NNに関する、そもそもの疑問

Question

https://qiita.com/takahiro_itazuri/items/d2bea1c643d7cca11352#comment-a59cd26161ee56ea1220
このpythonで書かれた「自分でNNを作ろう」の記事のコードですが、このコードが以下の２種類のデータを用いています。
mnist_dataset/mnist_train.csv
mnist_dataset/mnist_test.csv
で、このいずれも、データの数を３つとかにしてみました。

mnist_test
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,84,185,159,151,60,36,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,222,254,254,254,254,241,198,198,198,198,198,198,198,198,170,52,0,0,0,0,0,0,0,0,0,0,0,0,67,114,72,114,163,227,254,225,254,254,254,250,229,254,254,140,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,17,66,14,67,67,67,59,21,236,254,106,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,83,253,209,18,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,22,・・・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

mnist_train
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,3,18,18,18,126,136,175,26,166,255,247,127,0,0,0,0,0,0,0,0,0,0,0,0,30,36,94,154,170,253,253,253,253,253,225,172,253,242,195,64,0,0,0,0,0,0,0,0,0,0,0,49,238,253,253,253,253,253,253,253,253,251,93,82,82,56,39,0,0,0,0,0,0,0,0,0,0,0,0,18,219,253,253,253,253,253,198,182,247,241,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,80,156,107,253,253,205,11,0,43,154,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,14,1,154,253,90,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,139,253,190,2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,11,190,253,70,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,35,241,225,160,108,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,81,240,253,253,119,25,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,45,186,253,253,150,27,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,16,93,252,253,187,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,249,253,249,64,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,46,130,183,253,253,207,2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,39,148,229,253,253,253,250,182,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,24,114,221,253,253,253,253,201,78,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,23,66,213,253,253,253,253,198,81,2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,18,171,219,253,253,253,253,195,80,9,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,55,172,226,253,253,253,253,244,133,11,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,136,253,253,253,212,135,132,16,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,51,159,253,159,50,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,48,238,252,252,252,237,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,54,227,253,252,239,233,252,57,6,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,10,60,224,252,253,252,202,84,252,253,122,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,163,252,252,252,253,252,252,96,189,253,167,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,51,238,253,253,190,114,253,228,47,79,255,168,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,48,238,252,252,179,12,75,121,21,0,0,253,243,50,0,0,0,0,0,0,0,0,0,0,0,0,0,38,165,253,233,208,84,0,0,0,0,0,0,253,252,165,0,0,0,0,0,0,0,0,0,0,0,0,7,178,252,240,71,19,28,0,0,0,0,0,0,253,252,195,0,0,0,0,0,0,0,0,0,0,0,0,57,252,252,63,0,0,0,0,0,0,0,0,0,253,252,195,0,0,0,0,0,0,0,0,0,0,0,0,198,253,190,0,0,0,0,0,0,0,0,0,0,255,253,196,0,0,0,0,0,0,0,0,0,0,0,76,246,252,112,0,0,0,0,0,0,0,0,0,0,253,252,148,0,0,0,0,0,0,0,0,0,0,0,85,252,230,25,0,0,0,0,0,0,0,0,7,135,253,186,12,0,0,0,0,0,0,0,0,0,0,0,85,252,223,0,0,0,0,0,0,0,0,7,131,252,225,71,0,0,0,0,0,0,0,0,0,0,0,0,85,252,145,0,0,0,0,0,0,0,48,165,252,173,0,0,0,0,0,0,0,0,0,0,0,0,0,0,86,253,225,0,0,0,0,0,0,114,238,253,162,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,85,252,249,146,48,29,85,178,225,253,223,167,56,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,85,252,252,252,229,215,252,252,252,196,130,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,28,199,252,252,253,252,252,233,145,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,25,128,252,253,252,141,37,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
2,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,67,232,39,0,0,0,0,0,0,0,0,0,62,81,0,0,0,0,0,0,0,0,0,0,0,0,0,0,120,180,39,0,0,0,0,0,0,0,0,0,126,163,0,0,0,0,0,0,0,0,0,0,0,0,0,2,153,210,40,0,0,0,0,0,0,0,0,0,220,163,0,0,0,0,0,0,0,0,0,0,0,0,0,27,254,162,0,0,0,0,0,0,0,0,0,0,222,163,0,0,0,0,0,0,0,0,0,0,0,0,0,183,254,125,0,0,0,0,0,0,0,0,0,46,245,163,0,0,0,0,0,0,0,0,0,0,0,0,0,198,254,56,0,0,0,0,0,0,0,0,0,120,254,163,0,0,0,0,0,0,0,0,0,0,0,0,23,231,254,29,0,0,0,0,0,0,0,0,0,159,254,120,0,0,0,0,0,0,0,0,0,0,0,0,163,254,216,16,0,0,0,0,0,0,0,0,0,159,254,67,0,0,0,0,0,0,0,0,0,14,86,178,248,254,91,0,0,0,0,0,0,0,0,0,0,159,254,85,0,0,0,47,49,116,144,150,241,243,234,179,241,252,40,0,0,0,0,0,0,0,0,0,0,150,253,237,207,207,207,253,254,250,240,198,143,91,28,5,233,250,0,0,0,0,0,0,0,0,0,0,0,0,119,177,177,177,177,177,98,56,0,0,0,0,0,102,254,220,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,169,254,137,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,169,254,57,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,169,254,57,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,169,255,94,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,169,254,96,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,169,254,153,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,169,255,153,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,96,254,153,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0

で、実行してみたところ、あまりにも精度が低いです。
また、少し改造したコードで、「全く同じデータ」をtestとtrainに忍ばせて、データ数１０個ぐらいで判別を試したのですが、全然判別できませんでした。
「全く同じデータ」が、データ数１０個とかで判別できないのに、データ数を１０００個に増やしたら判別ができるようになるんでしょうか？またそれはなぜですか？
１０個ですら、同じデータですら判別ができないなら、使い物にならないような気がするのですが。

Comments

[seastar3]

この質問にはpythonブログラムが示されていません。ライブラリを活用するのか、自作のメソッドを活用するのか、いずれにしてもニューラルネットワークの構成はどうなっているのかコードを示しましょう。
そもそも、python以前の機械学習とディープラーニングの質問ですからタグを適切に設定しましょう。これまでの質問もそうですよ。

Answer 1

「全く同じデータ」をtestとtrainに忍ばせて、データ数１０個ぐらいで判別を試したのですが、全然判別できませんでした。

MNISTから適当に取り出して、10クラス識別問題、1クラスあたり1個、testとtrainに同じデータを使う
という問題設定だと3〜4エポックで判別できるようになりましたよ。
ただの確認間違いか、たまたまうまくいかない初期値に当たったか、たまたまうまくいかないデータを選んだか、決して識別できないデータを作ってしまった(?)か、あるいはなにかを決定的に理解していないかだと思いますよ。

(略)
# 直接10個分のデータをリストとして保持
training_data_list = test_data_list = [
[5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,3,18,18,18,126,136,175,
26,166,255,247,127,0,0,0,0,0,0,0,0,0,0,0,0,30,36,94,154,170,253,(略)],
[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,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,51,159,253,159,50,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,48,238,252,
252,252,237,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,54,227,253,(略)],
(略)
]
(略)
    # 学習
    epoch = 10
    for e in range(epoch):
        print('#epoch ', e)
        data_size = len(training_data_list)
        for i in range(data_size):
            val = training_data_list[i]
            idata = (np.asfarray(val[1:]) / 255.0 * 0.99) + 0.01
            tdata = np.zeros(onodes) + 0.01
            tdata[val[0]] = 0.99
            nn.backprop(idata, tdata)

        scoreboard = []
        for val in test_data_list:
            idata = (np.asfarray(val[1:]) / 255.0 * 0.99) + 0.01
            tlabel = val[0]
            predict = nn.feedforward(idata)
            plabel = np.argmax(predict)
            scoreboard.append(tlabel == plabel)

        scoreboard_array = np.asarray(scoreboard)
        print(scoreboard)
        print('performance: ', scoreboard_array.sum() / scoreboard_array.size)

結果例

#epoch  0
[False, False, False, False, False, False, True, True, False, True]
performance:  0.3
#epoch  1
[True, False, True, True, False, True, True, True, True, True]
performance:  0.8
#epoch  2
[True, True, True, True, True, True, True, True, True, True]
performance:  1.0
#epoch  3
[True, True, True, True, True, True, True, True, True, True]
performance:  1.0
#epoch  4
[True, True, True, True, True, True, True, True, True, True]
performance:  1.0
#epoch  5
[True, True, True, True, True, True, True, True, True, True]
performance:  1.0
#epoch  6
[True, True, True, True, True, True, True, True, True, True]
performance:  1.0
#epoch  7
[True, True, True, True, True, True, True, True, True, True]
performance:  1.0
#epoch  8
[True, True, True, True, True, True, True, True, True, True]
performance:  1.0
#epoch  9
[True, True, True, True, True, True, True, True, True, True]
performance:  1.0

---

質問のmnist_test mnist_train は同じデータを忍ばせてないから、mnist_testを識別可能な学習ができるわけがないですね。（上の回答では無視しましたが）

3クラス識別問題で学習データが1クラスあたり1個だと、極端な話、どこか2個の値を見てどっちも1なら"0"、どっちも0なら"1"、こっちが0でこっちが1なら"2"、みたいにたまたま選んだ学習データに依存している識別可能な箇所があって当たり前で、それを見つけるだけで終わってしまう。
ニューラルネットで学習させてもたまたま当てられるという程度のパラメータが見つかったらもう学習が止まって（なぜなら当てられる=誤差がなくなるのだから）未知のデータは全然当てられないものしかできない。
というだけの話。

Comments

[退会済みユーザー]

1や2じゃだめなんですかね。

[quickquip]

3クラス識別問題、1クラスあたり1個、testとtrainに同じデータを使う
は1〜2エポックで学習できましたよ。

[jbpb0]

質問者さん

> 「全く同じデータ」をtestとtrainに忍ばせて、データ数１０個ぐらいで判別を試したのですが、全然判別できませんでした。

「忍ばせる」(一部のデータだけ同じにする)のではなくて、全部同じにしたら判別できませんか？

Answer 2

NNの動作原理をちゃんと理解されてから進めべきだと思います。

とりあえず、epochを10000とかにしてやってみるとどうでしょうね？
同じ3つのデータで10000回学習させるという意味ですよ。