status": "failure", "check": "ERROR", "buildurl": "https://github.com/r-universe/mneunhoe/actions/runs/38725475208" }, { "r": "4.6.0", "os": "wasm", "version": "0.2.2", "date": "2026-10-15T11:16:36.000Z", "commit": "fe64f21fd23b52de03b718277de8136adcaf2616", "fileid": "https://r2.ropensci.org/0cb0562587b290d27767c834f45864e7a40879829e72bea7165b4fff1e1e891b", "status": "success", "buildurl": "https://github.com/r-universe/mneunhoe/actions/runs/38725475208" }, { "r": "4.7.0", "os": "win", "version": "0.2.2", "date": "2026-10-15T11:15:58.000Z", "commit": "fe64f21fd23b52de03b718277de8136adcaf2616", "fileid": "https://r2.ropensci.org/e85cc027c8788f0fdfbcde0f867b92724f57b50bd1134b1281d23bc0c4c74956", "status": "failure", "check": "ERROR", "buildurl": "https://github.com/r-universe/mneunhoe/actions/runs/38725475208" }, { "r": "4.5.3", "os": "win", "version": "0.2.2", "date": "2026-10-15T11:15:54.000Z", "commit": "fe64f21fd23b52de03b718277de8136adcaf2616", "fileid": "https://r2.ropensci.org/ef60946337ec52559749854c097e0a273f387e6f570da20d97510562cf36001d", "status": "failure", "check": "ERROR", "buildurl": "https://github.com/r-universe/mneunhoe/actions/runs/38725475208" }, { "r": "4.6.1", "os": "win", "version": "0.2.2", "date": "2026-10-15T11:16:05.000Z", "commit": "fe64f21fd23b52de03b718277de8136adcaf2616", "fileid": "https://r2.ropensci.org/6473acc00fe029f9d379ef32d8112d42581b7807380abe7fefb1649137e0144c", "status": "failure", "check": "ERROR", "buildurl": "https://github.com/r-universe/mneunhoe/actions/runs/38725475208" } ] }