machine learning

Computer vision and image analysis are major application examples of deep learning. While computer vision and image analysis deal with existing images and produce features of these images (images to features), tomographic imaging produces images of multi-dimensional structures from experimentally measured “encoded” data as various tomographic features (integrals, harmonics, and so on, of underlying images) (features to images). Recently, deep learning is being actively developed worldwide for tomographic imaging, forming a new area of imaging research.

In this talk, I will provide an introduction to the use of machine learning and convolutional neural networks (CNNs) in the area of MR image reconstruction. Building on a general framework of inverse problems and variational optimization, I will focus on application examples from image reconstruction for accelerated Magnetic Resonance (MR) imaging. I will cover both methodological developments as well as clinical translation and validation.

We present a general method for privacy-preserving Bayesian inference in Poisson factorization, a broad class of models that includes some of the most widely used models in the social sciences. Our method satisfies limited precision local privacy, a generalization of local differential privacy, which we introduce to formulate privacy guarantees appropriate for sparse count data. We develop an MCMC algorithm that approximates the locally private posterior over model parameters given data that has been locally privatized by the geometric mechanism (Ghosh et al., 2012).

Neural networks solve complex computer vision problems with human-like accuracy. However, it has recently been observed that neural nets are easily fooled and manipulated by adversarial examples, in which an attacker manipulates the network by making tiny changes to its inputs. In this talk, I give a high-level overview of adversarial examples, and then discuss a newer type of attack called data poisoning, in which a network is manipulated at train time rather than test time.

Current guidelines for treatment decision making largely rely on data from ran- domized controlled trials (RCTs) studying average treatment effects. They may be inadequate to make individualized treatment decisions in real-world settings. Large- scale electronic health records (EHR) provide opportunities to fulfill the goals of personalized medicine and learn individualized treatment rules (ITRs) depending on patient-specific characteristics from real-world patient data.

Many real-world analytics problems involve two significant challenges: prediction and optimization. Due to the typically complex nature of each challenge, the standard paradigm is to predict, then optimize. By and large, machine learning tools are intended to minimize prediction error and do not account for how the predictions will be used in a downstream optimization problem.

Predictive analytics is increasingly used to guide decision-making in many applications. However, in practice, we often have limited data on the true outcome that we wish to predict, but copious data on an intermediate or proxy outcome. Practitioners often train predictive models on proxies since it achieves more accurate predictions.

Forensic analyses are often concerned with identifying the spatial source of biological residue. Using recent advances in high-throughput sequencing technologies, dust collected from nearly any object can be shown to harbor DNA fragments from thousands of bacteria and fungi species which may be informative of the source of the dust. We show that training collections of deep neural network classifiers on random Voronoi partitions of a spatial domain yields remarkably accurate geolocation predictions.

The fast moving of the cancer immunotherapy field has generated tremendous excitement regarding new therapeutic strategies and will likely change the paradigm of therapeutic interventions for cancer. Neoantigens, generated by tumor-specific DNA alterations that result in the formation of novel protein sequences and only in cancer cells, represent an optimal target for the immune system and make possible a new class of highly personalized vaccines with the potential for significant efficacy with reduced side effects.

Estimating individualized treatment rules is a central task of personalized or precision medicine. In this presentation, we review several new developments in outcome weighted learning for identifying individualized treatment rules for two challenging settings: dose finding and treatment selection under right censoring. In the former, we develop an approach which involves the use of two different kernels; and in the later, we use random forests to address censoring before applying outcome weighted learning.