Search

Understanding the large-scale causal relationship among brain regions is crucial for elucidating the information flow that the brain integrates external stimuli and generates behaviors. Despite the availability of neurostimulation and computational methods to infer causal relationships among a limited number of regions, these approaches are not capable of mapping the causal network of the entire brain, also known as the effective brain connectome (EBC). To address this gap, we propose a data-driven framework called Neural Perturbational Inference (NPI) and map the human EBC for the first time. NPI uses an artificial neural network trained to learn large-scale neural dynamics as a surrogate brain. By perturbing each region of the surrogate brain and observing the resulting responses in all other regions, the human EBC is obtained. This connectome captures the directionality, strength, and excitatory-inhibitory distinction of brain-wide causal relationships, offering mechanistic insights into cognitive processes. EBC provides a complete picture of information flow both within and across brain functional networks as well as reveals the large-scale hierarchy of the organization of excitatory and inhibitory ECs. As EBC captures the neurostimulation transmission pathways in the brain, it has great potential to guide the target selection in personalized neurostimulation of neurological disorders.

Recent studies show that deep learning models achieve good performance on medical imaging tasks such as diagnosis prediction. Among the models, multimodality has been an emerging trend, integrating different forms of data such as chest X-ray (CXR) images and electronic medical records (EMRs). However, most existing methods incorporate them in a model-free manner, which lacks theoretical support and ignores the intrinsic relations between different data sources. To address this problem, we propose a knowledge-driven and data-driven framework for lung disease diagnosis. By incorporating domain knowledge, machine learning models can reduce the dependence on labeled data and improve interpretability. We formulate diagnosis rules according to authoritative clinical medicine guidelines and learn the weights of rules from text data. Finally, a multimodal fusion consisting of text and image data is designed to infer the marginal probability of lung disease. We conduct experiments on a real-world dataset collected from a hospital. The results show that the proposed method outperforms the state-of-the-art multimodal baselines in terms of accuracy and interpretability.

Background Seroconversion of hepatitis B e antigen (HBeAg) is a critical event in the natural course of hepatitis B virus (HBV) infection. Aim We herein characterize the virological factors associated with postpartum spontaneous HBeAg seroconversion. Methods A total of 214 pregnant women positive for both hepatitis B surface antigen (HBsAg) and HBeAg were followed up at 7–12 months postpartum. Results Of the subjects, 26 (12.1%) achieved spontaneous HBeAg seroconversion. Receiver operating curve analysis indicated that HBV DNA level Conclusion These results suggest that quantitative determination of virological markers and sequencing PC and BCP can predict spontaneous HBeAg seroconversion, which could be valuable in deciding antiviral therapy against HBV.

In this paper, we provide a deep dive into the deployment of inference accelerators at Facebook. Many of our ML workloads have unique characteristics, such as sparse memory accesses, large model sizes, as well as high compute, memory and network bandwidth requirements. We co-designed a high-performance, energy-efficient inference accelerator platform based on these requirements. We describe the inference accelerator platform ecosystem we developed and deployed at Facebook: both hardware, through Open Compute Platform (OCP), and software framework and tooling, through Pytorch/Caffe2/Glow. A characteristic of this ecosystem from the start is its openness to enable a variety of AI accelerators from different vendors. This platform, with six low-power accelerator cards alongside a single-socket host CPU, allows us to serve models of high complexity that cannot be easily or efficiently run on CPUs. We describe various performance optimizations, at both platform and accelerator level, which enables this platform to serve production traffic at Facebook. We also share deployment challenges, lessons learned during performance optimization, as well as provide guidance for future inference hardware co-design.