Auto IV: Counterfactual Prediction via Automatic Instrumental Variable Decomposition

Instrumental variables (IVs), sources of treatment randomization that are conditionally independent of the outcome, play an important role in causal inference with unobserved confounders. However, the existing IV-based counterfactual prediction methods need well-predefined IVs, while it’s an art rather than science to find valid IVs in many real-world scenes. Moreover, the predefined hand-made IVs could be weak or erroneous by violating the conditions of valid IVs. These thorny facts hinder the application of the IV-based counterfactual prediction methods. In this article, we propose a novel Automatic Instrumental Variable decomposition (AutoIV) algorithm to automatically generate representations serving the role of IVs from observed variables (IV candidates). Specifically, we let the learned IV representations satisfy the relevance condition with the treatment and exclusion condition with the outcome via mutual information maximization and minimization constraints, respectively. We also learn confounder representations by encouraging them to be relevant to both the treatment and the outcome. The IV and confounder representations compete for the information with their constraints in an adversarial game, which allows us to get valid IV representations for IV-based counterfactual prediction. Extensive experiments demonstrate that our method generates valid IV representations for accurate IV-based counterfactual prediction.

Knowledge-Guided Disentangled Representation Learning for Recommender Systems

In recommender systems, it is essential to understand the underlying factors that affect user-item interaction. Recently, several studies have utilized disentangled representation learning to discover such hidden factors from user-item interaction data, which shows promising results. However, without any external guidance signal, the learned disentangled representations lack clear meanings, and are easy to suffer from the data sparsity issue. In light of these challenges, we study how to leverage knowledge graph (KG) to guide the disentangled representation learning in recommender systems. The purpose for incorporating KG is twofold, making the disentangled representations interpretable and resolving data sparsity issue. However, it is not straightforward to incorporate KG for improving disentangled representations, because KG has very different data characteristics compared with user-item interactions. We propose a novel K nowledge-guided D isentangled R epresentations approach ( KDR ) to utilizing KG to guide the disentangled representation learning in recommender systems. The basic idea, is to first learn more interpretable disentangled dimensions (explicit disentangled representations) based on structural KG, and then align implicit disentangled representations learned from user-item interaction with the explicit disentangled representations. We design a novel alignment strategy based on mutual information maximization. It enables the KG information to guide the implicit disentangled representation learning, and such learned disentangled representations will correspond to semantic information derived from KG. Finally, the fused disentangled representations are optimized to improve the recommendation performance. Extensive experiments on three real-world datasets demonstrate the effectiveness of the proposed model in terms of both performance and interpretability.

Deep attributed network embedding by mutual information maximization

There is a growing body of literature that recognizes the importance of network embedding. It intends to encode the graph structure information into a low-dimensional vector for each node in the graph, which benefits the downstream tasks. Most of recent works focus on supervised learning. But they are usually not feasible in real-world datasets owing to the high cost to obtain labels. To address this issue, we design a new unsupervised attributed network embedding method, deep attributed network embedding by mutual information maximization (DMIM). Our method focuses on maximizing mutual information between the hidden representations of the global topological structure and the node attributes, which allows us to obtain the node embedding without manual labeling. To illustrate the effectiveness of our method, we carry out the node classification task using the learned node embeddings. Compared with the state-of-the-art unsupervised methods, our method achieves superior results on various datasets.

Comparison Analysis of Gene Expression Profiles Proximity Metrics

The problems of gene regulatory network (GRN) reconstruction and the creation of disease diagnostic effective systems based on genes expression data are some of the current directions of modern bioinformatics. In this manuscript, we present the results of the research focused on the evaluation of the effectiveness of the most used metrics to estimate the gene expression profiles’ proximity, which can be used to extract the groups of informative gene expression profiles while taking into account the states of the investigated samples. Symmetry is very important in the field of both genes’ and/or proteins’ interaction since it undergirds essentially all interactions between molecular components in the GRN and extraction of gene expression profiles, which allows us to identify how the investigated biological objects (disease, state of patients, etc.) contribute to the further reconstruction of GRN in terms of both the symmetry and understanding the mechanism of molecular element interaction in a biological organism. Within the framework of our research, we have investigated the following metrics: Mutual information maximization (MIM) using various methods of Shannon entropy calculation, Pearson’s χ2 test and correlation distance. The accuracy of the investigated samples classification was used as the main quality criterion to evaluate the appropriate metric effectiveness. The random forest classifier (RF) was used during the simulation process. The research results have shown that results of the use of various methods of Shannon entropy within the framework of the MIM metric disagree with each other. As a result, we have proposed the modified mutual information maximization (MMIM) proximity metric based on the joint use of various methods of Shannon entropy calculation and the Harrington desirability function. The results of the simulation have also shown that the correlation proximity metric is less effective in comparison to both the MMIM metric and Pearson’s χ2 test. Finally, we propose the hybrid proximity metric (HPM) that considers both the MMIM metric and Pearson’s χ2 test. The proposed metric was investigated within the framework of one-cluster structure effectiveness evaluation. To our mind, the main benefit of the proposed HPM is in increasing the objectivity of mutually similar gene expression profiles extraction due to the joint use of the various effective proximity metrics that can contradict with each other when they are used alone.