pluto notebooks

Codes/Julia/Part-3/gan/gan.jl

@@ -1,13 +1,29 @@
+### A Pluto.jl notebook ###
+# v0.20.3
+
+using Markdown
+using InteractiveUtils
+
+# ╔═╡ e085fb51-8865-4b12-80e9-e5a0ebcdb5da
 ####################################
 #= Generative Adverserial Network =#
 ####################################
 # `versioninfo()` -> 1.11.1
 
 using Flux # v0.14.25
+
+# ╔═╡ 01ca737b-829c-4697-aa5d-6b907cd4c3fd
 using Images: Gray
+
+# ╔═╡ 189414a9-03b9-42d7-93ca-15ed58a20ce8
 using ProgressMeter
 
 ## Generator: noise vector -> synthetic sample.
+
+# ╔═╡ 281920a0-6d99-4896-a8dd-3de46f1182fb
+using Plots
+
+# ╔═╡ eb7aa3c4-4340-4ed4-918e-2974002f613d
 function generator(; latent_dim=16, img_shape=(28,28,1,1))
     return Chain(
         Dense(latent_dim, 128, relu),
@@ -18,6 +34,8 @@ function generator(; latent_dim=16, img_shape=(28,28,1,1))
 end
 
 ## Discriminator : sample -> score indicating the probability that the sample is real.
+
+# ╔═╡ 69b4af29-b689-4b20-a773-78742bb68281
 function discriminator(; img_shape=(28,28,1,1))
     return Chain(
         x -> reshape(x, :, size(x, 4)),
@@ -28,9 +46,13 @@ function discriminator(; img_shape=(28,28,1,1))
 end
 
 ## Loss functions
+
+# ╔═╡ 11471f2f-df52-4764-b511-9fe1a8848432
 bce_loss(y_true, y_pred) = Flux.logitbinarycrossentropy(y_pred, y_true)
 
 ## Training function
+
+# ╔═╡ e866f93b-221d-4444-9fff-28e7c20028d6
 function train_gan(gen, disc, gen_opt, disc_opt; n_epochs=16, latent_dim=16)
     @showprogress for epoch in 1:n_epochs
 
@@ -56,28 +78,66 @@ function train_gan(gen, disc, gen_opt, disc_opt; n_epochs=16, latent_dim=16)
 end
 
 ## Setup the GAN
+
+# ╔═╡ 6c3f364c-0d0d-4a34-832e-f3653f15ea54
 gen = generator()
+
+# ╔═╡ 3b43ad53-e53e-4849-82db-0441ed8daa3b
 disc = discriminator()
 
+# ╔═╡ 4e20b5ff-8b80-4850-9d59-4058a3c16a24
 gen_opt = Adam(0.001)
+
+# ╔═╡ 2fc91e4e-b736-4b24-856f-e997cea5e497
 disc_opt = Adam(0.0002)
 
 ## Train the GAN
+
+# ╔═╡ 0362230d-17e3-424e-b923-1e3b1c5359b5
 train_gan(gen, disc, gen_opt, disc_opt)
 
 ## Generate and plot some images
+
+# ╔═╡ b0bfaa5f-4e77-4b52-9970-dc0adb47bddb
 latent_dim = 16
+
+# ╔═╡ c7707565-b11c-45d1-a670-43b546d447d4
 noise = randn(Float32, latent_dim, 16)
-generated_images = [ gen(noise[:, i]) for i in 1:16 ]
 
-using Plots
+# ╔═╡ c685204f-dd86-48ad-96d6-3d94ac30321a
+generated_images = [ gen(noise[:, i]) for i in 1:16 ]
 
+# ╔═╡ e50e02cb-8601-4b8b-adad-e9c8f39e7ce5
 plot_images = [ plot(Gray.(generated_images[i])[:,:,1,1]) for i in 1:16 ]
+
+# ╔═╡ 62f695e1-6cba-49df-b722-99ccfa5f8c73
 titles = reshape([string(i) for i in 1:16], 1, :);
 
+# ╔═╡ 036ebfaf-e14e-4558-b795-d04923922489
 plot(
     plot_images...,
     layout = (4, 4), 
     title = titles, titleloc=:right, titlefont=font(8),
     size = (800, 800)
 )
+
+# ╔═╡ Cell order:
+# ╠═e085fb51-8865-4b12-80e9-e5a0ebcdb5da
+# ╠═01ca737b-829c-4697-aa5d-6b907cd4c3fd
+# ╠═189414a9-03b9-42d7-93ca-15ed58a20ce8
+# ╠═eb7aa3c4-4340-4ed4-918e-2974002f613d
+# ╠═69b4af29-b689-4b20-a773-78742bb68281
+# ╠═11471f2f-df52-4764-b511-9fe1a8848432
+# ╠═e866f93b-221d-4444-9fff-28e7c20028d6
+# ╠═6c3f364c-0d0d-4a34-832e-f3653f15ea54
+# ╠═3b43ad53-e53e-4849-82db-0441ed8daa3b
+# ╠═4e20b5ff-8b80-4850-9d59-4058a3c16a24
+# ╠═2fc91e4e-b736-4b24-856f-e997cea5e497
+# ╠═0362230d-17e3-424e-b923-1e3b1c5359b5
+# ╠═b0bfaa5f-4e77-4b52-9970-dc0adb47bddb
+# ╠═c7707565-b11c-45d1-a670-43b546d447d4
+# ╠═c685204f-dd86-48ad-96d6-3d94ac30321a
+# ╠═281920a0-6d99-4896-a8dd-3de46f1182fb
+# ╠═e50e02cb-8601-4b8b-adad-e9c8f39e7ce5
+# ╠═62f695e1-6cba-49df-b722-99ccfa5f8c73
+# ╠═036ebfaf-e14e-4558-b795-d04923922489

Codes/Julia/Part-3/nlp/nlp.jl

@@ -1,36 +1,84 @@
+### A Pluto.jl notebook ###
+# v0.20.3
+
+using Markdown
+using InteractiveUtils
+
+# ╔═╡ d8292e75-d298-4e8e-b7bb-acc848e01b4a
 #################################
 #= Natural Language Processing =#
 #################################
 # `versioninfo()` -> 1.11.1
 
 using Markdown
+
+# ╔═╡ bf1da4bc-48f4-4042-a7ab-e042bbd32d41
 using TextAnalysis
 
+# ╔═╡ c1fb388b-9c59-429b-b036-9bc4395ab5af
+using Embeddings
+
+# ╔═╡ fdd4f0b1-1cae-467a-9fbf-6fbbe1c35c1c
+using LinearAlgebra
+
+# ╔═╡ f336e7e2-99ae-4b91-8f93-0607eede9773
 txt = "The quick brown fox is jumping over the lazy dog" # Pangram [modif.]
 
+# ╔═╡ 40388d9d-16db-4e3c-9384-bc4638ba3663
 md"Create a `Corpus` using `txt`"
+
+# ╔═╡ 4febd42e-b2c5-46e2-af27-3e81f825fdab
 crps = Corpus([StringDocument(txt)])
+
+# ╔═╡ f38b2eee-ee30-4a4f-8500-2c076000ff95
 lexicon(crps)
+
+# ╔═╡ beb39317-c83c-4c0e-ac1d-ab64cde4307d
 update_lexicon!(crps)
+
+# ╔═╡ 2c3d4f5f-a0fd-42e0-ba1d-ba390c6da171
 lexicon(crps)
+
+# ╔═╡ 59170576-05f3-4edc-b26b-8b48152a4477
 lexical_frequency(crps, "fox")
 
+# ╔═╡ 2388744f-24b6-4ac1-8673-0feb223bfe37
 md"Create a `StringDocument` using `txt`"
+
+# ╔═╡ 811cdd9b-7cc8-444b-a621-479f785af38c
 sd = StringDocument(txt)
+
+# ╔═╡ 75322fd2-a075-4c89-b219-95b86753457a
 md"Get a smaller set of words `text(sd)`"
+
+# ╔═╡ ce5adafd-e6d3-4571-a289-c31940e3ebfc
 prepare!(sd, strip_articles | strip_numbers | strip_punctuation | strip_case | strip_whitespace)
+
+# ╔═╡ 356963d7-9999-43bd-a21b-49b3056a172e
 stem!(sd)
 
+# ╔═╡ 560a06c5-9078-49ad-9da6-9838656a0948
 md"Get the tokens of `sd`"
+
+# ╔═╡ 5b336441-b1f6-4ce1-a979-64e363bbc5fc
 the_tokens = tokens(sd)
 
+# ╔═╡ 25388144-78af-4f7d-9f24-7c190141e61f
 md"Get the stemmed tokens of `sd`"
+
+# ╔═╡ 6c5b1442-6297-44cb-8f97-200572f5dfb9
 stemmer = Stemmer("english")
+
+# ╔═╡ 1cf0c275-c1cc-4453-a5b8-0cfcd541faf8
 stemmed_tokens = stem(stemmer, the_tokens)
 
+# ╔═╡ 8295d84f-8aa6-4ad9-be45-fcc9aa3ae087
 println("Original tokens: ", the_tokens)
+
+# ╔═╡ 7a155470-e751-4137-9c87-b047bc2d44c6
 println("Stemmed tokens: ", stemmed_tokens)
 
+# ╔═╡ e1566e3f-4c08-4485-b345-9cefda319e45
 md"**Part-of-speech tags**"
 
 #= 
@@ -56,57 +104,177 @@ pos = PoSTagger()
 pos(crps)
 =#
 
+# ╔═╡ 38639182-7aaa-44e7-b11f-60aebe6a54d8
 md"**Word embeddings**"
-using Embeddings
+
+# ╔═╡ 88d15a75-7d44-405c-a8e4-19c68708d0a5
 embtab = load_embeddings(GloVe{:en}, max_vocab_size=5)
 
+# ╔═╡ eb53e251-d8e4-4b66-86fd-279fbbabe7e4
 embtab.vocab
+
+# ╔═╡ d1f06dd1-2078-4b4e-a33a-40191b96a1bb
 embtab.embeddings
 
+# ╔═╡ eadd26c5-d4ea-4408-ab78-5bc1cce4cd12
 glove = load_embeddings(GloVe{:en}, 3, max_vocab_size=10_000)
+
+# ╔═╡ d77c50ba-84af-419e-b127-a2deb9a3859c
 const word_to_index = Dict(word => ii for (ii,word) in enumerate(glove.vocab))
+
+# ╔═╡ 2be17f28-55e6-43c2-bece-77702ad72923
 function get_word_vector(word)
     idx = word_to_index[word]
     return glove.embeddings[:, idx]
 end
 
-using LinearAlgebra
+# ╔═╡ 6e1668bb-43af-49be-9d9d-623a101dba03
 function cosine_similarity(v1::Vector{Float32}, v2::Vector{Float32})
     return *(v1', v2) / *(norm(v1), norm(v2))
 end
 
+# ╔═╡ 937e7d17-34cc-4e59-b2bf-98cf9e153bb5
 md"_e.g. - \"king\" - \"man\" + \"woman\" ≈ \"queen\"_"
+
+# ╔═╡ 37ee5b81-3444-49ab-8732-530004b00035
 king = get_word_vector("king")
+
+# ╔═╡ 1f128c17-8da6-4752-939b-25e75e8a8725
 queen = get_word_vector("queen")
+
+# ╔═╡ 4cc8fb94-43a6-44cd-8754-bb1ef0190573
 man = get_word_vector("man")
+
+# ╔═╡ 9c71279a-6ebd-4c86-82e2-c0af535f33b9
 woman = get_word_vector("woman")
 
+# ╔═╡ accb5901-6b0b-41f6-8d7c-3dd5a33d571a
 cosine_similarity(king - man + woman, queen)
 
+# ╔═╡ 07d3ff60-b7e9-4d99-9a73-601e421e1b64
 md"_e.g. - \"Madrid\" - \"Spain\" + \"France\" ≈ \"Paris\"_"
+
+# ╔═╡ 6bb3a773-c5c4-4d06-a28e-03737ce02797
 Madrid = get_word_vector("madrid")
+
+# ╔═╡ 3b8f4450-4248-4b60-a61c-5411c311d5d6
 Spain = get_word_vector("spain")
+
+# ╔═╡ cb9b1c25-9407-4ada-98eb-2b2462479729
 France = get_word_vector("france")
+
+# ╔═╡ 17e527a4-b80f-43bd-94a4-3e0411bcc642
 Paris = get_word_vector("paris")
 
+# ╔═╡ dca54445-4704-4726-99ce-3f60431459ba
 cosine_similarity(Madrid - Spain + France, Paris)
 
+# ╔═╡ 74d57c25-34f0-4f60-9ba9-6d6b507b61f2
 md"**Text classification**"
+
+# ╔═╡ 65f6b972-a4fe-4b62-8dc8-80f564de8fbf
 md"https://github.com/JuliaText/TextAnalysis.jl/blob/master/docs/src/classify.md"
-m = NaiveBayesClassifier([:legal, :financial])
+
+# ╔═╡ 1653799c-175e-4024-b9a8-ca08446cdc79
+md"**Semantic analysis**"
+
+# ╔═╡ 3af49352-fd14-4181-8f2e-5f5aeaf6aff3
 fit!(m, "this is financial doc", :financial)
+
+# ╔═╡ 8c93b998-3e9a-4c6f-8ae2-14d1292942ce
 fit!(m, "this is legal doc", :legal)
-predict(m, "this should be predicted as a legal document")
 
-md"**Semantic analysis**"
-m = DocumentTermMatrix(crps)
+# ╔═╡ 33272333-27b7-4ed1-918c-fac97e0e18fd
+predict(m, "this should be predicted as a legal document")
 
+# ╔═╡ 3a6c37d3-7ea9-4d5b-b486-e3c48be5e076
 md"*Latent Semantic Analysis*"
+
+# ╔═╡ 2ded08e4-b048-41e7-a02c-8733e02ef928
 lsa(m)
 
+# ╔═╡ c011919b-0378-4116-a085-21ee18231104
 md"*Latent Dirichlet Allocation*"
+
+# ╔═╡ 4202f7f4-499b-44df-8d79-4779609c93fe
 k = 2              # number of topics
+
+# ╔═╡ cddae5b7-8f84-4407-9a09-9f7d19aac071
 iterations = 1000  # number of Gibbs sampling iterations
+
+# ╔═╡ 46c868e8-23f8-4c22-9742-2126c5b7e15e
 α = 0.1            # hyper parameter
+
+# ╔═╡ e4ea104e-0bd9-4d54-a4f3-800ad6a3dde9
 β  = 0.1           # hyper parameter
-ϕ, θ  = lda(m, k, iterations, α, β) # 
+
+# ╔═╡ 92520517-1ba9-4436-9472-91699cb2b56c
+ϕ, θ  = lda(m, k, iterations, α, β) #
+
+# ╔═╡ 11fe766c-bb25-478a-bd94-e786f98d21aa
+m = NaiveBayesClassifier([:legal, :financial])
+
+# ╔═╡ 9120c455-745e-498a-acc7-827da3f8c9ae
+m = DocumentTermMatrix(crps)
+
+# ╔═╡ Cell order:
+# ╠═d8292e75-d298-4e8e-b7bb-acc848e01b4a
+# ╠═bf1da4bc-48f4-4042-a7ab-e042bbd32d41
+# ╠═f336e7e2-99ae-4b91-8f93-0607eede9773
+# ╠═40388d9d-16db-4e3c-9384-bc4638ba3663
+# ╠═4febd42e-b2c5-46e2-af27-3e81f825fdab
+# ╠═f38b2eee-ee30-4a4f-8500-2c076000ff95
+# ╠═beb39317-c83c-4c0e-ac1d-ab64cde4307d
+# ╠═2c3d4f5f-a0fd-42e0-ba1d-ba390c6da171
+# ╠═59170576-05f3-4edc-b26b-8b48152a4477
+# ╠═2388744f-24b6-4ac1-8673-0feb223bfe37
+# ╠═811cdd9b-7cc8-444b-a621-479f785af38c
+# ╠═75322fd2-a075-4c89-b219-95b86753457a
+# ╠═ce5adafd-e6d3-4571-a289-c31940e3ebfc
+# ╠═356963d7-9999-43bd-a21b-49b3056a172e
+# ╠═560a06c5-9078-49ad-9da6-9838656a0948
+# ╠═5b336441-b1f6-4ce1-a979-64e363bbc5fc
+# ╠═25388144-78af-4f7d-9f24-7c190141e61f
+# ╠═6c5b1442-6297-44cb-8f97-200572f5dfb9
+# ╠═1cf0c275-c1cc-4453-a5b8-0cfcd541faf8
+# ╠═8295d84f-8aa6-4ad9-be45-fcc9aa3ae087
+# ╠═7a155470-e751-4137-9c87-b047bc2d44c6
+# ╠═e1566e3f-4c08-4485-b345-9cefda319e45
+# ╠═38639182-7aaa-44e7-b11f-60aebe6a54d8
+# ╠═c1fb388b-9c59-429b-b036-9bc4395ab5af
+# ╠═88d15a75-7d44-405c-a8e4-19c68708d0a5
+# ╠═eb53e251-d8e4-4b66-86fd-279fbbabe7e4
+# ╠═d1f06dd1-2078-4b4e-a33a-40191b96a1bb
+# ╠═eadd26c5-d4ea-4408-ab78-5bc1cce4cd12
+# ╠═d77c50ba-84af-419e-b127-a2deb9a3859c
+# ╠═2be17f28-55e6-43c2-bece-77702ad72923
+# ╠═fdd4f0b1-1cae-467a-9fbf-6fbbe1c35c1c
+# ╠═6e1668bb-43af-49be-9d9d-623a101dba03
+# ╠═937e7d17-34cc-4e59-b2bf-98cf9e153bb5
+# ╠═37ee5b81-3444-49ab-8732-530004b00035
+# ╠═1f128c17-8da6-4752-939b-25e75e8a8725
+# ╠═4cc8fb94-43a6-44cd-8754-bb1ef0190573
+# ╠═9c71279a-6ebd-4c86-82e2-c0af535f33b9
+# ╠═accb5901-6b0b-41f6-8d7c-3dd5a33d571a
+# ╠═07d3ff60-b7e9-4d99-9a73-601e421e1b64
+# ╠═6bb3a773-c5c4-4d06-a28e-03737ce02797
+# ╠═3b8f4450-4248-4b60-a61c-5411c311d5d6
+# ╠═cb9b1c25-9407-4ada-98eb-2b2462479729
+# ╠═17e527a4-b80f-43bd-94a4-3e0411bcc642
+# ╠═dca54445-4704-4726-99ce-3f60431459ba
+# ╠═74d57c25-34f0-4f60-9ba9-6d6b507b61f2
+# ╠═65f6b972-a4fe-4b62-8dc8-80f564de8fbf
+# ╠═11fe766c-bb25-478a-bd94-e786f98d21aa
+# ╠═3af49352-fd14-4181-8f2e-5f5aeaf6aff3
+# ╠═8c93b998-3e9a-4c6f-8ae2-14d1292942ce
+# ╠═33272333-27b7-4ed1-918c-fac97e0e18fd
+# ╠═1653799c-175e-4024-b9a8-ca08446cdc79
+# ╠═9120c455-745e-498a-acc7-827da3f8c9ae
+# ╠═3a6c37d3-7ea9-4d5b-b486-e3c48be5e076
+# ╠═2ded08e4-b048-41e7-a02c-8733e02ef928
+# ╠═c011919b-0378-4116-a085-21ee18231104
+# ╠═4202f7f4-499b-44df-8d79-4779609c93fe
+# ╠═cddae5b7-8f84-4407-9a09-9f7d19aac071
+# ╠═46c868e8-23f8-4c22-9742-2126c5b7e15e
+# ╠═e4ea104e-0bd9-4d54-a4f3-800ad6a3dde9
+# ╠═92520517-1ba9-4436-9472-91699cb2b56c