Discriminative conditional restricted Boltzmann machine for discrete choice and latent variable modelling

Abstract

Conventional methods of estimating latent behaviour generally use attitudinal questions which are subjective and these survey questions may not always be available. We hypothesize that an alternative approach can be used for latent variable estimation through an undirected graphical models. For instance, non-parametric artificial neural networks. In this study, we explore the use of generative non-parametric modelling methods to estimate latent variables from prior choice distribution without the conventional use of measurement indicators. A restricted Boltzmann machine is used to represent latent behaviour factors by analyzing the relationship information between the observed choices and explanatory variables. The algorithm is adapted for latent behaviour analysis in discrete choice scenario and we use a graphical approach to evaluate and understand the semantic meaning from estimated parameter vector values. We illustrate our methodology on a financial instrument choice dataset and perform statistical analysis on parameter sensitivity and stability. Our findings show that through non-parametric statistical tests, we can extract useful latent information on the behaviour of latent constructs through machine learning methods and present strong and significant influence on the choice process. Furthermore, our modelling framework shows robustness in input variability through sampling and validation.

A bi-partite generative model framework for analyzing and simulating large scale multiple discrete-continuous travel behaviour data

“…To the best of our knowledge, the use of generative learning is limited to image and video data to capture motion and dynamics. Here, we extend our previous work on RBM based single discrete choice and latent variable models [48] to incorporate multiple discrete-continuous choices. We also propose a generic algorithm for estimating MDC models using generative machine learning.…”

“…The RBM generative model restricts lateral connections within layers, which provides independent and identically distributed (IID) assumption about the observed and latent variables. For prediction and forecasting, RBMs are typically used for learning latent features followed by either a generative simulation based classifier or directly as a multi-layer neural network classifier [48].…”

On the Expressive Power of Scientific Manuscripts

“…Since HDP is very generative in nature and is capable of producing large volume of topics, care is taken by the auto encoder to feed only the important segments of the research article as input. DSA-H stacked auto encoder utilizes three layers of hidden layer stack before arriving at the output layer (refer figure 5) The auto encoder pre-trained with restricted Boltzmann machine (RBM) [36], learns the inputs and further aims at considerable reduction of input dimension at every hidden layer thereby learning generative models of data. Provision of hidden layers is what contributes towards the sparseness of deep auto encoder.The improvement in average topic coherence of DSA-H for a sample journal full-text articles is given in figure 6.…”

Using Artificial Neural Networks for Recovering the Value-of-Travel-Time Distribution

“…Very recently, Artificial Neural Networks (ANNs) are gaining ground in the travel behaviour research arena [e.g. 11,[12][13][14][15][16][17][18][19][20]. A fundamental difference between discrete choice models and ANNs is the modelling paradigm to which they belong.…”