Machine learning cheat sheet

CS 229 – Machine Learning https://stanford.edu/~shervine
Super VIP Cheatsheet: Machine Learning
Afshine Amidi and Shervine Amidi
September 15, 2018
Contents
1 Supervised Learning 2
1.1 Introduction to Supervised Learning . . . . . . . . . . . . . . . . . . . 2
1.2 Notations and general concepts . . . . . . . . . . . . . . . . . . . . . 2
1.3 Linear models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3.1 Linear regression . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3.2 Classiﬁcation and logistic regression . . . . . . . . . . . . . . . 3
1.3.3 Generalized Linear Models . . . . . . . . . . . . . . . . . . . . 3
1.4 Support Vector Machines . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Generative Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5.1 Gaussian Discriminant Analysis . . . . . . . . . . . . . . . . . 4
1.5.2 Naive Bayes . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.6 Tree-based and ensemble methods . . . . . . . . . . . . . . . . . . . . 4
1.7 Other non-parametric approaches . . . . . . . . . . . . . . . . . . . . 4
1.8 Learning Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Unsupervised Learning 6
2.1 Introduction to Unsupervised Learning . . . . . . . . . . . . . . . . . 6
2.2 Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 Expectation-Maximization . . . . . . . . . . . . . . . . . . . . 6
2.2.2 k-means clustering . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.3 Hierarchical clustering . . . . . . . . . . . . . . . . . . . . . . 6
2.2.4 Clustering assessment metrics . . . . . . . . . . . . . . . . . . 6
2.3 Dimension reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3.1 Principal component analysis . . . . . . . . . . . . . . . . . . 7
2.3.2 Independent component analysis . . . . . . . . . . . . . . . . . 7
3 Deep Learning 8
3.1 Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Convolutional Neural Networks . . . . . . . . . . . . . . . . . . . . . 8
3.3 Recurrent Neural Networks . . . . . . . . . . . . . . . . . . . . . . . 8
3.4 Reinforcement Learning and Control . . . . . . . . . . . . . . . . . . . 9
4 Machine Learning Tips and Tricks 10
4.1 Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.1.1 Classiﬁcation . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.1.2 Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.2 Model selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.3 Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5 Refreshers 12
5.1 Probabilities and Statistics . . . . . . . . . . . . . . . . . . . . . . . . 12
5.1.1 Introduction to Probability and Combinatorics . . . . . . . . . 12
5.1.2 Conditional Probability . . . . . . . . . . . . . . . . . . . . . 12
5.1.3 Random Variables . . . . . . . . . . . . . . . . . . . . . . . . 13
5.1.4 Jointly Distributed Random Variables . . . . . . . . . . . . . . 13
5.1.5 Parameter estimation . . . . . . . . . . . . . . . . . . . . . . 14
5.2 Linear Algebra and Calculus . . . . . . . . . . . . . . . . . . . . . . . 14
5.2.1 General notations . . . . . . . . . . . . . . . . . . . . . . . . 14
5.2.2 Matrix operations . . . . . . . . . . . . . . . . . . . . . . . . 15
5.2.3 Matrix properties . . . . . . . . . . . . . . . . . . . . . . . . 15
5.2.4 Matrix calculus . . . . . . . . . . . . . . . . . . . . . . . . . 16
Stanford University 1 Fall 2018

CS 229 – Machine Learning Shervine Amidi & Afshine Amidi
1 Supervised Learning
1.1 Introduction to Supervised Learning
Given a set of data points {x(1), ..., x(m)} associated to a set of outcomes {y(1), ..., y(m)}, we
want to build a classifier that learns how to predict y from x.
Ì Type of prediction – The different types of predictive models are summed up in the table
below:
Regression Classifier
Outcome Continuous Class
Examples Linear regression Logistic regression, SVM, Naive Bayes
Ì Type of model – The different models are summed up in the table below:
Discriminative model Generative model
Goal Directly estimate P(y|x) Estimate P(x|y) to deduce P(y|x)
What’s learned Decision boundary Probability distributions of the data
Illustration
Examples Regressions, SVMs GDA, Naive Bayes
1.2 Notations and general concepts
Ì Hypothesis – The hypothesis is noted hθ and is the model that we choose. For a given input
data x(i), the model prediction output is hθ(x(i)).
Ì Loss function – A loss function is a function L : (z,y) ∈ R × Y −→ L(z,y) ∈ R that takes as
inputs the predicted value z corresponding to the real data value y and outputs how different
they are. The common loss functions are summed up in the table below:
Least squared Logistic Hinge Cross-entropy
1
2
(y − z)2
log(1 + exp(−yz)) max(0,1 − yz) − y log(z) + (1 − y) log(1 − z)
Linear regression Logistic regression SVM Neural Network
Ì Cost function – The cost function J is commonly used to assess the performance of a model,
and is defined with the loss function L as follows:
J(θ) =
m
i=1
L(hθ(x(i)
), y(i)
)
Ì Gradient descent – By noting α ∈ R the learning rate, the update rule for gradient descent
is expressed with the learning rate and the cost function J as follows:
θ ←− θ − α J(θ)
Remark: Stochastic gradient descent (SGD) is updating the parameter based on each training
example, and batch gradient descent is on a batch of training examples.
Ì Likelihood – The likelihood of a model L(θ) given parameters θ is used to find the optimal
parameters θ through maximizing the likelihood. In practice, we use the log-likelihood (θ) =
log(L(θ)) which is easier to optimize. We have:
θopt
= arg max
θ
L(θ)
Ì Newton’s algorithm – The Newton’s algorithm is a numerical method that finds θ such
that (θ) = 0. Its update rule is as follows:
θ ← θ −
(θ)
(θ)
Remark: the multidimensional generalization, also known as the Newton-Raphson method, has
the following update rule:
θ ← θ − 2
θ (θ)
−1
θ (θ)
1.3 Linear models
1.3.1 Linear regression
We assume here that y|x; θ ∼ N(µ,σ2)
Ì Normal equations – By noting X the matrix design, the value of θ that minimizes the cost
function is a closed-form solution such that:
θ = (XT
X)−1
XT
y

Ì LMS algorithm – By noting α the learning rate, the update rule of the Least Mean Squares
(LMS) algorithm for a training set of m data points, which is also known as the Widrow-Hoff
learning rule, is as follows:
∀j, θj ← θj + α
m
i=1
y(i)
− hθ(x(i)
) x
(i)
j
Remark: the update rule is a particular case of the gradient ascent.
Ì LWR – Locally Weighted Regression, also known as LWR, is a variant of linear regression that
weights each training example in its cost function by w(i)(x), which is defined with parameter
τ ∈ R as:
w(i)
(x) = exp −
(x(i) − x)2
2τ2
1.3.2 Classification and logistic regression
Ì Sigmoid function – The sigmoid function g, also known as the logistic function, is defined
as follows:
∀z ∈ R, g(z) =
1
1 + e−z
∈]0,1[
Ì Logistic regression – We assume here that y|x; θ ∼ Bernoulli(φ). We have the following
form:
φ = p(y = 1|x; θ) =
1
1 + exp(−θT x)
= g(θT
x)
Remark: there is no closed form solution for the case of logistic regressions.
Ì Softmax regression – A softmax regression, also called a multiclass logistic regression, is
used to generalize logistic regression when there are more than 2 outcome classes. By convention,
we set θK = 0, which makes the Bernoulli parameter φi of each class i equal to:
φi =
exp(θT
i x)
K
j=1
exp(θT
j x)
1.3.3 Generalized Linear Models
Ì Exponential family – A class of distributions is said to be in the exponential family if it can
be written in terms of a natural parameter, also called the canonical parameter or link function,
η, a sufficient statistic T(y) and a log-partition function a(η) as follows:
p(y; η) = b(y) exp(ηT(y) − a(η))
Remark: we will often have T(y) = y. Also, exp(−a(η)) can be seen as a normalization param-
eter that will make sure that the probabilities sum to one.
Here are the most common exponential distributions summed up in the following table:
Distribution η T(y) a(η) b(y)
Bernoulli log φ
1−φ
y log(1 + exp(η)) 1
Gaussian µ y η2
2
1√
2π
exp −y2
2
Poisson log(λ) y eη 1
y!
Geometric log(1 − φ) y log eη
1−eη 1
Ì Assumptions of GLMs – Generalized Linear Models (GLM) aim at predicting a random
variable y as a function fo x ∈ Rn+1 and rely on the following 3 assumptions:
(1) y|x; θ ∼ ExpFamily(η) (2) hθ(x) = E[y|x; θ] (3) η = θT
x
Remark: ordinary least squares and logistic regression are special cases of generalized linear
models.
1.4 Support Vector Machines
The goal of support vector machines is to find the line that maximizes the minimum distance to
the line.
Ì Optimal margin classifier – The optimal margin classifier h is such that:
h(x) = sign(wT
x − b)
where (w, b) ∈ Rn × R is the solution of the following optimization problem:
min
1
2
||w||2
such that y(i)
(wT
x(i)
− b) 1
Remark: the line is defined as wT
x − b = 0 .
Ì Hinge loss – The hinge loss is used in the setting of SVMs and is defined as follows:
L(z,y) = [1 − yz]+ = max(0,1 − yz)

Ì Kernel – Given a feature mapping φ, we define the kernel K to be defined as:
K(x,z) = φ(x)T
φ(z)
In practice, the kernel K defined by K(x,z) = exp −
||x−z||2
2σ2 is called the Gaussian kernel
and is commonly used.
Remark: we say that we use the "kernel trick" to compute the cost function using the kernel
because we actually don’t need to know the explicit mapping φ, which is often very complicated.
Instead, only the values K(x,z) are needed.
Ì Lagrangian – We define the Lagrangian L(w,b) as follows:
L(w,b) = f(w) +
l
i=1
βihi(w)
Remark: the coefficients βi are called the Lagrange multipliers.
1.5 Generative Learning
A generative model first tries to learn how the data is generated by estimating P(x|y), which
we can then use to estimate P(y|x) by using Bayes’ rule.
1.5.1 Gaussian Discriminant Analysis
Ì Setting – The Gaussian Discriminant Analysis assumes that y and x|y = 0 and x|y = 1 are
such that:
y ∼ Bernoulli(φ)
x|y = 0 ∼ N(µ0,Σ) and x|y = 1 ∼ N(µ1,Σ)
Ì Estimation – The following table sums up the estimates that we find when maximizing the
likelihood:
φ µj (j = 0,1) Σ
1
m
m
i=1
1{y(i)=1}
m
i=1
1{y(i)=j}x(i)
m
i=1
1{y(i)=j}
1
m
m
i=1
(x(i)
− µy(i) )(x(i)
− µy(i) )T
1.5.2 Naive Bayes
Ì Assumption – The Naive Bayes model supposes that the features of each data point are all
independent:
P(x|y) = P(x1,x2,...|y) = P(x1|y)P(x2|y)... =
n
i=1
P(xi|y)
Ì Solutions – Maximizing the log-likelihood gives the following solutions, with k ∈ {0,1},
l ∈ [[1,L]]
P(y = k) =
1
m
× #{j|y(j)
= k} and P(xi = l|y = k) =
#{j|y(j) = k and x
(j)
i = l}
#{j|y(j) = k}
Remark: Naive Bayes is widely used for text classification and spam detection.
1.6 Tree-based and ensemble methods
These methods can be used for both regression and classification problems.
Ì CART – Classification and Regression Trees (CART), commonly known as decision trees,
can be represented as binary trees. They have the advantage to be very interpretable.
Ì Random forest – It is a tree-based technique that uses a high number of decision trees
built out of randomly selected sets of features. Contrary to the simple decision tree, it is highly
uninterpretable but its generally good performance makes it a popular algorithm.
Remark: random forests are a type of ensemble methods.
Ì Boosting – The idea of boosting methods is to combine several weak learners to form a
stronger one. The main ones are summed up in the table below:
Adaptive boosting Gradient boosting
- High weights are put on errors to - Weak learners trained
improve at the next boosting step on remaining errors
- Known as Adaboost
1.7 Other non-parametric approaches
Ì k-nearest neighbors – The k-nearest neighbors algorithm, commonly known as k-NN, is a
non-parametric approach where the response of a data point is determined by the nature of its
k neighbors from the training set. It can be used in both classification and regression settings.
Remark: The higher the parameter k, the higher the bias, and the lower the parameter k, the
higher the variance.

1.8 Learning Theory
Ì Union bound – Let A1, ..., Ak be k events. We have:
P(A1 ∪ ... ∪ Ak) P(A1) + ... + P(Ak)
Ì Hoeffding inequality – Let Z1, .., Zm be m iid variables drawn from a Bernoulli distribution
of parameter φ. Let φ be their sample mean and γ > 0 fixed. We have:
P(|φ − φ| > γ) 2 exp(−2γ2
m)
Remark: this inequality is also known as the Chernoff bound.
Ì Training error – For a given classifier h, we define the training error (h), also known as the
empirical risk or empirical error, to be as follows:
(h) =
1
m
m
i=1
1{h(x(i))=y(i)}
Ì Probably Approximately Correct (PAC) – PAC is a framework under which numerous
results on learning theory were proved, and has the following set of assumptions:
• the training and testing sets follow the same distribution
• the training examples are drawn independently
Ì Shattering – Given a set S = {x(1),...,x(d)}, and a set of classifiers H, we say that H shatters
S if for any set of labels {y(1), ..., y(d)}, we have:
∃h ∈ H, ∀i ∈ [[1,d]], h(x(i)
) = y(i)
Ì Upper bound theorem – Let H be a finite hypothesis class such that |H| = k and let δ and
the sample size m be fixed. Then, with probability of at least 1 − δ, we have:
(h) min
h∈H
(h) + 2
1
2m
log
2k
δ
Ì VC dimension – The Vapnik-Chervonenkis (VC) dimension of a given infinite hypothesis
class H, noted VC(H) is the size of the largest set that is shattered by H.
Remark: the VC dimension of H = {set of linear classifiers in 2 dimensions} is 3.
Ì Theorem (Vapnik) – Let H be given, with VC(H) = d and m the number of training
examples. With probability at least 1 − δ, we have:
(h) min
h∈H
(h) + O
d
m
log
m
d
+
1
m
log
1
δ

2 Unsupervised Learning
2.1 Introduction to Unsupervised Learning
Ì Motivation – The goal of unsupervised learning is to find hidden patterns in unlabeled data
{x(1),...,x(m)}.
Ì Jensen’s inequality – Let f be a convex function and X a random variable. We have the
following inequality:
E[f(X)] f(E[X])
2.2 Clustering
2.2.1 Expectation-Maximization
Ì Latent variables – Latent variables are hidden/unobserved variables that make estimation
problems difficult, and are often denoted z. Here are the most common settings where there are
latent variables:
Setting Latent variable z x|z Comments
Mixture of k Gaussians Multinomial(φ) N(µj,Σj) µj ∈ Rn, φ ∈ Rk
Factor analysis N(0,I) N(µ + Λz,ψ) µj ∈ Rn
Ì Algorithm – The Expectation-Maximization (EM) algorithm gives an efficient method at
estimating the parameter θ through maximum likelihood estimation by repeatedly constructing
a lower-bound on the likelihood (E-step) and optimizing that lower bound (M-step) as follows:
• E-step: Evaluate the posterior probability Qi(z(i)) that each data point x(i) came from
a particular cluster z(i) as follows:
Qi(z(i)
) = P(z(i)
|x(i)
; θ)
• M-step: Use the posterior probabilities Qi(z(i)) as cluster specific weights on data points
x(i) to separately re-estimate each cluster model as follows:
θi = argmax
θ
i
ˆ
z(i)
Qi(z(i)
) log
P(x(i),z(i); θ)
Qi(z(i))
dz(i)
2.2.2 k-means clustering
We note c(i) the cluster of data point i and µj the center of cluster j.
Ì Algorithm – After randomly initializing the cluster centroids µ1,µ2,...,µk ∈ Rn, the k-means
algorithm repeats the following step until convergence:
c(i)
= arg min
j
||x(i)
− µj||2
and µj =
m
i=1
1{c(i)=j}x(i)
m
i=1
1{c(i)=j}
Ì Distortion function – In order to see if the algorithm converges, we look at the distortion
function defined as follows:
J(c,µ) =
m
i=1
||x(i)
− µc(i) ||2
2.2.3 Hierarchical clustering
Ì Algorithm – It is a clustering algorithm with an agglomerative hierarchical approach that
build nested clusters in a successive manner.
Ì Types – There are different sorts of hierarchical clustering algorithms that aims at optimizing
different objective functions, which is summed up in the table below:
Ward linkage Average linkage Complete linkage
Minimize within cluster Minimize average distance Minimize maximum distance
distance between cluster pairs of between cluster pairs
2.2.4 Clustering assessment metrics
In an unsupervised learning setting, it is often hard to assess the performance of a model since
we don’t have the ground truth labels as was the case in the supervised learning setting.
Ì Silhouette coefficient – By noting a and b the mean distance between a sample and all
other points in the same class, and between a sample and all other points in the next nearest
cluster, the silhouette coefficient s for a single sample is defined as follows:
s =
b − a
max(a,b)

Ì Calinski-Harabaz index – By noting k the number of clusters, Bk and Wk the between
and within-clustering dispersion matrices respectively defined as
Bk =
k
j=1
nc(i) (µc(i) − µ)(µc(i) − µ)T
, Wk =
m
i=1
(x(i)
− µc(i) )(x(i)
− µc(i) )T
the Calinski-Harabaz index s(k) indicates how well a clustering model defines its clusters, such
that the higher the score, the more dense and well separated the clusters are. It is defined as
follows:
s(k) =
Tr(Bk)
Tr(Wk)
×
N − k
k − 1
2.3 Dimension reduction
2.3.1 Principal component analysis
It is a dimension reduction technique that finds the variance maximizing directions onto which
to project the data.
Ì Eigenvalue, eigenvector – Given a matrix A ∈ Rn×n, λ is said to be an eigenvalue of A if
there exists a vector z ∈ Rn{0}, called eigenvector, such that we have:
Az = λz
Ì Spectral theorem – Let A ∈ Rn×n. If A is symmetric, then A is diagonalizable by a real
orthogonal matrix U ∈ Rn×n. By noting Λ = diag(λ1,...,λn), we have:
∃Λ diagonal, A = UΛUT
Remark: the eigenvector associated with the largest eigenvalue is called principal eigenvector of
matrix A.
Ì Algorithm – The Principal Component Analysis (PCA) procedure is a dimension reduction
technique that projects the data on k dimensions by maximizing the variance of the data as
follows:
• Step 1: Normalize the data to have a mean of 0 and standard deviation of 1.
x
(i)
j ←
x
(i)
j − µj
σj
where µj =
1
m
m
i=1
x
(i)
j and σ2
j =
1
m
m
i=1
(x
(i)
j − µj)2
• Step 2: Compute Σ =
1
m
m
i=1
x(i)
x(i)T
∈ Rn×n
, which is symmetric with real eigenvalues.
• Step 3: Compute u1, ..., uk ∈ Rn the k orthogonal principal eigenvectors of Σ, i.e. the
orthogonal eigenvectors of the k largest eigenvalues.
• Step 4: Project the data on spanR(u1,...,uk). This procedure maximizes the variance
among all k-dimensional spaces.
2.3.2 Independent component analysis
It is a technique meant to find the underlying generating sources.
Ì Assumptions – We assume that our data x has been generated by the n-dimensional source
vector s = (s1,...,sn), where si are independent random variables, via a mixing and non-singular
matrix A as follows:
x = As
The goal is to find the unmixing matrix W = A−1 by an update rule.
Ì Bell and Sejnowski ICA algorithm – This algorithm finds the unmixing matrix W by
following the steps below:
• Write the probability of x = As = W−1s as:
p(x) =
n
i=1
ps(wT
i x) · |W|
• Write the log likelihood given our training data {x(i), i ∈ [[1,m]]} and by noting g the
sigmoid function as:
l(W) =
m
i=1
n
j=1
log g (wT
j x(i)
) + log |W|
Therefore, the stochastic gradient ascent learning rule is such that for each training example
x(i), we update W as follows:
W ←− W + α






1 − 2g(wT
1 x(i))
1 − 2g(wT
2 x(i))
...
1 − 2g(wT
n x(i))


x(i)T
+ (WT
)−1




3 Deep Learning
3.1 Neural Networks
Neural networks are a class of models that are built with layers. Commonly used types of neural
networks include convolutional and recurrent neural networks.
Ì Architecture – The vocabulary around neural networks architectures is described in the
figure below:
By noting i the ith layer of the network and j the jth hidden unit of the layer, we have:
z
[i]
j = w
[i]
j
T
x + b
[i]
j
where we note w, b, z the weight, bias and output respectively.
Ì Activation function – Activation functions are used at the end of a hidden unit to introduce
non-linear complexities to the model. Here are the most common ones:
Sigmoid Tanh ReLU Leaky ReLU
g(z) =
1
1 + e−z
g(z) =
ez − e−z
ez + e−z
g(z) = max(0,z) g(z) = max( z,z)
with 1
Ì Cross-entropy loss – In the context of neural networks, the cross-entropy loss L(z,y) is
commonly used and is defined as follows:
L(z,y) = − y log(z) + (1 − y) log(1 − z)
Ì Learning rate – The learning rate, often noted η, indicates at which pace the weights get
updated. This can be fixed or adaptively changed. The current most popular method is called
Adam, which is a method that adapts the learning rate.
Ì Backpropagation – Backpropagation is a method to update the weights in the neural network
by taking into account the actual output and the desired output. The derivative with respect
to weight w is computed using chain rule and is of the following form:
∂L(z,y)
∂w
=
∂L(z,y)
∂a
×
∂a
∂z
×
∂z
∂w
As a result, the weight is updated as follows:
w ←− w − η
∂L(z,y)
∂w
Ì Updating weights – In a neural network, weights are updated as follows:
• Step 1: Take a batch of training data.
• Step 2: Perform forward propagation to obtain the corresponding loss.
• Step 3: Backpropagate the loss to get the gradients.
• Step 4: Use the gradients to update the weights of the network.
Ì Dropout – Dropout is a technique meant at preventing overfitting the training data by
dropping out units in a neural network. In practice, neurons are either dropped with probability
p or kept with probability 1 − p.
3.2 Convolutional Neural Networks
Ì Convolutional layer requirement – By noting W the input volume size, F the size of the
convolutional layer neurons, P the amount of zero padding, then the number of neurons N that
fit in a given volume is such that:
N =
W − F + 2P
S
+ 1
Ì Batch normalization – It is a step of hyperparameter γ, β that normalizes the batch {xi}.
By noting µB, σ2
B the mean and variance of that we want to correct to the batch, it is done as
follows:
xi ←− γ
xi − µB
σ2
B +
+ β
It is usually done after a fully connected/convolutional layer and before a non-linearity layer and
aims at allowing higher learning rates and reducing the strong dependence on initialization.
3.3 Recurrent Neural Networks
Ì Types of gates – Here are the different types of gates that we encounter in a typical recurrent
neural network:
Input gate Forget gate Output gate Gate
Write to cell or not? Erase a cell or not? Reveal a cell or not? How much writing?
Ì LSTM – A long short-term memory (LSTM) network is a type of RNN model that avoids
the vanishing gradient problem by adding ’forget’ gates.

3.4 Reinforcement Learning and Control
The goal of reinforcement learning is for an agent to learn how to evolve in an environment.
Ì Markov decision processes – A Markov decision process (MDP) is a 5-tuple (S,A,{Psa},γ,R)
where:
• S is the set of states
• A is the set of actions
• {Psa} are the state transition probabilities for s ∈ S and a ∈ A
• γ ∈ [0,1[ is the discount factor
• R : S × A −→ R or R : S −→ R is the reward function that the algorithm wants to
maximize
Ì Policy – A policy π is a function π : S −→ A that maps states to actions.
Remark: we say that we execute a given policy π if given a state s we take the action a = π(s).
Ì Value function – For a given policy π and a given state s, we deﬁne the value function V π
as follows:
V π
(s) = E R(s0) + γR(s1) + γ2
R(s2) + ...|s0 = s,π
Ì Bellman equation – The optimal Bellman equations characterizes the value function V π∗
of the optimal policy π∗:
V π∗
(s) = R(s) + max
a∈A
γ
s ∈S
Psa(s )V π∗
(s )
Remark: we note that the optimal policy π∗ for a given state s is such that:
π∗
(s) = argmax
a∈A
s ∈S
Psa(s )V ∗
(s )
Ì Value iteration algorithm – The value iteration algorithm is in two steps:
• We initialize the value:
V0(s) = 0
• We iterate the value based on the values before:
Vi+1(s) = R(s) + max
a∈A
s ∈S
γPsa(s )Vi(s )
Ì Maximum likelihood estimate – The maximum likelihood estimates for the state transition
probabilities are as follows:
Psa(s ) =
#times took action a in state s and got to s
#times took action a in state s
Ì Q-learning – Q-learning is a model-free estimation of Q, which is done as follows:
Q(s,a) ← Q(s,a) + α R(s,a,s ) + γ max
a
Q(s ,a ) − Q(s,a)

4 Machine Learning Tips and Tricks
4.1 Metrics
Given a set of data points {x(1), ..., x(m)}, where each x(i) has n features, associated to a set of
outcomes {y(1), ..., y(m)}, we want to assess a given classifier that learns how to predict y from
x.
4.1.1 Classification
In a context of a binary classification, here are the main metrics that are important to track to
assess the performance of the model.
Ì Confusion matrix – The confusion matrix is used to have a more complete picture when
assessing the performance of a model. It is defined as follows:
Predicted class
+ –
Actual class
TP FN
+ False Negatives
True Positives
Type II error
FP TN
– False Positives
True Negatives
Type I error
Ì Main metrics – The following metrics are commonly used to assess the performance of
classification models:
Metric Formula Interpretation
Accuracy
TP + TN
TP + TN + FP + FN
Overall performance of model
Precision
TP
TP + FP
How accurate the positive predictions are
Recall
TP
TP + FN
Coverage of actual positive sample
Sensitivity
Specificity
TN
TN + FP
Coverage of actual negative sample
F1 score
2TP
2TP + FP + FN
Hybrid metric useful for unbalanced classes
Ì ROC – The receiver operating curve, also noted ROC, is the plot of TPR versus FPR by
varying the threshold. These metrics are are summed up in the table below:
Metric Formula Equivalent
True Positive Rate
TP
TP + FN
Recall, sensitivity
TPR
False Positive Rate
FP
TN + FP
1-specificity
FPR
Ì AUC – The area under the receiving operating curve, also noted AUC or AUROC, is the
area below the ROC as shown in the following figure:
4.1.2 Regression
Ì Basic metrics – Given a regression model f, the following metrics are commonly used to
assess the performance of the model:
Total sum of squares Explained sum of squares Residual sum of squares
SStot =
m
i=1
(yi − y)2
SSreg =
m
i=1
(f(xi) − y)2
SSres =
m
i=1
(yi − f(xi))2
Ì Coefficient of determination – The coefficient of determination, often noted R2 or r2,
provides a measure of how well the observed outcomes are replicated by the model and is defined
as follows:
R2
= 1 −
SSres
SStot
Ì Main metrics – The following metrics are commonly used to assess the performance of
regression models, by taking into account the number of variables n that they take into consid-
eration:
Mallow’s Cp AIC BIC Adjusted R2
SSres + 2(n + 1)σ2
m
2 (n + 2) − log(L) log(m)(n + 2) − 2 log(L) 1 −
(1 − R2)(m − 1)
m − n − 1
where L is the likelihood and σ2 is an estimate of the variance associated with each response.

4.2 Model selection
Ì Vocabulary – When selecting a model, we distinguish 3 different parts of the data that we
have as follows:
Training set Validation set Testing set
- Model is trained - Model is assessed - Model gives predictions
- Usually 80% of the dataset - Usually 20% of the dataset - Unseen data
- Also called hold-out
or development set
Once the model has been chosen, it is trained on the entire dataset and tested on the unseen
test set. These are represented in the figure below:
Ì Cross-validation – Cross-validation, also noted CV, is a method that is used to select a
model that does not rely too much on the initial training set. The different types are summed
up in the table below:
k-fold Leave-p-out
- Training on k − 1 folds and - Training on n − p observations and
assessment on the remaining one assessment on the p remaining ones
- Generally k = 5 or 10 - Case p = 1 is called leave-one-out
The most commonly used method is called k-fold cross-validation and splits the training data
into k folds to validate the model on one fold while training the model on the k − 1 other folds,
all of this k times. The error is then averaged over the k folds and is named cross-validation
error.
Ì Regularization – The regularization procedure aims at avoiding the model to overfit the
data and thus deals with high variance issues. The following table sums up the different types
of commonly used regularization techniques:
LASSO Ridge Elastic Net
- Shrinks coefficients to 0 Makes coefficients smaller Tradeoff between variable
- Good for variable selection selection and small coefficients
... + λ||θ||1 ... + λ||θ||2
2 ... + λ (1 − α)||θ||1 + α||θ||2
2
λ ∈ R λ ∈ R λ ∈ R, α ∈ [0,1]
Ì Model selection – Train model on training set, then evaluate on the development set, then
pick best performance model on the development set, and retrain all of that model on the whole
training set.
4.3 Diagnostics
Ì Bias – The bias of a model is the difference between the expected prediction and the correct
model that we try to predict for given data points.
Ì Variance – The variance of a model is the variability of the model prediction for given data
points.
Ì Bias/variance tradeoff – The simpler the model, the higher the bias, and the more complex
the model, the higher the variance.
Underfitting Just right Overfitting
- High training error - Training error - Low training error
Symptoms - Training error close slightly lower than - Training error much
to test error test error lower than test error
- High bias - High variance
Regression

Classification
Deep learning
- Complexify model - Regularize
Remedies - Add more features - Get more data
- Train longer
Ì Error analysis – Error analysis is analyzing the root cause of the difference in performance
between the current and the perfect models.
Ì Ablative analysis – Ablative analysis is analyzing the root cause of the difference in perfor-
mance between the current and the baseline models.
5 Refreshers
5.1 Probabilities and Statistics
5.1.1 Introduction to Probability and Combinatorics
Ì Sample space – The set of all possible outcomes of an experiment is known as the sample
space of the experiment and is denoted by S.
Ì Event – Any subset E of the sample space is known as an event. That is, an event is a set
consisting of possible outcomes of the experiment. If the outcome of the experiment is contained
in E, then we say that E has occurred.
Ì Axioms of probability – For each event E, we denote P(E) as the probability of event E
occuring. By noting E1,...,En mutually exclusive events, we have the 3 following axioms:
(1) 0 P(E) 1 (2) P(S) = 1 (3) P
n
i=1
Ei =
n
i=1
P(Ei)
Ì Permutation – A permutation is an arrangement of r objects from a pool of n objects, in a
given order. The number of such arrangements is given by P(n, r), defined as:
P(n, r) =
n!
(n − r)!
Ì Combination – A combination is an arrangement of r objects from a pool of n objects, where
the order does not matter. The number of such arrangements is given by C(n, r), defined as:
C(n, r) =
P(n, r)
r!
=
n!
r!(n − r)!
Remark: we note that for 0 r n, we have P(n,r) C(n,r).
5.1.2 Conditional Probability
Ì Bayes’ rule – For events A and B such that P(B) > 0, we have:
P(A|B) =
P(B|A)P(A)
P(B)
Remark: we have P(A ∩ B) = P(A)P(B|A) = P(A|B)P(B).
Ì Partition – Let {Ai, i ∈ [[1,n]]} be such that for all i, Ai = ∅. We say that {Ai} is a partition
if we have:
∀i = j, Ai ∩ Aj = ∅ and
n
i=1
Ai = S
Remark: for any event B in the sample space, we have P(B) =
n
i=1
P(B|Ai)P(Ai).

Ì Extended form of Bayes’ rule – Let {Ai, i ∈ [[1,n]]} be a partition of the sample space.
We have:
P(Ak|B) =
P(B|Ak)P(Ak)
n
i=1
P(B|Ai)P(Ai)
Ì Independence – Two events A and B are independent if and only if we have:
P(A ∩ B) = P(A)P(B)
5.1.3 Random Variables
Ì Random variable – A random variable, often noted X, is a function that maps every element
in a sample space to a real line.
Ì Cumulative distribution function (CDF) – The cumulative distribution function F,
which is monotonically non-decreasing and is such that lim
x→−∞
F(x) = 0 and lim
x→+∞
F(x) = 1, is
deﬁned as:
F(x) = P(X x)
Remark: we have P(a < X B) = F(b) − F(a).
Ì Probability density function (PDF) – The probability density function f is the probability
that X takes on values between two adjacent realizations of the random variable.
Ì Relationships involving the PDF and CDF – Here are the important properties to know
in the discrete (D) and the continuous (C) cases.
Case CDF F PDF f Properties of PDF
(D) F(x) =
xi x
P(X = xi) f(xj) = P(X = xj) 0 f(xj) 1 and
j
f(xj) = 1
(C) F(x) =
ˆ x
−∞
f(y)dy f(x) =
dF
dx
f(x) 0 and
ˆ +∞
−∞
f(x)dx = 1
Ì Variance – The variance of a random variable, often noted Var(X) or σ2, is a measure of the
spread of its distribution function. It is determined as follows:
Var(X) = E[(X − E[X])2
] = E[X2
] − E[X]2
Ì Standard deviation – The standard deviation of a random variable, often noted σ, is a
measure of the spread of its distribution function which is compatible with the units of the
actual random variable. It is determined as follows:
σ = Var(X)
Ì Expectation and Moments of the Distribution – Here are the expressions of the expected
value E[X], generalized expected value E[g(X)], kth moment E[Xk] and characteristic function
ψ(ω) for the discrete and continuous cases:
Case E[X] E[g(X)] E[Xk] ψ(ω)
(D)
n
i=1
xif(xi)
n
i=1
g(xi)f(xi)
n
i=1
xk
i f(xi)
n
i=1
f(xi)eiωxi
(C)
ˆ +∞
−∞
xf(x)dx
ˆ +∞
−∞
g(x)f(x)dx
ˆ +∞
−∞
xk
f(x)dx
ˆ +∞
−∞
f(x)eiωx
dx
Remark: we have eiωx = cos(ωx) + i sin(ωx).
Ì Revisiting the kth moment – The kth moment can also be computed with the characteristic
function as follows:
E[Xk
] =
1
ik
∂kψ
∂ωk
ω=0
Ì Transformation of random variables – Let the variables X and Y be linked by some
function. By noting fX and fY the distribution function of X and Y respectively, we have:
fY (y) = fX (x)
dx
dy
Ì Leibniz integral rule – Let g be a function of x and potentially c, and a, b boundaries that
may depend on c. We have:
∂
∂c
ˆ b
a
g(x)dx =
∂b
∂c
· g(b) −
∂a
∂c
· g(a) +
ˆ b
a
∂g
∂c
(x)dx
Ì Chebyshev’s inequality – Let X be a random variable with expected value µ and standard
deviation σ. For k, σ > 0, we have the following inequality:
P(|X − µ| kσ)
1
k2
5.1.4 Jointly Distributed Random Variables
Ì Conditional density – The conditional density of X with respect to Y , often noted fX|Y ,
is deﬁned as follows:
fX|Y (x) =
fXY (x,y)
fY (y)
Ì Independence – Two random variables X and Y are said to be independent if we have:
fXY (x,y) = fX (x)fY (y)

Ì Marginal density and cumulative distribution – From the joint density probability
function fXY , we have:
Case Marginal density Cumulative function
(D) fX (xi) =
j
fXY (xi,yj) FXY (x,y) =
xi x yj y
fXY (xi,yj)
(C) fX (x) =
ˆ +∞
−∞
fXY (x,y)dy FXY (x,y) =
ˆ x
−∞
ˆ y
−∞
fXY (x ,y )dx dy
Ì Distribution of a sum of independent random variables – Let Y = X1 + ... + Xn with
X1, ..., Xn independent. We have:
ψY (ω) =
n
k=1
ψXk
(ω)
Ì Covariance – We define the covariance of two random variables X and Y , that we note σ2
XY
or more commonly Cov(X,Y ), as follows:
Cov(X,Y ) σ2
XY = E[(X − µX )(Y − µY )] = E[XY ] − µX µY
Ì Correlation – By noting σX , σY the standard deviations of X and Y , we define the correlation
between the random variables X and Y , noted ρXY , as follows:
ρXY =
σ2
XY
σX σY
Remarks: For any X, Y , we have ρXY ∈ [−1,1]. If X and Y are independent, then ρXY = 0.
Ì Main distributions – Here are the main distributions to have in mind:
Type Distribution PDF ψ(ω) E[X] Var(X)
X ∼ B(n, p) P(X = x) =
n
x
px
qn−x
(peiω + q)n np npq
Binomial x ∈ [[0,n]]
(D)
X ∼ Po(µ) P(X = x) =
µx
x!
e−µ
eµ(eiω
−1) µ µ
Poisson x ∈ N
X ∼ U(a, b) f(x) =
1
b − a
eiωb − eiωa
(b − a)iω
a + b
2
(b − a)2
12
Uniform x ∈ [a,b]
(C) X ∼ N(µ, σ) f(x) =
1
√
2πσ
e
− 1
2
x−µ
σ
2
eiωµ− 1
2
ω2
σ2
µ σ2
Gaussian x ∈ R
X ∼ Exp(λ) f(x) = λe−λx 1
1 − iω
λ
1
λ
1
λ2
Exponential x ∈ R+
5.1.5 Parameter estimation
Ì Random sample – A random sample is a collection of n random variables X1, ..., Xn that
are independent and identically distributed with X.
Ì Estimator – An estimator ˆθ is a function of the data that is used to infer the value of an
unknown parameter θ in a statistical model.
Ì Bias – The bias of an estimator ˆθ is defined as being the difference between the expected
value of the distribution of ˆθ and the true value, i.e.:
Bias(ˆθ) = E[ˆθ] − θ
Remark: an estimator is said to be unbiased when we have E[ˆθ] = θ.
Ì Sample mean and variance – The sample mean and the sample variance of a random
sample are used to estimate the true mean µ and the true variance σ2 of a distribution, are
noted X and s2 respectively, and are such that:
X =
1
n
n
i=1
Xi and s2
= ˆσ2
=
1
n − 1
n
i=1
(Xi − X)2
Ì Central Limit Theorem – Let us have a random sample X1, ..., Xn following a given
distribution with mean µ and variance σ2, then we have:
X ∼
n→+∞
N µ,
σ
√
n
5.2 Linear Algebra and Calculus
5.2.1 General notations
Ì Vector – We note x ∈ Rn a vector with n entries, where xi ∈ R is the ith entry:
x =
x1
x2
...
xn
∈ Rn
Ì Matrix – We note A ∈ Rm×n a matrix with m rows and n columns, where Ai,j ∈ R is the
entry located in the ith row and jth column:
A =
A1,1 · · · A1,n
...
...
Am,1 · · · Am,n
∈ Rm×n
Remark: the vector x defined above can be viewed as a n × 1 matrix and is more particularly
called a column-vector.
Ì Identity matrix – The identity matrix I ∈ Rn×n is a square matrix with ones in its diagonal
and zero everywhere else:
I =



1 0 · · · 0
0
...
...
...
...
...
... 0
0 · · · 0 1




Remark: for all matrices A ∈ Rn×n, we have A × I = I × A = A.
Ì Diagonal matrix – A diagonal matrix D ∈ Rn×n is a square matrix with nonzero values in
its diagonal and zero everywhere else:
D =



d1 0 · · · 0
0
...
...
...
...
...
... 0
0 · · · 0 dn



Remark: we also note D as diag(d1,...,dn).
5.2.2 Matrix operations
Ì Vector-vector multiplication – There are two types of vector-vector products:
• inner product: for x,y ∈ Rn, we have:
xT
y =
n
i=1
xiyi ∈ R
• outer product: for x ∈ Rm, y ∈ Rn, we have:
xyT
=
x1y1 · · · x1yn
...
...
xmy1 · · · xmyn
∈ Rm×n
Ì Matrix-vector multiplication – The product of matrix A ∈ Rm×n and vector x ∈ Rn is a
vector of size Rn, such that:
Ax =


aT
r,1x
...
aT
r,mx

 =
n
i=1
ac,ixi ∈ Rn
where aT
r,i are the vector rows and ac,j are the vector columns of A, and xi are the entries
of x.
Ì Matrix-matrix multiplication – The product of matrices A ∈ Rm×n and B ∈ Rn×p is a
matrix of size Rn×p, such that:
AB =


aT
r,1bc,1 · · · aT
r,1bc,p
...
...
aT
r,mbc,1 · · · aT
r,mbc,p

 =
n
i=1
ac,ibT
r,i ∈ Rn×p
where aT
r,i, bT
r,i are the vector rows and ac,j, bc,j are the vector columns of A and B respec-
tively.
Ì Transpose – The transpose of a matrix A ∈ Rm×n, noted AT , is such that its entries are
ﬂipped:
∀i,j, AT
i,j = Aj,i
Remark: for matrices A,B, we have (AB)T = BT AT .
Ì Inverse – The inverse of an invertible square matrix A is noted A−1 and is the only matrix
such that:
AA−1
= A−1
A = I
Remark: not all square matrices are invertible. Also, for matrices A,B, we have (AB)−1 =
B−1A−1
Ì Trace – The trace of a square matrix A, noted tr(A), is the sum of its diagonal entries:
tr(A) =
n
i=1
Ai,i
Remark: for matrices A,B, we have tr(AT ) = tr(A) and tr(AB) = tr(BA)
Ì Determinant – The determinant of a square matrix A ∈ Rn×n, noted |A| or det(A) is
expressed recursively in terms of Ai,j, which is the matrix A without its ith row and jth
column, as follows:
det(A) = |A| =
n
j=1
(−1)i+j
Ai,j|Ai,j|
Remark: A is invertible if and only if |A| = 0. Also, |AB| = |A||B| and |AT | = |A|.
5.2.3 Matrix properties
Ì Symmetric decomposition – A given matrix A can be expressed in terms of its symmetric
and antisymmetric parts as follows:
A =
A + AT
2
Symmetric
+
A − AT
2
Antisymmetric
Ì Norm – A norm is a function N : V −→ [0, + ∞[ where V is a vector space, and such that
for all x,y ∈ V , we have:
• N(x + y) N(x) + N(y)
• N(ax) = |a|N(x) for a scalar
• if N(x) = 0, then x = 0
For x ∈ V , the most commonly used norms are summed up in the table below:

Norm Notation Definition Use case
Manhattan, L1 ||x||1
n
i=1
|xi| LASSO regularization
Euclidean, L2 ||x||2
n
i=1
x2
i Ridge regularization
p-norm, Lp ||x||p
n
i=1
xp
i
1
p
Hölder inequality
Infinity, L∞ ||x||∞ max
i
|xi| Uniform convergence
Ì Linearly dependence – A set of vectors is said to be linearly dependent if one of the vectors
in the set can be defined as a linear combination of the others.
Remark: if no vector can be written this way, then the vectors are said to be linearly independent.
Ì Matrix rank – The rank of a given matrix A is noted rank(A) and is the dimension of the
vector space generated by its columns. This is equivalent to the maximum number of linearly
independent columns of A.
Ì Positive semi-definite matrix – A matrix A ∈ Rn×n is positive semi-definite (PSD) and
is noted A 0 if we have:
A = AT
and ∀x ∈ Rn
, xT
Ax 0
Remark: similarly, a matrix A is said to be positive definite, and is noted A 0, if it is a PSD
matrix which satisfies for all non-zero vector x, xT Ax > 0.
Ì Eigenvalue, eigenvector – Given a matrix A ∈ Rn×n, λ is said to be an eigenvalue of A if
there exists a vector z ∈ Rn{0}, called eigenvector, such that we have:
Az = λz
Ì Spectral theorem – Let A ∈ Rn×n. If A is symmetric, then A is diagonalizable by a real
orthogonal matrix U ∈ Rn×n. By noting Λ = diag(λ1,...,λn), we have:
∃Λ diagonal, A = UΛUT
Ì Singular-value decomposition – For a given matrix A of dimensions m × n, the singular-
value decomposition (SVD) is a factorization technique that guarantees the existence of U m×m
unitary, Σ m × n diagonal and V n × n unitary matrices, such that:
A = UΣV T
5.2.4 Matrix calculus
Ì Gradient – Let f : Rm×n → R be a function and A ∈ Rm×n be a matrix. The gradient of f
with respect to A is a m × n matrix, noted Af(A), such that:
Af(A)
i,j
=
∂f(A)
∂Ai,j
Remark: the gradient of f is only defined when f is a function that returns a scalar.
Ì Hessian – Let f : Rn → R be a function and x ∈ Rn be a vector. The hessian of f with
respect to x is a n × n symmetric matrix, noted 2
xf(x), such that:
2
xf(x)
i,j
=
∂2f(x)
∂xi∂xj
Remark: the hessian of f is only defined when f is a function that returns a scalar.
Ì Gradient operations – For matrices A,B,C, the following gradient properties are worth
having in mind:
Atr(AB) = BT
AT f(A) = ( Af(A))T
Atr(ABAT
C) = CAB + CT
ABT
A|A| = |A|(A−1
)T

Machine learning cheat sheet

More Related Content

Similar to Machine learning cheat sheet

More from Hany Sewilam Abdel Hamid

Recently uploaded

Machine learning cheat sheet