Mixed logit model#


Below is documentation for the Matlab code in “full_info_NNE_mixed_logit” folder at this GitHub repository. The code uses full-info NNE to estimate a mixed logit model. Full-info NNE shows no computational or accuracy advantages here, but it serves as a good example to illustrate that the two-part architecture works in practice.

Workflow#

The following shows how to conduct a Monte Carlo experiment that estimates the mixed logit model on a simulated dataset.

>> monte_carlo_data        % simulate a dataset using mixed logit model and save it in data.mat
>> nne_gen                 % generate the training examples
>> nne_train               % train a neural net and apply it to the data

Description of each file#

model_mixed_logit.m#

This function codes the mixed logit model.

Y = mix_logit_model(rs, par, X, consumer_idx)

  • Inputs:

    • rs: a random stream (to control randomness)

    • par: mixed logit model parameter vector

    • X: product attributes

    • consumer_id: indices of consumers (or search sessions)

  • Outputs:

    • Y: dummies indicating if products are bought


nne_gen.m#

This script generates the training, validation, and test examples.

  • It uses model_mixed_logit.m to simulate the data.


nne_train.m#

This script trains a neural net, using the examples from nne_gen.m.

  • Validation loss is reported. You can use it to choose neural net hyperparameters, such as the numbers of hidden nodes.

  • After training, it draws parameter recovery plots using the test examples.

  • After training, it applies the neural net on data.mat.

  • In the end, it saves the trained neural net to trained_nne.mat.


learn.m#

This function codes the training loop, and is used by nne_train.m. This is a custom training loop based on Matlab’s built-in back-propogation and adam algorithms.

[ema_net, train_pred, val_pred, test_pred] = ...
            learn(net, opt, nne, train_dataY, train_label, val_dataY, val_label, test_dataY)

  • Inputs:

    • net: the initial neural net

    • opt: options including batch size, number of iterations, etc.

    • train_dataY, train_label: training examples

    • val_dataY, val_label: validation examples

    • test_dataY: test examples

  • Outputs:

    • ema_net: the final trained neural net

    • train_pred: predictions for training examples

    • val_pred: predictions for validation examples

    • test_pred: predictions for test examples