Pembelajaran mendalam telah menjadi alat penting dalam bidang kecerdasan buatan Berasal daripada penyelidikan rangkaian saraf tiruan, perceptron berbilang lapisan dengan berbilang lapisan tersembunyi ialah struktur pembelajaran mendalam. Proses mencari parameter berat dan berat sebelah lapisan tersembunyi sering dirujuk sebagai proses "pembelajaran" Prinsip asas yang diikuti adalah untuk meminimumkan ralat keluaran akhir rangkaian. Dalam rangkaian saraf, fungsi pengaktifan adalah salah satu daripada banyak parameter yang mesti dipilih supaya rangkaian saraf memperoleh hasil dan prestasi yang optimum.
Apakah fungsi pengaktifan yang biasa digunakan? Bagaimana untuk memilih?
Fungsi pengaktifan ialah fungsi yang berjalan pada neuron rangkaian saraf tiruan Ia bertanggungjawab untuk memetakan input neuron kepada fungsi pengaktifan Rangkaian menukarkan jumlah isyarat input kepada isyarat keluaran. Kebanyakan fungsi pengaktifan adalah fungsi tak linear, yang boleh menukar output perceptron berbilang lapisan kepada tak linear, supaya rangkaian saraf boleh sewenang-wenangnya menghampiri mana-mana fungsi tak linear, dan boleh digunakan pada banyak model tak linear.
Dalam erti kata lain, fungsi pengaktifan tak linear boleh mencipta hubungan pemetaan yang kompleks antara kunci input dan output, dan rangkaian saraf juga boleh mengemas kini parameter melalui "pembelajaran". Selain itu, kerana derivatif fungsi tak linear berkaitan dengan input, kecerunan boleh dikira melalui algoritma perambatan ke belakang, dan rangkaian saraf berbilang lapisan juga boleh dibina untuk menangani masalah yang kompleks.
Fungsi pengaktifan biasa termasuk siri sigmoid untuk rangkaian cetek, siri ReLU untuk rangkaian dalam, siri tanh dan siri Softmax untuk rangkaian rekursif, dsb.
Fungsi sigmoid juga dipanggil fungsi Logistik Ia digunakan untuk output neuron lapisan tersembunyi Ia boleh memetakan nilai (−∞, +∞) kepada selang (0,1) Fungsi ini biasanya digunakan apabila menyatakan nilai ramalan dalam bentuk kemungkinan. Kod Python bagi fungsi pengaktifan sigmod adalah seperti berikut:
import numpy <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">as</span> np<br><br>def sigmoid<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>s <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">1</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">/</span> <span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">1</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">+</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span><span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>return s
Imej fungsi adalah seperti berikut:
Kelebihan Sigmoid fungsi ialah ia boleh dibezakan, dan nilai Domain adalah antara 0 dan 1, yang menormalkan output neuron dan merupakan fungsi pengaktifan terawal yang digunakan dalam rangkaian saraf. Kelemahannya juga jelas Apabila ia meningkat atau menurun ke tahap tertentu, nilai fungsi berubah sangat sedikit Ini adalah apa yang dipanggil "penghilangan kecerunan", yang menyebabkan rangkaian berkumpul dengan perlahan dan menggunakan sumber pengkomputeran. Juga, nilai output tidak berpusat pada 0, tetapi pada 0.5.
Fungsi Sigmoid umum digunakan pada rangkaian cetek.
Berdasarkan Sigmoid, terdapat juga HardSigmoid, kerana apabila nilai input cenderung kepada infiniti, nilai output cenderung kepada 1; apabila nilai input cenderung kepada infinitesimal, nilai output cenderung kepada 0. Oleh itu, seperti namanya, HardSigmoid adalah berdasarkan Sigmoid dan secara paksa menetapkan 1 dan 0 apabila nilai input melebihi julat tertentu. Kod python HardSigmoid adalah seperti berikut:
def Hard_sigmoid<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>y <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> <span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">[</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">]</span><br>for i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">in</span> x<span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>if i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);"> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">2.5</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>y_i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">0</span><br>elif i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">>=</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">2.5</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">and</span> i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);"> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">2.5</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>y_i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">0.2</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">*</span> i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">+</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">0.5</span><br>else<span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>y_i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">1</span><br>y<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.append</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>y_i<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>return y</span></span>
Imej fungsi fungsi pengaktifan HardSigmoid adalah seperti berikut:
Ungkapan swish ialah : f ( x ) = x ⋅ s i g m o i d ( b x ), kod python adalah seperti berikut:
def Swish<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>return x <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">/</span> <span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">1</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">+</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span><span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span>b<span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">*</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span>
di mana b ialah parameter yang boleh dipelajari, Swish mempunyai ciri-ciri tiada sempadan atas dan sempadan bawah, kelancaran dan tidak monoton.
Swish berfungsi lebih baik daripada ReLU pada model dalam. Sebagai contoh, hanya menggantikan ReLU dengan unit Swish boleh meningkatkan ketepatan klasifikasi NASNetA Mudah Alih pada ImageNet sebanyak 0.9%.
Maxout boleh dilihat sebagai menambah lapisan fungsi pengaktifan pada rangkaian pembelajaran mendalam, termasuk parameter k. Berbanding dengan ReLU, sigmoid, dsb., lapisan ini istimewa kerana ia menambah k neuron dan kemudian mengeluarkan nilai dengan nilai pengaktifan terbesar.
maksout ialah penganggar fungsi Untuk rangkaian MLP standard, jika terdapat cukup neuron dalam lapisan tersembunyi, ia secara teorinya boleh menganggarkan sebarang fungsi. Maxout mempunyai keupayaan pemasangan yang sangat kuat dan boleh memuatkan mana-mana fungsi cembung Ia mempunyai semua kelebihan ReLU, lineariti dan ketidaktepuan, tetapi tidak mempunyai beberapa kelemahan ReLU, seperti kematian neuron.
Relu (Rectified Linear Unit)称为“线性整流函数”或者“修正线性单元”,通常就直接称为 ReLU 函数,是解决梯度消失问题的方法。将 ReLU 函数引入神经网络时,也引入了很大的稀疏性。然而,由于稀疏性,时间和空间复杂度更低,不涉及成本更高的指数运算,允许网络快速收敛。
尽管Relu看起来像线性函数,但它具有导数函数并允许反向传播,python 代码如下:
import numpy <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">as</span> np<br><br>def relu<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>s <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.where</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);"> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">0</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">0</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span> x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>return s</span>
ReLU引入了神经元死亡问题,当输入接近零或为负时,函数的梯度变为零,网络将无法执行反向传播,也无法学习,也就是说,网络的大部分分量都永远不会更新,另外,它不能避免梯度爆炸问题。
ReLU是现在DNN模型中比较常用的激活函数。
指数线性单元激活函数ELU解决了 ReLU 的一些问题,同时也保留了一些好的方面。这种激活函数要选取一个 α 值;常见的取值是在 0.1 到 0.3 之间。但α =0.3时的函数图像如下:
ELU能避免神经元死亡问题,能得到负值输出,这能帮助网络向正确的方向推动权重和偏置变化,在计算梯度时能得到激活,而不是让它们等于 0。ELU 的python 代码如下:
import numpy <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">as</span> np<br><br>def elu<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>s <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.where</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">>=</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">0</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span> x<span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span> α<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">1</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>return s
但是,由于包含了指数运算,计算时间更长,同样无法避免梯度爆炸问题,另外,神经网络不学习 α 值。
渗漏型整流线性单元激活函数也有一个 α 值,通常取值在 0.1 到 0.3 之间。Leaky ReLU 激活函数很常用,相比于 ELU 也有一些缺陷,但比 ReLU 具有一些优势。
LeakyReLU的负值斜率很小,而不是平坦的斜率。斜率系数需要在训练前确定,即在训练过程中不学习。这种类型的激活函数在可能遇到稀疏梯度的任务中很流行,例如训练生成式对抗网络。
import numpy <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">as</span> np<br><br>def lrelu<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>s <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.where</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">>=</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">0</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span> x<span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span> αx<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>return s
类似 ELU,Leaky ReLU 也能避免死亡 ReLU 问题,因为其在计算导数时允许较小的梯度,由于不包含指数运算,所以计算速度比 ELU 快。
SELU 激活能够对神经网络进行自归一化,归一化就是首先减去均值,然后除以标准差。因此,经过归一化之后,网络的组件(权重、偏置和激活)的均值为 0,标准差为 1,而这正是 SELU 激活函数的输出值。通过归一化,网络参数会被初始化一个正态分布。
通过归一化,网络参数会被初始化一个正态分布。
def SeLU<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span>alpha<span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">1.6732632423543772848170429916717</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span>scale<span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">1.0507009873554804934193349852946</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>y <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> <span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">[</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">]</span><br>for i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">in</span> x<span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>if i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">>=</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">0</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>y_i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> scale <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">*</span> i<br>else<span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>y_i <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> scale <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">*</span> alpha <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">*</span> <span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>i<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">1</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>y<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.append</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>y_i<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>return y
SELU内部归一化的速度比外部归一化快,这意味着网络能更快收敛,而且避免了出现梯度消失或爆炸问题,在CNN或RNN 网络架构中有所应用。
GELU是某些函数(比如双曲正切函数 tanh)与近似数值的组合,
当 x 大于 0 时,输出为 x;但 x=0 到 x=1 的区间除外,这时曲线更偏向于 y 轴。
import numpy <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">as</span> np<br>def tanh<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>s1 <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span><span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>s2 <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">+</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span><span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>s <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> s1 <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">/</span> s2<br>return s<br>gelu <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> lambda x<span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">0.5</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">*</span> x <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">*</span> <span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">1</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">+</span> tanh<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.sqrt</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">2</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">/</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.pi</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">*</span> <span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">+</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">0.044715</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">*</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.power</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span> <span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">3</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span>
GELU 在NLP 领域有较好表现,尤其在 Transformer 模型中表现最好,能避免梯度消失问题。
Tanh函数,即双曲正切函数,比sigmoid函数更受欢迎,能为多层神经网络提供更好的性能。
它的输出更多地以零为中心,这有助于加速收敛,尤其是在训练初期。双曲线正切函数的python代码如下:
import numpy <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">as</span> np<br><br>def tanh<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>s1 <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span><span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>s2 <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span> <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">+</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span><span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">-</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>s <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> s1 <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">/</span> s2<br>return s
Tanh函数的最大优点是输出值以 0为中心,即关于坐标原点对称,分属为正数和负数两大类别,函数及其导数都是单调的,收敛速度比sigmoid快,从而可以减少迭代次数。这使得它具有了Sigmoid函数的优势,又克服了某些不足。但是,“梯度消失”的问题都还存在,进而导致收敛速度变慢。
Tanh 一般用于递归神经网络。
Hardtanh激活函数是Tanh的线性分段近似。相较而言,它更易计算,这使得学习计算的速度更快,尽管首次派生值为零可能导致静默神经元/过慢的学习速率。
基于Tanh之上,计算输入输出的差值,即为TanhShrink,函数图像如下。
在当输入在0附近时,梯度近乎为0,而在输入极大或极小时,梯度反而为正常梯度。
Softmax函数比较适合作为多分类模型的激活函数,一般会与交叉熵损失函数相配。
通常,Softmax函数只应用于输出层,把一堆实数的值映射到0-1区间,并且使他们的和为1,可以理解为对应每个类别对应的预测概率。python代码如下:
def softmax<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">:</span><br>x_exp <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.exp</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x<span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>x_sum <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> np<span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">.sum</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">(</span>x_exp<span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span> axis<span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span><span style="color: rgb(0, 92, 197); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">1</span><span style="color: rgb(89, 89, 89); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">,</span> keepdims<span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span><span style="color: rgb(153, 0, 85); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">True</span><span style="color: rgb(153, 153, 119); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">)</span><br>s <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">=</span> x_exp <span style="color: rgb(215, 58, 73); margin: 0px; padding: 0px; background: none 0% 0% / auto repeat scroll padding-box border-box rgba(0, 0, 0, 0);">/</span> x_sum<br>return s
如果某一个zj大过其他z,那这个映射的分量就逼近于1,其他就逼近于0。
Softmax函数用于将输入进行归一化到(0,1),并且其和为1,普遍应用于分类模型(互斥)的预测概率值。事实上,但凡涉及到概率的地方基本都会用到softmax,典型的就比如attention layer当中,都会使用softmax来计算attention值。
LogSoftMax
LogSoftmax是基于Softmax函数之上,计算其对应的对数值,范围在(-∞,0)用来计算交叉熵损失函数(根据groundtruth的标签取出对应的值即可)。LogSoftMax 加快了运算速度,提高数据稳定性。
Softmin
Softmin是在Softmax的基础上,做相反变换。Softmin是在Softmax的基础上,做相反变换。 跟softmax类似,输入n维t数据,对它们进行重新缩放使得n维输出的每个元素都在[0, 1]区间内,且和为1。不同的是,softmax是单调递增而softmin是单调递减,意味着softmax操作会使得最大的值在激活操作后依然保持最大,而softmin会使得最小的数在经过了softmin后变成最大值。
以终为始,激活函数的选择也是为最终的任务目标服务的。不存在普遍适用各种神经网络的万能的激活函数,在选择激活函数的时候,要考虑不同的条件限制,例如,如果函数可导,求导数的计算难度如何?函数光滑程度如何?输出是否保持标准化?网络的收敛速度如何?等等。
一般地,在用于分类器时,Sigmoid函数及其组合通常效果更好。为了避免梯度消失问题,又需要避免使用Sigmoid和TanH。如果是回归模型,在输出层上可以使用线性激活函数。如果是浅层神经网络,如不超过4层的,可选择使用多种激励函数,没有太大的影响。如果网络中存在大量未激活神经元,可以考虑leaky ReLU函数。
ReLU函数是应用比较广泛的激活函数,可以作为默认选项。深度学习往往需要大量时间来处理大量数据,模型的收敛速度是尤为重要的所以要尽量选择输出具有zero-centered特点的激活函数以加快模型的收敛速度。
一个经验上的建议是:SELU > ELU > Leaky ReLU > ReLU> tanh > sigmoid,但是,如果网络的体系结构阻止自归一化,那么 ELU 可能是比 SELU 更好的选择。如果速度很重要,Leaky ReLU 将是比慢很多的 ELU 更好的选择。
更重要的是,激活函数仍在发展,需要跟踪业界的最新进展,并勇于探索和创新。
Fungsi pengaktifan ialah parameter penting dalam rangkaian saraf Secara amnya, siri Sigmoid digunakan untuk lapisan keluaran dua tugas klasifikasi, siri softmax digunakan untuk lapisan keluaran. tugas berbilang klasifikasi, dan siri tanh digunakan untuk model Lapisan tersembunyi, siri Relu digunakan untuk tugas regresi serta lapisan tersembunyi rangkaian saraf konvolusi. Tetapi tiada jaminan, dan penyelidikan baharu tentang fungsi pengaktifan masih muncul.
Lampiran, terdapat gambar fungsi pengaktifan pada reddit, ia agak menarik!
Atas ialah kandungan terperinci Fungsi pengaktifan biasa dalam rangkaian saraf. Untuk maklumat lanjut, sila ikut artikel berkaitan lain di laman web China PHP!