Encoder Analysis: Simple NN vs. scVI NN for BGPLVM

Table of Links

Abstract and 1. Introduction

2. Background

2.1 Amortized Stochastic Variational Bayesian GPLVM

2.2 Encoding Domain Knowledge through Kernels

3. Our Model and Pre-Processing and Likelihood

3.2 Encoder

4. Results and Discussion and 4.1 Each Component is Crucial to Modifies Model Performance

4.2 Modified Model achieves Significant Improvements over Standard Bayesian GPLVM and is Comparable to SCVI

4.3 Consistency of Latent Space with Biological Factors

4. Conclusion, Acknowledgement, and References

A. Baseline Models

B. Experiment Details

C. Latent Space Metrics

D. Detailed Metrics

3.2 ENCODER

In the encoder analysis, we compare a simple encoder comprised of linear layers followed by SoftPlus activations (Simple NN) with the scVI’s more complex encoder (scVI NN). scVI NN incorporates batch information as input to the nonlinear mapping, so incorporating this encoder into the BGPLVM may help address batch effects observed in the raw count data. Additionally, the scVI encoder architecture includes batch normalizations, contributing to a more stable optimization process, which we leverage for our GPLVM implementation.