generate_data

generate_data

Generates data via a sparse latent variable model based on Witten et al. 2009. The generated data has two modalities with nexamples samples and nfeatx and nfeaty variables, respectively. The activex and activey inputs define the number of variables in the two data modalities that are associated with a Gaussian latent variable.

Syntax

  [X, Y, wX, wY] = generate_data(nexamples, nfeatx, nfeaty, activex, activey, noise)

Inputs

• nexamples [int]

  number of examples in generated data

• nfeatx [int]

  number of features in generated data \(\mathbf{X}\)

• nfeaty [int]

  number of features in generated data \(\mathbf{Y}\)

• activex [int]

  number of active features in generated data \(\mathbf{X}\) associated with the latent variable

• activey [int]

  number of active features in generated data \(\mathbf{Y}\) associated with the latent variable

• noise [float]

  noise level in the generative model

Outputs

• X [2D numeric array]

  generated data \(\mathbf{X}\) with nexamples rows and nfeatx columns

• Y [2D numeric array]

  generated data \(\mathbf{Y}\) with nexamples rows and nfeaty columns

• wX [numeric array]

  true weights used to generate data \(\mathbf{X}\) from the latent variable, which has activex non-zero values

• wY [numeric array]

  true weights used to generate data \(\mathbf{Y}\) from the latent variable, which has activey non-zero values

Examples

  % Example 1
  [X, Y, wX, wY] = generate_data(1000, 100, 100, 10, 10, 1);

---