Deep learning has given rise to the field of representation learning, which aims to automatically extract rich semantics from data. However, there have been several challenges in the generalization capabilities of deep learning models. Recent works have highlighted beneficial properties of causal models that are desirable for learning robust models under distribution shifts. Thus, there has been a growing interest in causal representation learning for achieving generalizability in tasks involving reasoning and planning. The goal of my dissertation is to develop theoretical intuitions and practical algorithms that uncover the nature of causal representations and their applications. In my work, I focus on causal generative modeling with an emphasis on either representation or generation. For representation learning, I investigate the disentanglement of causal representations through the lens of independent causal mechanisms. For generation tasks, I develop algorithms for counterfactual generation under weak supervision settings by leveraging recent advances in generative modeling. The proposed approaches have been empirically shown to be effective in achieving disentanglement and generating counterfactuals.

Large language models (LLMs) have shown remarkable ability in various language tasks, especially with their emergent in-context learning capability.Extending LLMs to incorporate visual inputs, large vision-language models (LVLMs) have shown impressive performance in tasks such as recognition and visual question answering (VQA).Despite increasing interest in the utility of LLMs in causal reasoning tasks such as causal discovery and counterfactual reasoning, there has been relatively little work showcasing the abilities of LVLMs on visual causal reasoning tasks.We take this opportunity to formally introduce a comprehensive causal reasoning benchmark for multi-modal in-context learning from LVLMs.Our CausalVLBench encompasses three representative tasks: causal structure inference, intervention target prediction, and counterfactual prediction.We evaluate the ability of state-ofthe-art open-source LVLMs on our causal reasoning tasks across three causal representation learning datasets and demonstrate their fundamental strengths and weaknesses.We hope that our benchmark elucidates the drawbacks of existing vision-language models and motivates new directions and paradigms in improving the visual causal reasoning abilities of LVLMs.We make our code and data available at https: //github.com/Akomand/CausalVLBench.

Diffusion probabilistic models (DPMs) have become the state-of-the-art in high-quality image generation. However, DPMs have an arbitrary noisy latent space with no interpretable or controllable semantics. Although there has been significant research effort to improve image sample quality, there is little work on representation-controlled generation using diffusion models. Specifically, causal modeling and controllable counterfactual generation using DPMs is an underexplored area. In this work, we propose CausalDiffAE, a diffusion-based causal representation learning framework to enable counterfactual generation according to a specified causal model. Our key idea is to use an encoder to extract high-level semantically meaningful causal variables from high-dimensional data and model stochastic variation using reverse diffusion. We propose a causal encoding mechanism that maps high-dimensional data to causally related latent factors and parameterize the causal mechanisms among latent factors using neural networks. To enforce the disentanglement of causal variables, we formulate a variational objective and leverage auxiliary label information in a prior to regularize the latent space. We propose a DDIM-based counterfactual generation procedure subject to do-interventions. Finally, to address the limited label supervision scenario, we also study the application of CausalDiffAE when a part of the training data is unlabeled, which also enables granular control over the strength of interventions in generating counterfactuals during inference. We empirically show that CausalDiffAE learns a disentangled latent space and is capable of generating high-quality counterfactual images.

The goal of causal representation learning is to map low-level observations to high-level causal concepts to learn interpretable and robust representations for various downstream tasks. Latent variable models such as the variational autoencoder (VAE) are frequently leveraged to learn disentangled representations. However, there are often complex non-linear causal relationships underlying the observed data that cannot be captured through disentangled representations or linear dependence assumptions. Further, an independent conditional prior assumption can make learning causal dependencies in the latent space more challenging. We propose a framework, coined SCM-VAE, which uses apriori causal knowledge, a structural causal prior, and a non-linear additive noise structural causal model (SCM) to learn independent causal mechanisms and identifiable causal representations. We conduct theoretical analysis and perform experiments on synthetic and real-world datasets to show the improved quality of learned causal representations and robustness under interventions.

In the modern age of social media and networks, graph representations of real-world phenomena have become an incredibly useful source to mine insights. Often, we are interested in understanding how entities in a graph are interconnected. The Graph Neural Network (GNN) has proven to be a very useful tool in a variety of graph learning tasks including node classification, link prediction, and edge classification. However, in most of these tasks, the graph data we are working with may be noisy and may contain spurious edges. That is, there is a lot of uncertainty associated with the underlying graph structure. Recent approaches to modeling uncertainty have been to use a Bayesian framework and view the graph as a random variable with probabilities associated with model parameters. Introducing the Bayesian paradigm to graph-based models, specifically for semi-supervised node classification, has been shown to yield higher classification accuracies. However, the method of graph inference proposed in recent work does not take into account the structure of the graph. In this paper, we propose a novel algorithm called Bayesian Graph Convolutional Network using Neighborhood Random Walk Sampling (BGCN-NRWS), which uses a Markov Chain Monte Carlo (MCMC) based graph sampling algorithm utilizing graph structure, reduces overfitting by using a variational inference layer, and yields consistently competitive classification results compared to the state-of-the-art in semi-supervised node classification.