A collection of classical ML equations in Latex . Some of them are provided with simple notes and paper link. Hopes to help writings such as papers and blogs.
Better viewed at https://blmoistawinde.github.io/ml_equations_latex/
encoder hidden state at time step
, with input token embedding
decoder hidden state at time step
, with input token embedding
h_t = RNN_{enc}(x_t, h_{t-1})
s_t = RNN_{dec}(y_t, s_{t-1})
The ,
are usually either
-
LSTM (paper: Long short-term memory)
-
GRU (paper: Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation).
The attention weight , the
th decoder step over the
th encoder step, resulting in context vector
c_i = \sum_{j=1}^{T_x} \alpha_{ij}h_j
\alpha_{ij} = \frac{\exp(e_{ij})}{\sum_{k=1}^{T_x} \exp(e_{ik})}
e_{ij} = a(s_{i-1}, h_j)
is an specific attention function, which can be
Paper: Neural Machine Translation by Jointly Learning to Align and Translate
e_{ij} = v^T tanh(W[s_{i-1}; h_j])
Paper: Effective Approaches to Attention-based Neural Machine Translation
If and
has same number of dimension.
otherwise
e_{ij} = s_{i-1}^T h_j
e_{ij} = s_{i-1}^T W h_j
Finally, the output is produced by:
s_t = tanh(W[s_{t-1};y_t;c_t])
o_t = softmax(Vs_t)
Paper: Attention Is All You Need
Attention(Q, K, V) = softmax(\frac{QK^T}{\sqrt{d_k}})V
where is the dimension of the key vector
and query vector
.
where
MultiHead(Q, K, V) = Concat(head_1, ..., head_h)W^O
head_i = Attention(Q W^Q_i, K W^K_i, V W^V_i)
Paper: Generative Adversarial Networks
\min_{G}\max_{D}V(D, G) = \mathbb{E}_{x\sim p_{\text{data}}(x)}[\log{D(x)}] + \mathbb{E}_{z\sim p_{\text{z}}(z)}[\log{(1-D(G(z)))}]
Paper: Auto-Encoding Variational Bayes
To produce a latent variable z such that , we sample
, than z is produced by
z \sim q_{\mu, \sigma}(z) = \mathcal{N}(\mu, \sigma^2)
\epsilon \sim \mathcal{N}(0,1)
z = \mu + \epsilon \cdot \sigma
Above is for 1-D case. For a multi-dimensional (vector) case we use:
\epsilon \sim \mathcal{N}(0, \textbf{I})
\vec{z} \sim \mathcal{N}(\vec{\mu}, \sigma^2 \textbf{I})
Related to Logistic Regression. For single-label/multi-label binary classification.
\sigma(z) = \frac{1} {1 + e^{-z}}
tanh(x) = \frac{e^x - e^{-x}}{e^x + e^{-x}} = \frac{1 - e^{-2x}}{1 + e^{-2x}}
For multi-class single label classification.
\sigma(z_i) = \frac{e^{z_{i}}}{\sum_{j=1}^K e^{z_{j}}} \ \ \ for\ i=1,2,\dots,K
Relu(z) = max(0, z)
where is the cumulative distribution function of Gaussian distribution.
Gelu(x) = x\Phi(x)
Below and
are
dimensional vectors, and
denotes the value on the
th dimension of
.
\sum_{i=1}^{D}|x_i-y_i|
\sum_{i=1}^{D}(x_i-y_i)^2
It’s less sensitive to outliers than the MSE as it treats error as square only inside an interval.
L_{\delta}=
\left\{\begin{matrix}
\frac{1}{2}(y - \hat{y})^{2} & if \left | (y - \hat{y}) \right | < \delta\\
\delta ((y - \hat{y}) - \frac1 2 \delta) & otherwise
\end{matrix}\right.
- In binary classification, where the number of classes
equals 2, Binary Cross-Entropy(BCE) can be calculated as:
- If
(i.e. multiclass classification), we calculate a separate loss for each class label per observation and sum the result.
-{(y\log(p) + (1 - y)\log(1 - p))}
-\sum_{c=1}^My_{o,c}\log(p_{o,c})
M - number of classes
log - the natural log
y - binary indicator (0 or 1) if class label c is the correct classification for observation o
p - predicted probability observation o is of class c
Minimizing negative loglikelihood
is equivalent to Maximum Likelihood Estimation(MLE).
Here is a scaler instead of vector. It is the value of the single dimension where the ground truth
lies. It is thus equivalent to cross entropy (See wiki).\
NLL(y) = -{\log(p(y))}
\min_{\theta} \sum_y {-\log(p(y;\theta))}
\max_{\theta} \prod_y p(y;\theta)
Used in Support Vector Machine(SVM).
max(0, 1 - y \cdot \hat{y})
KL(\hat{y} || y) = \sum_{c=1}^{M}\hat{y}_c \log{\frac{\hat{y}_c}{y_c}}
JS(\hat{y} || y) = \frac{1}{2}(KL(y||\frac{y+\hat{y}}{2}) + KL(\hat{y}||\frac{y+\hat{y}}{2}))
The below can be any of the above loss.
A regression model that uses L1 regularization technique is called Lasso Regression.
Loss = Error(Y - \widehat{Y}) + \lambda \sum_1^n |w_i|
A regression model that uses L1 regularization technique is called Ridge Regression.
Loss = Error(Y - \widehat{Y}) + \lambda \sum_1^n w_i^{2}
Some of them overlaps with loss, like MAE, KL-divergence.
Accuracy = \frac{TP+TN}{TP+TN+FP+FN}
Precision = \frac{TP}{TP+FP}
Recall = \frac{TP}{TP+FN}
F1 = \frac{2*Precision*Recall}{Precision+Recall} = \frac{2*TP}{2*TP+FP+FN}
Sensitivity = Recall = \frac{TP}{TP+FN}
Specificity = \frac{TN}{FP+TN}
AUC is calculated as the Area Under the (TPR)-
(FPR) Curve.
MAE, MSE, equation above.
The Mutual Information is a measure of the similarity between two labels of the same data. Where is the number of the samples in cluster
and
is the number of the samples in cluster
, the Mutual Information between cluster
and
is given as:
MI(U,V)=\sum_{i=1}^{|U|} \sum_{j=1}^{|V|} \frac{|U_i\cap V_j|}{N}
\log\frac{N|U_i \cap V_j|}{|U_i||V_j|}
Normalized Mutual Information (NMI) is a normalization of the Mutual Information (MI) score to scale the results between 0 (no mutual information) and 1 (perfect correlation). In this function, mutual information is normalized by some generalized mean of H(labels_true) and H(labels_pred)), See wiki.
Skip RI, ARI for complexity.
Also skip metrics for related tasks (e.g. modularity for community detection[graph clustering], coherence score for topic modeling[soft clustering]).
Skip nDCG (Normalized Discounted Cumulative Gain) for its complexity.
Average Precision is calculated as:
\text{AP} = \sum_n (R_n - R_{n-1}) P_n
where and
are the precision and recall at the
th threshold.
AP can also be regarded as the area under the precision-recall curve.
MAP is the mean of AP over all the queries.
Cosine(x,y) = \frac{x \cdot y}{|x||y|}
Similarity of two sets and
.
Jaccard(U,V) = \frac{|U \cap V|}{|U \cup V|}
Relevance of two events and
.
PMI(x;y) = \log{\frac{p(x,y)}{p(x)p(y)}}
For example, and
is the frequency of word
and
appearing in corpus and
is the frequency of the co-occurrence of the two.
This repository now only contains simple equations for ML. They are mainly about deep learning and NLP now due to personal research interests.
For time issues, elegant equations in traditional ML approaches like SVM, SVD, PCA, LDA are not included yet.
Moreover, there is a trend towards more complex metrics, which have to be calculated with complicated program (e.g. BLEU, ROUGE, METEOR), iterative algorithms (e.g. PageRank), optimization (e.g. Earth Mover Distance), or even learning based (e.g. BERTScore). They thus cannot be described using simple equations.
https://blog.floydhub.com/gans-story-so-far/
https://ermongroup.github.io/cs228-notes/extras/vae/
Thanks for a-rodin's solution to show Latex in Github markdown, which I have wrapped into latex2pic.py.