Why Not to Use Zero Imputation? Correcting Sparsity Bias in Training Neural Networks

Abstract

Handling missing data is one of the most fundamental problems in machine learning. Among many approaches, the simplest and most intuitive way is zero imputation, which treats the value of a missing entry simply as zero. However, many studies have experimentally confirmed that zero imputation results in suboptimal performances in training neural networks. Yet, none of the existing work has explained what brings such performance degradations. In this paper, we introduce the variable sparsity problem (VSP), which describes a phenomenon where the output of a predictive model largely varies with respect to the rate of missingness in the given input, and show that it adversarially affects the model performance. We first theoretically analyze this phenomenon and propose a simple yet effective technique to handle missingness, which we refer to as Sparsity Normalization (SN), that directly targets and resolves the VSP. We further experimentally validate SN on diverse benchmark datasets, to show that debiasing the effect of input-level sparsity improves the performance and stabilizes the training of neural networks.

arXiv comment27 pages

References in corpus (19)

• Adam: A Method for Stochastic Optimization
• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Efficient Estimation of Word Representations in Vector Space
• Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
• Graph Convolutional Matrix Completion
• GAIN: Missing Data Imputation using Generative Adversarial Nets
• DropBlock: A regularization method for convolutional networks
• BRITS: Bidirectional Recurrent Imputation for Time Series
• MADE: Masked Autoencoder for Distribution Estimation
• Modeling Missing Data in Clinical Time Series with RNNs
• A Neural Autoregressive Approach to Collaborative Filtering
• Neural Network Matrix Factorization
• Processing of missing data by neural networks
• Training Deep AutoEncoders for Collaborative Filtering
• Learning Algorithms for Active Learning
• Normalizing the Normalizers: Comparing and Extending Network Normalization Schemes
• Classification with Low Rank and Missing Data
• Collaborative Filtering with User-Item Co-Autoregressive Models
• Regularizing by the Variance of the Activations' Sample-Variances

Cited by in corpus (4)

• MAIN: Multihead-Attention Imputation Networks
• FedNI: Federated Graph Learning with Network Inpainting for Population-Based Disease Prediction
• Graph Convolutional Networks for Graphs Containing Missing Features
• Cloned Identity Detection in Social-Sensor Clouds based on Incomplete Profiles