No abstract

Learning individual-level causal effects from observational data, such as inferring the most effective medication for a specific patient, is a problem of growing importance for policy makers. The most important aspect of inferring causal effects from observational data is the handling of confounders, factors that affect both an intervention and its outcome. A carefully designed observational study attempts to measure all important confounders. However, even if one does not have direct access to all confounders, there may exist noisy and uncertain measurement of proxies for confounders. We build on recent advances in latent variable modeling to simultaneously estimate the unknown latent space summarizing the confounders and the causal effect. Our method is based on Variational Autoencoders (VAE) which follow the causal structure of inference with proxies. We show our method is significantly more robust than existing methods, and matches the state-of-the-art on previous benchmarks focused on individual treatment effects.

Argument extraction is the task of identifying arguments, along with their components in text. Arguments can be usually decomposed into a claim and one or more premises justifying it. Among the novel aspects of this work is the thematic domain itself which relates to Social Media, in contrast to traditional research in the area, which concentrates mainly on law documents and scientific publications. The huge increase of social media communities, along with their user tendency to debate, makes the identification of arguments in these texts a necessity. Argument extraction from Social Media is more challenging because texts may not always contain arguments, as is the case of legal documents or scientific publications usually studied. In addition, being less formal in nature, texts in Social Media may not even have proper syntax or spelling. This paper presents a two-step approach for argument extraction from social media texts. During the first step, the proposed approach tries to classify the sentences into “sentences that contain arguments” and “sentences that don’t contain arguments”. In the second step, it tries to identify the exact fragments that contain the premises from the sentences that contain arguments, by utilizing conditional random fields. The results exceed significantly the base line approach, and according to literature, are quite promising.

Neural network quantization has become an important research area due to its great impact on deployment of large models on resource constrained devices. In order to train networks that can be effectively discretized without loss of performance, we introduce a differentiable quantization procedure. Differentiability can be achieved by transforming continuous distributions over the weights and activations of the network to categorical distributions over the quantization grid. These are subsequently relaxed to continuous surrogates that can allow for efficient gradient-based optimization. We further show that stochastic rounding can be seen as a special case of the proposed approach and that under this formulation the quantization grid itself can also be optimized with gradient descent. We experimentally validate the performance of our method on MNIST, CIFAR 10 and Imagenet classification.

No abstract