Towards a Coherent Notation

[jeroarenas]

We need to adapt all notebooks to follow the following notation criteria

• k for the samples indexes (k=0, ..., K-1)

• i for the components (i=0, ..., m-1)

• n for the time index (in learning rules)

• {\bf x}_k, and y_k are the samples

• x_{k,i} is component i of sample {\bf x}_k

• x_i would be the ith component of a generic sample

As shown, all indexes should run from 0

[jeroarenas]

List of folders adapted to new notation:

• A1.Optimization

[jeroarenas]

• R1. Regression Intro
• R2. Regression k-NN

[Orieus]

• R4.Regression_ML

[Orieus]

• R5.Bayesian_Regression

[Orieus]

• R3. Least squares

[Orieus]

• R6.Gaussian_Processes
• R7.Regression_Overview
• R_lab1_Bay_GP_Regresion_v_py2
• R_lab1_ML_Bay_Regresion
• R_lab2_GP

[jeroarenas]

• Ejercicios 1 - 16 (optimización y regresión)

[Orieus]

• Exercises 17-end.

[Orieus]

• C1 Intro Classification
• C2 kNN classification
• C3 Classification LogReg

[jeroarenas]

• C4 Classification SVM
• C lab 1 Classification (no tenía notación, únicamente erratas menores)

[Orieus]

• U1 K-means: usual input sample indexing, k=0,..., K-1, is replaced here by $\ell =0, ..., L-1$ to use K for the number of groups and k for the groups.

[Orieus]

• TM1-NLP. No notation changes.

[Orieus]

• U2 Spectral clustering. $K$ is used here for the kernel function, the affinity matrix and the number of K-means groups. To avoid overusing $K$, I have used $N$ for the number of samples. Also, I have not used $k$ for the sample index (so as to avoid K_{kl}). Using i, j and n in this notebook makes a bit more sense, specially because samples i and j are used to compute component Kij of the affinity matrix.

[Orieus]

• U_lab1: a single notation change. U_lab1 contains draft versions only, at this time.

[Orieus]

• TM2, TM3, TM4. No notation changes.

[Orieus]

• E_lab1: old notation unchanged, because it is being used for other courses (TMDE).