机器学习应用建议（二）

偏差和方差的判别

高偏差和高方差本质上为学习模型的欠拟合和过拟合问题。

对于高偏差和高方差问题，即学习模型的欠拟合和过拟合问题，我们通常绘制如下图表进行判断：

高偏差——欠拟合问题

• J_train(Θ)误差大
• J_CV(Θ)误差 ≈ J_train(Θ)误差

高方差——过拟合问题

• J_train(Θ)误差小
• J_CV(Θ)误差 >> J_train(Θ)误差

补充笔记

Diagnosing Bias vs. Variance

In this section we examine the relationship between the degree of the polynomial d and the underfitting or overfitting of our hypothesis.

• We need to distinguish whether bias or variance is the problem contributing to bad predictions.
• High bias is underfitting and high variance is overfitting. Ideally, we need to find a golden mean between these two.

The training error will tend to decrease as we increase the degree d of the polynomial.

At the same time, the cross validation error will tend to decrease as we increase d up to a point, and then it will increase as d is increased, forming a convex curve.

High bias (underfitting): both J_train(Θ) and J_CV(Θ) will be high. Also, J_CV(Θ)≈J_train(Θ).

High variance (overfitting): J_train(Θ) will be low and J_CV(Θ) will be much greater than J_train(Θ).

The is summarized in the figure below:

正则化的偏差与方差

在训练模型的过程中，为了避免过拟合问题我们通常使用正则化方法。但对于正则化参数λ的选择，我们是需要谨慎考虑的。

之前，我们在考虑正则化参数λ的选择时，只是考虑单变量的情况。现在，我们要考虑在多项式的情况下，正则化参数λ的取值问题。

例如：对于某一多项式模型，我们使用正则化方法。其中，正则化参数λ=0,0.01,0.02,0.04,0.08,0.16,0.32,0.64,1.28,2.56,5.12,10。现求出最佳的正则化参数λ的值。

首先，我们将数据集分为训练集、交叉验证集和测试集三部分。

然后，当正则化参数λ=0,0.01,0.02,0.04,0.08,0.16,0.32,0.64,1.28,2.56,5.12,10时，我们分别求出J_tran(θ)和J_CV(θ)。

最后，我们利用测试集对J_CV(θ)最小时的某个正则化参数λ值进行计算，求出其J_test(θ)。

图中，假设正则化参数λ=0.08时，J_CV(θ)最小。

为了便于理解，以及便于找到最佳的正则化参数λ的值，我们可以画出下图：

补充笔记

Regularization and Bias/Variance

In the figure above, we see that as λ increases, our fit becomes more rigid. On the other hand, as λ approaches 0, we tend to over overfit the data. So how do we choose our parameter λ to get it 'just right' ? In order to choose the model and the regularization term λ, we need to:

1. Create a list of lambdas (i.e. λ∈{0,0.01,0.02,0.04,0.08,0.16,0.32,0.64,1.28,2.56,5.12,10.24});
2. Create a set of models with different degrees or any other variants.
3. Iterate through the λs and for each λ go through all the models to learn some Θ.
4. Compute the cross validation error using the learned Θ (computed with λ) on the J_CV(Θ) without regularization or λ = 0.
5. Select the best combo that produces the lowest error on the cross validation set.
6. Using the best combo Θ and λ, apply it on J_test(Θ) to see if it has a good generalization of the problem.

学习曲线

通过绘制学习曲线可以帮助我们了解学习算法是否运行正常。学习曲线为训练集误差、交叉验证集误差与训练集样本数量m之间的函数关系图。

上图中，假设函数为h_θ(x) = θ₀ + θ₁x + θ₂x²，且此处不考虑正则化。当m = 1时，我们的假设函数h_θ(x)能完美拟合训练集，其J_train(θ) = 0，但对于交叉验证集而言，假设函数h_θ(x)的泛化能力差，其J_CV(θ)的值将较大；当m=2时，我们的假设函数h_θ能较好地拟合训练集，其J_train(θ)的值将稍微增大，但对于交叉验证集而言，假设函数h_θ(x)的泛化能力依旧较差，其J_CV(θ)的值将较比之前有略微减小；······；但m足够大时，J_train(θ)的值将增大到某一特定值后保持水平，J_CV(θ)的值将减小到某一特定值后保持水平，且J_train(θ)的值与J_CV(θ)的值非常接近。

因此，当学习算法处于高偏差的情况时，我们增加训练集样本数量是毫无用处的。

上图中，我们的假设函数h_θ(x) = θ₀ + θ₁x + θ₂x² + ... + θ₁₀₀x¹⁰⁰，此处考虑正则化，其中正则化参数λ的值很小。当m = 5时，假设函数h_θ(x)能够较好地拟合训练集，其J_train(θ)的值较小，但假设函数h_θ(x)的泛化能力较差，其J_CV(θ)的值较大；当m = 12时，假设函数h_θ(x)依旧能够较好地拟合训练集，但其J_train(θ)的值稍微增大一些，J_CV(θ)的值略微减小一些；······；当m足够大时，J_train(θ)的值逐渐增大，J_CV(θ)的值逐渐减小。

因此，此时学习算法处于高偏差的情况时，我们增加训练集样本数量可能会有些帮助。

注：当m足够大时，J_train(θ)的值逐渐增大，J_CV(θ)的值逐渐减小，这两者是否会相交，视频中尚未交代清楚。

补充笔记

Learning Curves

Training an algorithm on a very few number of data points (such as 1, 2 or 3) will easily have 0 errors because we can always find a quadratic curve that touches exactly those number of points. Hence:

• As the training set gets larger, the error for a quadratic function increases.
• The error value will plateau out after a certain m, or training set size.

Experiencing high bias:

Low training set size: causes J_train(Θ) to be low and J_CV(Θ) to be high.

Large training set size: causes both J_train(Θ) and J_CV(Θ) to be high with J_train(Θ)≈J_CV(Θ).

If a learning algorithm is suffering from high bias, getting more training data will not (by itself) help much.

Experiencing high variance:

Low training set size: J_train(Θ) will be low and J_CV(Θ) will be high.

Large training set size: J_train(Θ) increases with training set size and J_CV(Θ) continues to decrease without leveling off. Also, J_train(Θ) < J_CV(Θ) but the difference between them remains significant.

If a learning algorithm is suffering from high variance, getting more training data is likely to help.

下一步决定做什么

在机器学习应用建议（一）一文的开头，我们就预测结果存在高误差而提出了如下的解决方法：

• 获取更多的样本
• 尝试减少特征变量的数量
• 尝试获取更多的特征变量
• 尝试增加多项式特征
• 尝试减小正则化参数λ的值
• 尝试增大正则化参数λ的值

对于这些方法，我们分别进行了研究得出了如下结论：

• 获取更多的样本——适合高方差（过拟合）问题
• 尝试减少特征变量的数量——适合高方差（过拟合）问题
• 尝试获取更多的特征变量——适合高偏差（欠拟合）问题
• 尝试增加多项式特征——适合高偏差（欠拟合）问题
• 尝试减小正则化参数λ的值——适合高偏差（欠拟合）问题
• 尝试增大正则化参数λ的值 ——适合高方差（过拟合）问题

对于神经网络模型而言，使用“小”的模型，其容易出现高偏差（欠拟合）问题，但其优势在于计算代价较小；使用“大”的模型（即隐藏层激活单元较多或有多个隐藏层。），其容易出现高方差（过拟合）问题，且其计算代价较大。但一般而言，正则化的神经网络模型越“大”其性能越好。

通常我们选择只含有一层隐藏层的神经网络模型。但对于其他情况，只含有一层隐藏层的神经网络模型并不是最优的模型。因此，我们可以将数据集分为训练集、交叉验证集和测试集三部分，分别对隐藏层层数不同的神经网络模型进行训练，找到一个J_CV(Θ)最小的神经网络模型为止。

补充笔记

Deciding What to Do Next Revisited

Our decision process can be broken down as follows:

• Getting more training examples: Fixes high variance
• Trying smaller sets of features: Fixes high variance
• Adding features: Fixes high bias
• Adding polynomial features: Fixes high bias
• Decreasing λ: Fixes high bias
• Increasing λ: Fixes high variance.

Diagnosing Neural Networks

• A neural network with fewer parameters is prone to underfitting. It is also computationally cheaper.
• A large neural network with more parameters is prone to overfitting. It is also computationally expensive. In this case you can use regularization (increase λ) to address the overfitting.

Using a single hidden layer is a good starting default. You can train your neural network on a number of hidden layers using your cross validation set. You can then select the one that performs best.

Model Complexity Effects:

• Lower-order polynomials (low model complexity) have high bias and low variance. In this case, the model fits poorly consistently.
• Higher-order polynomials (high model complexity) fit the training data extremely well and the test data extremely poorly. These have low bias on the training data, but very high variance.
• In reality, we would want to choose a model somewhere in between, that can generalize well but also fits the data reasonably well.