# Fitting models to 2d data

I am reading the documentation on tf.lattice (https://www.tensorflow.org/lattice/overview)

I am wondering how the training data was created/trained. Is it represented by (x,y) intpositions and some intensity value output?

" The function can capture non-linear interactions between features. You can think of the lattice parameters as the height of poles set in the ground on a regular grid, and the resulting function is like cloth pulled tight against the four poles.

With features and 2 vertices along each dimension, a regular lattice will have parameters. To fit a more flexible function, you can specify a finer-grained lattice over the feature space with more vertices along each dimension. Lattice regression functions are continuous and piecewise infinitely differentiable.

" EDIT-

Specifically for these pictures, they have given the example of dateset ranking restaurants(links are below). These graphs shows number of reviews(Y-Axis)/Average rating (x-axis).It has source code also. https://www.tensorflow.org/lattice/tutorials/shape_constraints https://www.tensorflow.org/lattice

• No I don't see the answer here, this does not explain how other models are fit to the same data or what the dataset is. Commented May 18, 2020 at 0:04
• umm...they have given this example here,-tensorflow.org/lattice/tutorials/keras_layers. are you looking for dataset and model "specifically" which generated these model in given pictures? or in general how Lattice uses data? may be I should have asked it earlier. Commented May 18, 2020 at 11:20
• I think the latter is answered in the post you answered but the former is not, I am curious how the other models were generated with the same data that the lattice model was generated. Commented May 18, 2020 at 15:56
• Here you go.. its a dataset for ranking restaurants. tensorflow.org/lattice/tutorials/shape_constraints Commented May 18, 2020 at 17:31
• thank you that's what I am looking for. Commented May 18, 2020 at 19:41