2017/10/30

Python - numpy 的使用方法

markdown [NumPy Reference](https://docs.scipy.org/doc/numpy-1.13.0/reference/index.html) 數列生成 ``` >>> np.zeros(5) #array([ 0., 0., 0., 0., 0.]) >>> np.ones(5) #array([ 1., 1., 1., 1., 1.]) >>> np.arange(5) #array([0, 1, 2, 3, 4]) ``` [更多的數列生成](https://docs.scipy.org/doc/numpy-1.13.0/reference/routines.array-creation.html) 數列變形 ``` >>> np.arange(12) #array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]) >>> np.arange(12).reshape(2,6) #array([[ 0, 1, 2, 3, 4, 5], # [ 6, 7, 8, 9, 10, 11]]) >>> np.arange(12).reshape(3,4) #array([[ 0, 1, 2, 3], # [ 4, 5, 6, 7], # [ 8, 9, 10, 11]]) >>> np.arange(12).reshape(4,3) #array([[ 0, 1, 2], # [ 3, 4, 5], # [ 6, 7, 8], # [ 9, 10, 11]]) >>> np.arange(12).reshape(6,2) #array([[ 0, 1], # [ 2, 3], # [ 4, 5], # [ 6, 7], # [ 8, 9], # [10, 11]]) >>> np.arange(12).reshape(12,1).flatten() #array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]) ``` 查看數列形狀 ``` >>> np.arange(12).reshape(1,2,3,2) #array([[[[ 0, 1], # [ 2, 3], # [ 4, 5]], # # [[ 6, 7], # [ 8, 9], # [10, 11]]]]) >>> np.arange(12).reshape(1,2,3,2).shape #(1, 2, 3, 2) ``` [更多的數列操作](https://docs.scipy.org/doc/numpy-1.13.0/reference/routines.array-manipulation.html)

2017/10/26

Rails - Apartment 的使用方法

markdown #簡介 Apartment : [https://github.com/influitive/apartment](https://github.com/influitive/apartment) 這個套件的功能是讓你在 rails 可以透過切換 db schema,來存取不同的資料。 在 postgresql 中,預設的 schema 名稱為 public ,而所有的 table 都在 schema 'public' 下。 你可以簡單地把 schema 看成 namespace,在 apartment 中,這被稱為是 tenant。 #安裝步驟 請參考 : [https://github.com/influitive/apartment](https://github.com/influitive/apartment) #使用方法 ## 在 rails c 環境下 新增 tenant ``` Apartment::Tenant.create('abc') ``` 這個指令會在 schema 'abc' 下複製 public 上的所有表格。所以 tenant 就是具有相同表格的 schema。 在以下的示範,看到 abc 的地方,代表的是要填入你的 schema name。 顯示目前所在的 tenant ``` Apartment::Tenant.current # "public" ``` 切換 tenant ``` Apartment::Tenant.switch!('abc') Apartment::Tenant.current # "abc" Apartment::Tenant.switch!('public') Apartment::Tenant.current # "public" Apartment::Tenant.switch('abc') do Apartment::Tenant.current # "abc" end Apartment::Tenant.current # "public" ``` 有暫時切換跟永久切換兩種 ## 在 rails db 環境下 列出所有 schema ``` \dn # List of schemas # Name | Owner #--------+---------- # public | postgres # abc | etrex # (2 rows) ``` 顯示目前所在的 schema ``` SHOW search_path; # search_path #----------------- # "$user", public # (1 row) ``` 切換 schema ``` SET search_path TO abc; # SET SET search_path TO 'abc'; # SET ``` 字串要不要加引號都可以,當 search_path 的值不在 schema 的列表上時不會跳 error,可以想像成連接到一個空的 schema,而這樣並不代表新增了一個 schema。 在 select 的當下使用 schema ``` SELECT * FROM schema_name.table_name; ``` ## 關於 excluded_models 在 apartment 提供的功能中有一個功能,這個功能是可以設定在切換 schema 的時候,讓某些表格不要被切換,也就是做成全域的表格。 舉例來說,假設我有兩個 table,是使用以下 code 所生成。 ``` # 這裡是 bash rails g model a rails g model b a:references rails db:migrate ``` 然後我將 B 做成全域的表格 ``` # 這裡是 /config/initializers/apartment.rb 約 18 行的位置 config.excluded_models = %w{ B } ``` 然後建立一個 tenant 'abc' 並且切換過去,在這個環境下做測試 ``` # 這裡是 rails c 環境下 Apartment::Tenant.create('abc') Apartment::Tenant.switch!('abc') ``` 對 A 做一些事情 ``` A.count # (3.8ms) SELECT COUNT(*) FROM "as" # => 0 ``` 對 B 也做一些事情 ``` B.count # (0.5ms) SELECT COUNT(*) FROM "public"."bs" # => 0 ``` 在這裡可以看到,當你對 B 進行操作時,其實是會強制加上 "public" 的,也就是說,對於 B ,不管在哪一個 tenant 下,都只會去存取 schema 'public' 下的表格,所以其他的 schema 的表格 bs 應該會都是空的。 嘗試在 tenant 'abc' 下新增資料 ``` a = A.create() # (0.2ms) BEGIN # SQL (0.4ms) INSERT INTO "as" ("created_at", "updated_at") VALUES ($1, $2) RETURNING "id" [["created_at", "2017-10-25 17:16:22.084365"], ["updated_at", "2017-10-25 17:16:22.084365"]] b = B.create(a:a) # (0.2ms) BEGIN # SQL (1.5ms) INSERT INTO "public"."bs" ("a_id", "created_at", "updated_at") VALUES ($1, $2, $3) RETURNING "id" [["a_id", 3], ["created_at", "2017-10-25 17:16:24.957753"], ["updated_at", "2017-10-25 17:16:24.957753"]] # (0.2ms) ROLLBACK #ActiveRecord::InvalidForeignKey: PG::ForeignKeyViolation: ERROR: insert or update on table "bs" violates foreign #key constraint "fk_rails_ddf8c0c4b5" #DETAIL: Key (a_id)=(3) is not present in table "as". #: INSERT INTO "public"."bs" ("a_id", "created_at", "updated_at") VALUES ($1, $2, $3) RETURNING "id" # from (irb):31 ``` 這時候出大問題了,因為 A 建立在 abc 下,但是 B 建立在 public 下。而 public.bs 有一個外來鍵限制是 a_id 必須要在 public.as 中有出現。 使用 [pgAdmin](https://www.pgadmin.org/) 連進去看,對 db_name/Schemas/public/Tables/bs/Constraints/fk_rails_ddf8c04b5 按右鍵選 CREATE Script: 。
會看到生成的 script 內容為: ``` -- Constraint: fk_rails_ddf8c0c4b5 -- ALTER TABLE public.bs DROP CONSTRAINT fk_rails_ddf8c0c4b5; ALTER TABLE public.bs ADD CONSTRAINT fk_rails_ddf8c0c4b5 FOREIGN KEY (a_id) REFERENCES public."as" (id) MATCH SIMPLE ON UPDATE NO ACTION ON DELETE NO ACTION; ``` 在這裡明確地指出跟 public.bs 有關係的表格是 public.as ,而不是其他 schema 下的 as,但是 Apartment 不處理,所以會出問題。

2017/10/19

Network debug 方法

markdown ## linux 環境下 查詢所有 process ``` ps ``` 輸出 ``` PID TTY TIME CMD 21968 ttys000 0:00.16 -bash ``` 查詢 port 3000 被誰占用 ``` lsof -i :3000 ``` 輸出 ``` COMMAND PID USER FD TYPE DEVICE SIZE/OFF NODE NAME ruby 20308 etrex 25u IPv4 0xd000000000000 0t0 TCP *:hbci (LISTEN) ``` 刪除某個 process ``` kill 20308 ``` ## 參考文件 kill process:[https://blog.gtwang.org/linux/linux-kill-killall-xkill/](https://blog.gtwang.org/linux/linux-kill-killall-xkill/)

Ruby - RSpec 的使用方法

markdown ## RSpec 簡介 RSpec 在執行的時候不保證執行順序,每個測試產生的資料不會自動被清除。 新增測試的generator語法: ``` rails generate rspec:model user ``` 這樣寫會建立一個檔案在 spec/models/user_spec.rb 執行測試的指令是在 rails 專案目錄下輸入以下指令: ``` # 這裡是 bash # 執行所有測試 rspec # 執行某個資料夾下的所有測試 rspec ./spec/models # 執行某個檔案裡的所有測試 rspec ./spec/models/user_spec.rb # 執行某個檔案裡的某個測試 rspec ./spec/models/user_spec.rb:8 ``` 簡單的範例 ``` require 'rails_helper' RSpec.describe "規格說明" do describe "處於某個狀態下" do # 設定狀態變數 let(:a) { 1 } it "should be 1" do puts "should be 1" expect(a).to eq(1) end end end ``` let 在每次測試裡,第一次存取變數時就會執行對應的程式 ``` require 'rails_helper' RSpec.describe "規格說明" do describe "處於某個狀態下" do let(:a) { puts "let a"; 1 } it "1" do puts "1" puts "a=#{a}" end it "2" do puts "2" puts "a=#{a}" puts "a=#{a}" puts "a=#{a}" end end end ``` 輸出 ``` 1 let a a=1 .2 let a a=1 a=1 a=1 . ``` before 和 after 會在所有測試的執行前後做事 ``` RSpec.describe "規格說明" do describe "處於某個狀態下" do before { puts "before" } after { puts "after" } it "1" do puts "1" end it "2" do puts "2" end end end ``` 執行結果 ``` before 1 after .before 2 after . ``` 因為測試對資料庫的操作會互相影響,如果想要確保每個測試都是在資料庫乾淨的狀態下,可以使用 [database_cleaner](https://github.com/DatabaseCleaner/database_cleaner)。 ``` require 'database_cleaner' ... RSpec.configure do |config| ... config.before(:suite) do DatabaseCleaner.strategy = :transaction DatabaseCleaner.clean_with(:truncation) end config.around(:each) do |example| DatabaseCleaner.cleaning do example.run end end ... end ``` 其中的 before、around 可以參考官網說明文件:[https://relishapp.com/rspec/rspec-core/v/2-13/docs/hooks/before-and-after-hooks#before/after-blocks-defined-in-config-are-run-in-order](https://relishapp.com/rspec/rspec-core/v/2-13/docs/hooks/before-and-after-hooks#before/after-blocks-defined-in-config-are-run-in-order) 如果想要控制 rspec 執行每一個 test 的順序,可以這樣寫: ``` # 這裡是 bash rspec --order defined rspec --seed 1 ``` 詳細的介紹可以參考 [https://relishapp.com/rspec/rspec-core/docs/command-line/order](https://relishapp.com/rspec/rspec-core/docs/command-line/order) 如果想要把結果輸出到檔案,可以這樣寫: ``` # 這裡是 bash rspec --out result.txt rspec --format documentation --out result.txt ``` 上面兩行會使檔案內容不同,詳細的介紹可以參考 [https://relishapp.com/rspec/rspec-core/v/2-4/docs/command-line/format-option](https://relishapp.com/rspec/rspec-core/v/2-4/docs/command-line/format-option)

Ruby - debug 方法

markdown ## 查詢繼承關係 ``` File.ancestors # [File, IO, File::Constants, Enumerable, Object, Kernel, BasicObject] ``` ## 從物件找方法 ``` # 查 File 的類別方法 File.methods # 查 File 的實體方法 File.instance_methods # 查 File 的實體方法 File.new('/').methods # 取得繼承樹上所有的方法 File.methods(true) # 只取得屬於 File 的方法 File.methods(false) ``` ## 從方法找定義 ``` File.method(:read) # #Method: File(IO).read IO.method(:read) # #Method: IO.read IO.method(:read).source_location # nil ``` 因為 IO.read 的定義是寫在 c 語言,所以就不顯示了。 ## 參考文件 為什麼File.method(:read)是nil:[https://ja.stackoverflow.com/questions/5755/ruby-file-read-%E3%83%A1%E3%82%BD%E3%83%83%E3%83%89%E3%81%AE%E8%AA%AC%E6%98%8E%E3%82%92api%E3%83%89%E3%82%AD%E3%83%A5%E3%83%A1%E3%83%B3%E3%83%88%E3%81%A7%E8%AA%BF%E3%81%B9%E3%81%9F%E3%81%84](https://ja.stackoverflow.com/questions/5755/ruby-file-read-%E3%83%A1%E3%82%BD%E3%83%83%E3%83%89%E3%81%AE%E8%AA%AC%E6%98%8E%E3%82%92api%E3%83%89%E3%82%AD%E3%83%A5%E3%83%A1%E3%83%B3%E3%83%88%E3%81%A7%E8%AA%BF%E3%81%B9%E3%81%9F%E3%81%84)

Python - 電腦字體數字辨識

markdown 續前篇:[Python - draw text on image and image to numpy array](http://etrex.blogspot.tw/2017/10/python-draw-text-on-image-and-image-to.html) ## 目標 嘗試建立一個簡單的類神經網路,只針對電腦字體數字 0~9 共 10 張圖片作訓練,訓練資料就是測試資料,在 40 次左右的訓練後正確度可達到 100%。 ## 本文包含 * keras 的使用 * keras Model 輸出成圖片 ## 程式碼 ``` import numpy as np from PIL import Image from PIL import ImageFont from PIL import ImageDraw image_size = (8,13) font_size = 10 x_train = np.array([]) y_train = np.array([]) for i in range(10): # 空白圖片生成 image = Image.new('L', image_size, 0) # 取得繪圖器 draw = ImageDraw.Draw(image) # 字型設定 # font = ImageFont.truetype("C:/Windows/Fonts/cour.ttf", font_size) font = ImageFont.truetype("C:/Windows/Fonts/msjh.ttc", font_size) # 關閉反鋸齒 draw.fontmode = '1' # 測量文字尺寸 text_size = draw.textsize(str(i),font) # print('text_size:', text_size) # 文字置中 text_position = ((image_size[0]-text_size[0])//2,(image_size[1]-text_size[1])//2) # print('text_position:', text_position) # 畫上文字 draw.text(text_position, str(i), 255, font) # 存檔 image.save(str(i)+'.bmp') # 轉成 numpy array na = np.array(image.getdata()).reshape(image.size[1], image.size[0]) # 加入訓練資料 x_train = np.append(x_train,na) y_train = np.append(y_train,i) import keras from keras.models import Sequential from keras.layers import Dense from keras.optimizers import RMSprop # 每次更新時所採用的資料筆數 batch_size = 1 # 總共有幾種數字 num_classes = 10 # 要更新幾次 epochs = 100 # 把資料轉成 model 需要的格式 x_train = x_train.reshape(10, 104) x_train = x_train.astype('float32') x_train /= 255 print(x_train.shape[0], 'train samples') y_train = keras.utils.to_categorical(y_train, num_classes) # 建立 model model = Sequential() # 一層 + softmax model.add(Dense(num_classes, input_shape=(104,), activation='softmax')) # ?? model.summary() # 選擇 loss function 和 optimizer model.compile(loss='categorical_crossentropy', optimizer=RMSprop(), metrics=['accuracy']) # 開始訓練 history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, verbose=1, validation_data=(x_train, y_train)) # 計算分數 score = model.evaluate(x_train, y_train, verbose=0) print('Test loss:', score[0]) print('Test accuracy:', score[1]) # 儲存 model 至圖片 from keras.utils import plot_model plot_model(model, to_file='model.png') ``` ## 輸出 ``` Using TensorFlow backend. 10 train samples _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= dense_1 (Dense) (None, 10) 1050 ================================================================= Total params: 1,050 Trainable params: 1,050 Non-trainable params: 0 _________________________________________________________________ Train on 10 samples, validate on 10 samples Epoch 1/100 10/10 [==============================] - ETA: 0s - loss: 2.5473 - acc: 0.0000e+00 - val_loss: 2.4555 - val_acc: 0.1000 Epoch 2/100 10/10 [==============================] - ETA: 0s - loss: 2.4563 - acc: 0.1000 - val_loss: 2.3924 - val_acc: 0.1000 Epoch 3/100 10/10 [==============================] - ETA: 0s - loss: 2.3926 - acc: 0.1000 - val_loss: 2.3334 - val_acc: 0.2000 Epoch 4/100 10/10 [==============================] - ETA: 0s - loss: 2.3336 - acc: 0.2000 - val_loss: 2.2760 - val_acc: 0.2000 Epoch 5/100 10/10 [==============================] - ETA: 0s - loss: 2.2766 - acc: 0.2000 - val_loss: 2.2196 - val_acc: 0.2000 Epoch 6/100 10/10 [==============================] - ETA: 0s - loss: 2.2203 - acc: 0.2000 - val_loss: 2.1636 - val_acc: 0.4000 Epoch 7/100 10/10 [==============================] - ETA: 0s - loss: 2.1640 - acc: 0.4000 - val_loss: 2.1090 - val_acc: 0.5000 Epoch 8/100 10/10 [==============================] - ETA: 0s - loss: 2.1096 - acc: 0.4000 - val_loss: 2.0545 - val_acc: 0.5000 Epoch 9/100 10/10 [==============================] - ETA: 0s - loss: 2.0556 - acc: 0.5000 - val_loss: 2.0013 - val_acc: 0.5000 Epoch 10/100 10/10 [==============================] - ETA: 0s - loss: 2.0023 - acc: 0.5000 - val_loss: 1.9484 - val_acc: 0.5000 Epoch 11/100 10/10 [==============================] - ETA: 0s - loss: 1.9493 - acc: 0.5000 - val_loss: 1.8971 - val_acc: 0.5000 Epoch 12/100 10/10 [==============================] - ETA: 0s - loss: 1.8985 - acc: 0.5000 - val_loss: 1.8453 - val_acc: 0.5000 Epoch 13/100 10/10 [==============================] - ETA: 0s - loss: 1.8464 - acc: 0.5000 - val_loss: 1.7949 - val_acc: 0.5000 Epoch 14/100 10/10 [==============================] - ETA: 0s - loss: 1.7962 - acc: 0.5000 - val_loss: 1.7448 - val_acc: 0.5000 Epoch 15/100 10/10 [==============================] - ETA: 0s - loss: 1.7464 - acc: 0.5000 - val_loss: 1.6958 - val_acc: 0.5000 Epoch 16/100 10/10 [==============================] - ETA: 0s - loss: 1.6971 - acc: 0.5000 - val_loss: 1.6471 - val_acc: 0.5000 Epoch 17/100 10/10 [==============================] - ETA: 0s - loss: 1.6486 - acc: 0.5000 - val_loss: 1.5994 - val_acc: 0.5000 Epoch 18/100 10/10 [==============================] - ETA: 0s - loss: 1.6007 - acc: 0.5000 - val_loss: 1.5520 - val_acc: 0.7000 Epoch 19/100 10/10 [==============================] - ETA: 0s - loss: 1.5538 - acc: 0.7000 - val_loss: 1.5064 - val_acc: 0.7000 Epoch 20/100 10/10 [==============================] - ETA: 0s - loss: 1.5078 - acc: 0.7000 - val_loss: 1.4612 - val_acc: 0.7000 Epoch 21/100 10/10 [==============================] - ETA: 0s - loss: 1.4629 - acc: 0.7000 - val_loss: 1.4168 - val_acc: 0.7000 Epoch 22/100 10/10 [==============================] - ETA: 0s - loss: 1.4190 - acc: 0.7000 - val_loss: 1.3736 - val_acc: 0.7000 Epoch 23/100 10/10 [==============================] - ETA: 0s - loss: 1.3758 - acc: 0.7000 - val_loss: 1.3311 - val_acc: 0.8000 Epoch 24/100 10/10 [==============================] - ETA: 0s - loss: 1.3331 - acc: 0.8000 - val_loss: 1.2891 - val_acc: 0.8000 Epoch 25/100 10/10 [==============================] - ETA: 0s - loss: 1.2913 - acc: 0.8000 - val_loss: 1.2488 - val_acc: 0.9000 Epoch 26/100 10/10 [==============================] - ETA: 0s - loss: 1.2513 - acc: 0.9000 - val_loss: 1.2091 - val_acc: 0.9000 Epoch 27/100 10/10 [==============================] - ETA: 0s - loss: 1.2114 - acc: 0.9000 - val_loss: 1.1701 - val_acc: 0.9000 Epoch 28/100 10/10 [==============================] - ETA: 0s - loss: 1.1721 - acc: 0.9000 - val_loss: 1.1324 - val_acc: 0.9000 Epoch 29/100 10/10 [==============================] - ETA: 0s - loss: 1.1349 - acc: 0.9000 - val_loss: 1.0957 - val_acc: 0.9000 Epoch 30/100 10/10 [==============================] - ETA: 0s - loss: 1.0984 - acc: 0.9000 - val_loss: 1.0596 - val_acc: 0.9000 Epoch 31/100 10/10 [==============================] - ETA: 0s - loss: 1.0620 - acc: 0.9000 - val_loss: 1.0243 - val_acc: 0.9000 Epoch 32/100 10/10 [==============================] - ETA: 0s - loss: 1.0269 - acc: 0.9000 - val_loss: 0.9905 - val_acc: 1.0000 Epoch 33/100 10/10 [==============================] - ETA: 0s - loss: 0.9932 - acc: 1.0000 - val_loss: 0.9571 - val_acc: 1.0000 Epoch 34/100 10/10 [==============================] - ETA: 0s - loss: 0.9597 - acc: 1.0000 - val_loss: 0.9247 - val_acc: 1.0000 Epoch 35/100 10/10 [==============================] - ETA: 0s - loss: 0.9276 - acc: 1.0000 - val_loss: 0.8934 - val_acc: 1.0000 Epoch 36/100 10/10 [==============================] - ETA: 0s - loss: 0.8962 - acc: 1.0000 - val_loss: 0.8629 - val_acc: 1.0000 Epoch 37/100 10/10 [==============================] - ETA: 0s - loss: 0.8655 - acc: 1.0000 - val_loss: 0.8330 - val_acc: 1.0000 Epoch 38/100 10/10 [==============================] - ETA: 0s - loss: 0.8359 - acc: 1.0000 - val_loss: 0.8047 - val_acc: 1.0000 Epoch 39/100 10/10 [==============================] - ETA: 0s - loss: 0.8077 - acc: 1.0000 - val_loss: 0.7766 - val_acc: 1.0000 Epoch 40/100 10/10 [==============================] - ETA: 0s - loss: 0.7797 - acc: 1.0000 - val_loss: 0.7501 - val_acc: 1.0000 Epoch 41/100 10/10 [==============================] - ETA: 0s - loss: 0.7529 - acc: 1.0000 - val_loss: 0.7239 - val_acc: 1.0000 Epoch 42/100 10/10 [==============================] - ETA: 0s - loss: 0.7270 - acc: 1.0000 - val_loss: 0.6988 - val_acc: 1.0000 Epoch 43/100 10/10 [==============================] - ETA: 0s - loss: 0.7018 - acc: 1.0000 - val_loss: 0.6744 - val_acc: 1.0000 Epoch 44/100 10/10 [==============================] - ETA: 0s - loss: 0.6773 - acc: 1.0000 - val_loss: 0.6511 - val_acc: 1.0000 Epoch 45/100 10/10 [==============================] - ETA: 0s - loss: 0.6542 - acc: 1.0000 - val_loss: 0.6283 - val_acc: 1.0000 Epoch 46/100 10/10 [==============================] - ETA: 0s - loss: 0.6312 - acc: 1.0000 - val_loss: 0.6065 - val_acc: 1.0000 Epoch 47/100 10/10 [==============================] - ETA: 0s - loss: 0.6097 - acc: 1.0000 - val_loss: 0.5852 - val_acc: 1.0000 Epoch 48/100 10/10 [==============================] - ETA: 0s - loss: 0.5882 - acc: 1.0000 - val_loss: 0.5647 - val_acc: 1.0000 Epoch 49/100 10/10 [==============================] - ETA: 0s - loss: 0.5677 - acc: 1.0000 - val_loss: 0.5451 - val_acc: 1.0000 Epoch 50/100 10/10 [==============================] - ETA: 0s - loss: 0.5482 - acc: 1.0000 - val_loss: 0.5261 - val_acc: 1.0000 Epoch 51/100 10/10 [==============================] - ETA: 0s - loss: 0.5291 - acc: 1.0000 - val_loss: 0.5078 - val_acc: 1.0000 Epoch 52/100 10/10 [==============================] - ETA: 0s - loss: 0.5108 - acc: 1.0000 - val_loss: 0.4900 - val_acc: 1.0000 Epoch 53/100 10/10 [==============================] - ETA: 0s - loss: 0.4928 - acc: 1.0000 - val_loss: 0.4729 - val_acc: 1.0000 Epoch 54/100 10/10 [==============================] - ETA: 0s - loss: 0.4759 - acc: 1.0000 - val_loss: 0.4567 - val_acc: 1.0000 Epoch 55/100 10/10 [==============================] - ETA: 0s - loss: 0.4596 - acc: 1.0000 - val_loss: 0.4409 - val_acc: 1.0000 Epoch 56/100 10/10 [==============================] - ETA: 0s - loss: 0.4440 - acc: 1.0000 - val_loss: 0.4258 - val_acc: 1.0000 Epoch 57/100 10/10 [==============================] - ETA: 0s - loss: 0.4286 - acc: 1.0000 - val_loss: 0.4113 - val_acc: 1.0000 Epoch 58/100 10/10 [==============================] - ETA: 0s - loss: 0.4143 - acc: 1.0000 - val_loss: 0.3971 - val_acc: 1.0000 Epoch 59/100 10/10 [==============================] - ETA: 0s - loss: 0.4001 - acc: 1.0000 - val_loss: 0.3835 - val_acc: 1.0000 Epoch 60/100 10/10 [==============================] - ETA: 0s - loss: 0.3864 - acc: 1.0000 - val_loss: 0.3706 - val_acc: 1.0000 Epoch 61/100 10/10 [==============================] - ETA: 0s - loss: 0.3735 - acc: 1.0000 - val_loss: 0.3580 - val_acc: 1.0000 Epoch 62/100 10/10 [==============================] - ETA: 0s - loss: 0.3611 - acc: 1.0000 - val_loss: 0.3459 - val_acc: 1.0000 Epoch 63/100 10/10 [==============================] - ETA: 0s - loss: 0.3487 - acc: 1.0000 - val_loss: 0.3343 - val_acc: 1.0000 Epoch 64/100 10/10 [==============================] - ETA: 0s - loss: 0.3370 - acc: 1.0000 - val_loss: 0.3232 - val_acc: 1.0000 Epoch 65/100 10/10 [==============================] - ETA: 0s - loss: 0.3261 - acc: 1.0000 - val_loss: 0.3125 - val_acc: 1.0000 Epoch 66/100 10/10 [==============================] - ETA: 0s - loss: 0.3152 - acc: 1.0000 - val_loss: 0.3023 - val_acc: 1.0000 Epoch 67/100 10/10 [==============================] - ETA: 0s - loss: 0.3049 - acc: 1.0000 - val_loss: 0.2925 - val_acc: 1.0000 Epoch 68/100 10/10 [==============================] - ETA: 0s - loss: 0.2953 - acc: 1.0000 - val_loss: 0.2828 - val_acc: 1.0000 Epoch 69/100 10/10 [==============================] - ETA: 0s - loss: 0.2854 - acc: 1.0000 - val_loss: 0.2736 - val_acc: 1.0000 Epoch 70/100 10/10 [==============================] - ETA: 0s - loss: 0.2762 - acc: 1.0000 - val_loss: 0.2648 - val_acc: 1.0000 Epoch 71/100 10/10 [==============================] - ETA: 0s - loss: 0.2675 - acc: 1.0000 - val_loss: 0.2563 - val_acc: 1.0000 Epoch 72/100 10/10 [==============================] - ETA: 0s - loss: 0.2588 - acc: 1.0000 - val_loss: 0.2481 - val_acc: 1.0000 Epoch 73/100 10/10 [==============================] - ETA: 0s - loss: 0.2506 - acc: 1.0000 - val_loss: 0.2403 - val_acc: 1.0000 Epoch 74/100 10/10 [==============================] - ETA: 0s - loss: 0.2429 - acc: 1.0000 - val_loss: 0.2327 - val_acc: 1.0000 Epoch 75/100 10/10 [==============================] - ETA: 0s - loss: 0.2351 - acc: 1.0000 - val_loss: 0.2254 - val_acc: 1.0000 Epoch 76/100 10/10 [==============================] - ETA: 0s - loss: 0.2279 - acc: 1.0000 - val_loss: 0.2184 - val_acc: 1.0000 Epoch 77/100 10/10 [==============================] - ETA: 0s - loss: 0.2209 - acc: 1.0000 - val_loss: 0.2117 - val_acc: 1.0000 Epoch 78/100 10/10 [==============================] - ETA: 0s - loss: 0.2140 - acc: 1.0000 - val_loss: 0.2053 - val_acc: 1.0000 Epoch 79/100 10/10 [==============================] - ETA: 0s - loss: 0.2078 - acc: 1.0000 - val_loss: 0.1989 - val_acc: 1.0000 Epoch 80/100 10/10 [==============================] - ETA: 0s - loss: 0.2011 - acc: 1.0000 - val_loss: 0.1929 - val_acc: 1.0000 Epoch 81/100 10/10 [==============================] - ETA: 0s - loss: 0.1954 - acc: 1.0000 - val_loss: 0.1872 - val_acc: 1.0000 Epoch 82/100 10/10 [==============================] - ETA: 0s - loss: 0.1894 - acc: 1.0000 - val_loss: 0.1817 - val_acc: 1.0000 Epoch 83/100 10/10 [==============================] - ETA: 0s - loss: 0.1841 - acc: 1.0000 - val_loss: 0.1763 - val_acc: 1.0000 Epoch 84/100 10/10 [==============================] - ETA: 0s - loss: 0.1784 - acc: 1.0000 - val_loss: 0.1711 - val_acc: 1.0000 Epoch 85/100 10/10 [==============================] - ETA: 0s - loss: 0.1732 - acc: 1.0000 - val_loss: 0.1662 - val_acc: 1.0000 Epoch 86/100 10/10 [==============================] - ETA: 0s - loss: 0.1684 - acc: 1.0000 - val_loss: 0.1614 - val_acc: 1.0000 Epoch 87/100 10/10 [==============================] - ETA: 0s - loss: 0.1636 - acc: 1.0000 - val_loss: 0.1568 - val_acc: 1.0000 Epoch 88/100 10/10 [==============================] - ETA: 0s - loss: 0.1590 - acc: 1.0000 - val_loss: 0.1523 - val_acc: 1.0000 Epoch 89/100 10/10 [==============================] - ETA: 0s - loss: 0.1542 - acc: 1.0000 - val_loss: 0.1480 - val_acc: 1.0000 Epoch 90/100 10/10 [==============================] - ETA: 0s - loss: 0.1501 - acc: 1.0000 - val_loss: 0.1438 - val_acc: 1.0000 Epoch 91/100 10/10 [==============================] - ETA: 0s - loss: 0.1457 - acc: 1.0000 - val_loss: 0.1399 - val_acc: 1.0000 Epoch 92/100 10/10 [==============================] - ETA: 0s - loss: 0.1418 - acc: 1.0000 - val_loss: 0.1360 - val_acc: 1.0000 Epoch 93/100 10/10 [==============================] - ETA: 0s - loss: 0.1379 - acc: 1.0000 - val_loss: 0.1323 - val_acc: 1.0000 Epoch 94/100 10/10 [==============================] - ETA: 0s - loss: 0.1343 - acc: 1.0000 - val_loss: 0.1287 - val_acc: 1.0000 Epoch 95/100 10/10 [==============================] - ETA: 0s - loss: 0.1306 - acc: 1.0000 - val_loss: 0.1252 - val_acc: 1.0000 Epoch 96/100 10/10 [==============================] - ETA: 0s - loss: 0.1271 - acc: 1.0000 - val_loss: 0.1218 - val_acc: 1.0000 Epoch 97/100 10/10 [==============================] - ETA: 0s - loss: 0.1237 - acc: 1.0000 - val_loss: 0.1185 - val_acc: 1.0000 Epoch 98/100 10/10 [==============================] - ETA: 0s - loss: 0.1204 - acc: 1.0000 - val_loss: 0.1154 - val_acc: 1.0000 Epoch 99/100 10/10 [==============================] - ETA: 0s - loss: 0.1172 - acc: 1.0000 - val_loss: 0.1124 - val_acc: 1.0000 Epoch 100/100 10/10 [==============================] - ETA: 0s - loss: 0.1141 - acc: 1.0000 - val_loss: 0.1095 - val_acc: 1.0000 Test loss: 0.109465420246 Test accuracy: 1.0 ``` ## 輸出的圖片 plot_model 的執行,需要安裝 pydot 和 graphviz ``` pip3 install pydot ``` ## plot_model 的環境安裝 在 windows 上,graphviz 必須手動安裝,安裝檔可以到他們官網下載。 [http://www.graphviz.org/Download_windows.php](http://www.graphviz.org/Download_windows.php) 安裝好之後必須在環境變數 path 加上 C:\Program Files (x86)\Graphviz2.30\bin 才能使用。 cmd 下輸入指令: ``` dot -version ``` 若顯示以下訊息,表示正確安裝。 ``` dot - graphviz version 2.30.1 (20130214.1330) libdir = "C:\Program Files (x86)\Graphviz2.30\bin" Activated plugin library: gvplugin_pango.dll Using textlayout: textlayout:cairo Activated plugin library: gvplugin_dot_layout.dll Using layout: dot:dot_layout Activated plugin library: gvplugin_core.dll Using render: dot:core Using device: dot:dot:core The plugin configuration file: C:\Program Files (x86)\Graphviz2.30\bin\config6 was successfully loaded. render : cairo dot fig gd gdiplus map pic pov ps svg tk vml vrml xdot layout : circo dot fdp neato nop nop1 nop2 osage patchwork sfdp twopi textlayout : textlayout device : bmp canon cmap cmapx cmapx_np dot emf emfplus eps fig gd gd2 gif gv imap imap_np ismap jpe jpeg jpg metafile pdf p ic plain plain-ext png pov ps ps2 svg svgz tif tiff tk vml vmlz vrml wbmp xdot loadimage : (lib) bmp eps gd gd2 gif jpe jpeg jpg png ps svg ``` ## 參考文件 plot_model 在 windows 上的環境建立:[https://zhuanlan.zhihu.com/p/28158957](https://zhuanlan.zhihu.com/p/28158957) graphviz 的使用:[https://www.openfoundry.org/tw/foss-programs/8820-graphviz-](https://www.openfoundry.org/tw/foss-programs/8820-graphviz-)

2017/10/17

Python - draw text on image and image to numpy array

markdown ## 目標 嘗試生成以微軟正黑體寫成的數字0~9並轉換成 numpy array ##本文包含 * 生成圖片和保存圖片 * 在圖片上寫出指定字型和大小的字 * 設定反鋸齒模式 * 圖片轉成 numpy array ## 程式碼 ``` import numpy from PIL import Image from PIL import ImageFont from PIL import ImageDraw image_size = (8,13) font_size = 10 for i in range(10): # 空白圖片生成 image = Image.new('L', image_size, 0) # 取得繪圖器 draw = ImageDraw.Draw(image) # 微軟正黑體 font = ImageFont.truetype("C:/Windows/Fonts/msjh.ttc", font_size) # 關閉反鋸齒 draw.fontmode = '1' # 測量文字尺寸 text_size = draw.textsize(str(i),font) # print('text_size:', text_size) # 文字置中 text_position = ((image_size[0]-text_size[0])//2,(image_size[1]-text_size[1])//2) # print('text_position:', text_position) # 畫上文字 draw.text(text_position, str(i), 255, font) # 存檔 image.save(str(i)+'.bmp') # 轉成 numpy array na = numpy.array(image.getdata()).reshape(image.size[1], image.size[0]) # 印出 print(na) ``` ## 輸出 ``` Using TensorFlow backend. [[ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 255 255 255 255 0 0] [ 0 0 255 0 0 255 255 0] [ 0 255 0 0 0 0 255 0] [ 0 255 0 0 0 0 255 0] [ 0 255 0 0 0 0 255 0] [ 0 255 0 0 0 0 255 0] [ 0 255 255 0 0 255 0 0] [ 0 0 255 255 255 255 0 0] [ 0 0 0 0 0 0 0 0]] [[ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 255 255 0 0 0] [ 0 0 255 255 255 0 0 0] [ 0 0 0 0 255 0 0 0] [ 0 0 0 0 255 0 0 0] [ 0 0 0 0 255 0 0 0] [ 0 0 0 0 255 0 0 0] [ 0 0 0 0 255 0 0 0] [ 0 0 255 255 255 255 255 0] [ 0 0 0 0 0 0 0 0]] [[ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 255 255 255 0 0] [ 0 0 255 0 0 0 255 0] [ 0 0 0 0 0 0 255 0] [ 0 0 0 0 0 0 255 0] [ 0 0 0 0 0 255 0 0] [ 0 0 0 0 255 0 0 0] [ 0 0 0 255 0 0 0 0] [ 0 0 255 255 255 255 255 0] [ 0 0 0 0 0 0 0 0]] [[ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 255 255 0 0 0] [ 0 0 255 0 0 255 0 0] [ 0 0 0 0 0 255 0 0] [ 0 0 0 255 255 0 0 0] [ 0 0 0 0 0 255 0 0] [ 0 0 0 0 0 255 0 0] [ 0 0 255 0 0 255 0 0] [ 0 0 255 255 255 0 0 0] [ 0 0 0 0 0 0 0 0]] [[ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 255 255 0 0] [ 0 0 0 0 255 255 0 0] [ 0 0 0 255 0 255 0 0] [ 0 0 255 0 0 255 0 0] [ 0 255 0 0 0 255 0 0] [ 0 255 255 255 255 255 255 0] [ 0 0 0 0 0 255 0 0] [ 0 0 0 0 0 255 0 0] [ 0 0 0 0 0 0 0 0]] [[ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 255 255 255 255 0 0] [ 0 0 255 0 0 0 0 0] [ 0 0 255 0 0 0 0 0] [ 0 0 255 255 255 0 0 0] [ 0 0 0 0 0 255 0 0] [ 0 0 0 0 0 255 0 0] [ 0 0 255 0 0 255 0 0] [ 0 0 255 255 255 0 0 0] [ 0 0 0 0 0 0 0 0]] [[ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 255 255 255 0] [ 0 0 0 255 0 0 0 0] [ 0 0 255 0 0 0 0 0] [ 0 0 255 0 255 255 0 0] [ 0 0 255 255 0 0 255 0] [ 0 0 255 0 0 0 255 0] [ 0 0 255 0 0 0 255 0] [ 0 0 0 255 255 255 0 0] [ 0 0 0 0 0 0 0 0]] [[ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 255 255 255 255 255 255 0] [ 0 0 0 0 0 255 0 0] [ 0 0 0 0 0 255 0 0] [ 0 0 0 0 255 0 0 0] [ 0 0 0 0 255 0 0 0] [ 0 0 0 255 0 0 0 0] [ 0 0 0 255 0 0 0 0] [ 0 0 255 0 0 0 0 0] [ 0 0 0 0 0 0 0 0]] [[ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 255 255 0 0 0] [ 0 0 255 0 0 255 0 0] [ 0 0 255 0 0 255 0 0] [ 0 0 0 255 255 0 0 0] [ 0 0 0 255 0 255 0 0] [ 0 0 255 0 0 0 255 0] [ 0 0 255 0 0 0 255 0] [ 0 0 0 255 255 255 0 0] [ 0 0 0 0 0 0 0 0]] [[ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 0 0 0 0 0] [ 0 0 0 255 255 255 0 0] [ 0 0 255 0 0 0 255 0] [ 0 0 255 0 0 0 255 0] [ 0 0 255 0 0 0 255 0] [ 0 0 0 255 255 255 255 0] [ 0 0 0 0 0 0 255 0] [ 0 0 0 0 0 255 0 0] [ 0 0 255 255 255 0 0 0] [ 0 0 0 0 0 0 0 0]] ``` ## 參考文件 PIL影像相關:[http://pillow.readthedocs.io/en/stable/reference/index.html](http://pillow.readthedocs.io/en/stable/reference/index.html)

2017/10/13

Python - 印出 mnist 的手寫圖形

markdown ## 安裝 keras 請參考 [https://etrex.blogspot.tw/2017/10/windows-10-keras.html](https://etrex.blogspot.tw/2017/10/windows-10-keras.html) ## 安裝 PIL 在 cmd 輸入以下指令: ``` pip3 install Image ``` 若想查看已安裝了哪些套件,可在 cmd 輸入以下指令: ``` pip3 list ``` ## 印出 MNIST 的手寫數字圖形 [MNIST](http://yann.lecun.com/exdb/mnist/) 是一個知名的手寫數字資料集,被當作機器學習界的 hello world 來使用。 ``` import keras from keras.datasets import mnist from PIL import Image (x_train, y_train), (x_test, y_test) = mnist.load_data() Image.fromarray(x_train[0,]).show() ``` x_train 是一個大小為 60000,28,28 的三維陣列,代表 60000 張 28x28 的灰階圖檔。 可輸入以下指令查看: ``` x_train.shape # (60000,28,28) ``` x_train[0,] 是一個大小為 28,28 的二維陣列,代表第一張圖片。 Image.fromarray(x_train[0,]).show() 把代表第一張圖片的二維陣列轉為圖片並顯示。

Python - debug 方法

印出變數值 ``` a = 1 print(a) # 1 ``` 印出變數型態 ``` a = 1 print(type(a)) # ``` 印出變數成員 ``` a = 1 print(dir(a)) # ['__abs__', '__add__', '__and__', '__bool__', '__ceil__', '__class__', '__delattr__', '__dir__', '__divmod__', '__doc__', '__eq__', '__float__', '__floor__', '__floordiv__', '__format__', '__ge__', '__getattribute__', '__getnewargs__', '__gt__', '__hash__', '__index__', '__init__', '__int__', '__invert__', '__le__', '__lshift__', '__lt__', '__mod__', '__mul__', '__ne__', '__neg__', '__new__', '__or__', '__pos__', '__pow__', '__radd__', '__rand__', '__rdivmod__', '__reduce__', '__reduce_ex__', '__repr__', '__rfloordiv__', '__rlshift__', '__rmod__', '__rmul__', '__ror__', '__round__', '__rpow__', '__rrshift__', '__rshift__', '__rsub__', '__rtruediv__', '__rxor__', '__setattr__', '__sizeof__', '__str__', '__sub__', '__subclasshook__', '__truediv__', '__trunc__', '__xor__', 'bit_length', 'conjugate', 'denominator', 'from_bytes', 'imag', 'numerator', 'real', 'to_bytes'] ``` 印出模組成員 ``` import json dir(json) # ['JSONDecodeError', 'JSONDecoder', 'JSONEncoder', '__all__', '__author__', '__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__', '__version__', '_default_decoder', '_default_encoder', 'decoder', 'dump', 'dumps', 'encoder', 'load', 'loads', 'scanner'] ``` 呼叫成員 ``` a = -1 print(a.__abs__) # print(a.__abs__()) # -1 ``` 印出json ``` import json print(json.dumps([1, 2, 3])) # '[1, 2, 3]' ```

JS - 重複組合

給兩個序列 a 跟 b ,在不改變 a 和 b 序列順序的情況下,列出所有 a 跟 b 混合後的結果 使用[重複組合](https://zh.wikipedia.org/wiki/%E7%B5%84%E5%90%88#.E9.87.8D.E8.A4.87.E7.B5.84.E5.90.88.E7.90.86.E8.AB.96.E8.88.87.E5.85.AC.E5.BC.8F)的概念,兩個序列混合後的結果等同於其中一個序列的 index 的組合 舉例來說序列 a 為 `--`,序列 b 為 `**`,其所有混和結果為: ``` --** -*-* *--* -**- *-*- **-- ``` 共六種,而我們可以把 a 序列所在的 index 寫出: ``` 1,2 1,3 2,3 1,4 2,4 3,4 ``` 而這是 C4取2 的所有組合結果。 假設已知 b 序列長度為 6,我們可以使用迴圈生成組合。 ``` var set = { "a":['b','b','b','b'], "b":['3','4','5','6','7','8'] } var n = set['a'].length + set['b'].length; var c = []; for(var i1 = 0 ; i1 < n ; i1 ++) for(var i2 = i1+1 ; i2 < n ; i2 ++) for(var i3 = i2+1 ; i3 < n ; i3 ++) for(var i4 = i3+1 ; i4 < n ; i4 ++) for(var i5 = i4+1 ; i5 < n ; i5 ++) for(var i6 = i5+1 ; i6 < n ; i6 ++){ var ab = []; for(var i = 0 ; i < n ; i ++) ab.push('a'); ab[i1] = 'b'; ab[i2] = 'b'; ab[i3] = 'b'; ab[i4] = 'b'; ab[i5] = 'b'; ab[i6] = 'b'; var c0 = []; var index = {"a":0,"b":0}; for(var i = 0 ; i < n ; i ++){ c0.push(set[ab[i]][index[ab[i]]]); index[ab[i]] = index[ab[i]] + 1; } c.push(c0); console.log(c0); } ```

2017/10/11

在 Windows 10 上安裝 Keras 的流程

markdown Keras 是一個在 Python 上的深度學習工具,用這個工具可以快速打造出複雜的類神經網路結構 他需要用到 TensorFlow,所以在安裝 Keras 之前要先安裝 TensorFlow 在安裝 TensorFlow 之前要先安裝 Python 2.7 或 3.5。 ##先裝 Python 我個人是安裝 Python 3.5,安裝流程請參考 [https://pygame.hackersir.org/Lessons/01/Python_install.html](https://pygame.hackersir.org/Lessons/01/Python_install.html), 正確安裝後,在 cmd 下輸入指令: ``` python -V ``` 若顯示 Python 3.5.2 表示正確安裝。 ##接著裝 TensorFlow 以下介紹cpu版本的安裝,若需要gpu版本的安裝流程請參考[https://rreadmorebooks.blogspot.tw/2017/04/win10cudacudnn.html](https://rreadmorebooks.blogspot.tw/2017/04/win10cudacudnn.html) 在 cmd 下輸入指令: ``` pip3 install --upgrade tensorflow ``` 裝好後,接著輸入 python 進入互動模式: ``` python ``` 再輸入以下的 python 程式碼(可直接複製貼上四行) ``` import tensorflow as tf hello = tf.constant('Hello, TensorFlow!') sess = tf.Session() print(sess.run(hello)) ``` 若顯示 b'Hello, TensorFlow!' 表示正確安裝,此時可輸入 exit() 離開 python 互動模式。 ##SciPy 在裝 Keras 之前,可能需要手動裝 SciPy,因為 SciPy 的自動安裝流程在 Windows 上似乎不完整。 先開啟[http://www.lfd.uci.edu/~gohlke/pythonlibs/#scipy](http://www.lfd.uci.edu/~gohlke/pythonlibs/#scipy), 然後點選 scipy‑1.0.0rc1‑cp35‑cp35m‑win_amd64.whl 下載檔案。 下載好了之後,在 cmd 切換目錄至下載目錄,然後輸入指令: ``` pip3 install scipy‑1.0.0rc1‑cp35‑cp35m‑win_amd64.whl ``` ##最後裝 Keras 直接在 cmd 輸入指令: ``` pip3 install keras ``` 裝好後,接著輸入 python 進入互動模式: ``` python ``` 再輸入以下的 python 程式碼 ``` import keras ``` 若顯示 Using TensorFlow backend. 表示正確安裝,此時可輸入 exit() 離開 python 互動模式。 ##參考連結: 安裝 Python [https://pygame.hackersir.org/Lessons/01/Python_install.html](https://pygame.hackersir.org/Lessons/01/Python_install.html) 安裝 TensorFlow [https://www.tensorflow.org/install/install_windows](https://www.tensorflow.org/install/install_windows) 安裝 SciPy [http://codewenda.com/windows%E4%B8%8A%E4%BD%BF%E7%94%A8pip%E5%AE%89%E8%A3%85scipy%E9%94%99%E8%AF%AF%EF%BC%9Arunning-setup-py-bdist_wheel-for-scipy-error/](http://codewenda.com/windows%E4%B8%8A%E4%BD%BF%E7%94%A8pip%E5%AE%89%E8%A3%85scipy%E9%94%99%E8%AF%AF%EF%BC%9Arunning-setup-py-bdist_wheel-for-scipy-error/) 安裝 Keras [https://keras-cn.readthedocs.io/en/latest/#_2](https://keras-cn.readthedocs.io/en/latest/#_2)

安裝 nokogiri 失敗時的解決方法

(參考此篇文章)[https://github.com/sparklemotion/nokogiri/issues/1483] 在 bash 輸入 xcode-select --install 可解決問題。 在我的 macbook 上的錯誤訊息如下: ``` Gem::Ext::BuildError: ERROR: Failed to build gem native extension. current directory: /Users/etrex/.rvm/gems/ruby-2.4.1@-global/gems/nokogiri-1.8.0/ext/nokogiri /Users/etrex/.rvm/rubies/ruby-2.4.1/bin/ruby -r ./siteconf20171011-80748-pgpdll.rb extconf.rb checking if the C compiler accepts ... yes checking if the C compiler accepts -Wno-error=unused-command-line-argument-hard-error-in-future... no Building nokogiri using packaged libraries. Using mini_portile version 2.2.0 checking for iconv.h... yes checking for gzdopen() in -lz... yes checking for iconv using --with-opt-* flags... yes ************************************************************************ IMPORTANT NOTICE: Building Nokogiri with a packaged version of libxml2-2.9.4 with the following patches applied: - 0001-Fix-comparison-with-root-node-in-xmlXPathCmpNodes.patch - 0002-Fix-XPointer-paths-beginning-with-range-to.patch - 0003-Disallow-namespace-nodes-in-XPointer-ranges.patch Team Nokogiri will keep on doing their best to provide security updates in a timely manner, but if this is a concern for you and want to use the system library instead; abort this installation process and reinstall nokogiri as follows: gem install nokogiri -- --use-system-libraries [--with-xml2-config=/path/to/xml2-config] [--with-xslt-config=/path/to/xslt-config] If you are using Bundler, tell it to use the option: bundle config build.nokogiri --use-system-libraries bundle install Note, however, that nokogiri is not fully compatible with arbitrary versions of libxml2 provided by OS/package vendors. ************************************************************************ Extracting libxml2-2.9.4.tar.gz into tmp/x86_64-apple-darwin15.4.0/ports/libxml2/2.9.4... OK Running git apply with /Users/etrex/.rvm/gems/ruby-2.4.1@-global/gems/nokogiri-1.8.0/patches/libxml2/0001-Fix-comparison-with-root-node-in-xmlXPathCmpNodes.patch... OK Running git apply with /Users/etrex/.rvm/gems/ruby-2.4.1@-global/gems/nokogiri-1.8.0/patches/libxml2/0002-Fix-XPointer-paths-beginning-with-range-to.patch... OK Running git apply with /Users/etrex/.rvm/gems/ruby-2.4.1@-global/gems/nokogiri-1.8.0/patches/libxml2/0003-Disallow-namespace-nodes-in-XPointer-ranges.patch... OK Running 'configure' for libxml2 2.9.4... OK Running 'compile' for libxml2 2.9.4... ERROR, review '/Users/etrex/.rvm/gems/ruby-2.4.1@-global/gems/nokogiri-1.8.0/ext/nokogiri/tmp/x86_64-apple-darwin15.4.0/ports/libxml2/2.9.4/compile.log' to see what happened. Last lines are: ======================================================================== unsigned short* in = (unsigned short*) inb; ^~~~~~~~~~~~~~~~~~~~~ encoding.c:815:27: warning: cast from 'unsigned char *' to 'unsigned short *' increases required alignment from 1 to 2 [-Wcast-align] unsigned short* out = (unsigned short*) outb; ^~~~~~~~~~~~~~~~~~~~~~ 4 warnings generated. CC error.lo CC parserInternals.lo CC parser.lo CC tree.lo CC hash.lo CC list.lo CC xmlIO.lo xmlIO.c:1450:52: error: use of undeclared identifier 'LZMA_OK' ret = (__libxml2_xzclose((xzFile) context) == LZMA_OK ) ? 0 : -1; ^ 1 error generated. make[2]: *** [xmlIO.lo] Error 1 make[1]: *** [all-recursive] Error 1 make: *** [all] Error 2 ======================================================================== *** extconf.rb failed *** Could not create Makefile due to some reason, probably lack of necessary libraries and/or headers. Check the mkmf.log file for more details. You may need configuration options. ```