Deep learning of value at risk through generative neural network models: The case of the Variational auto encoder.

We present in this paper a method to compute, using generative neural networks, an estimator of the "Value at Risk" for a financial asset. The method uses a Variational Auto Encoder with an 'energy' (a.k.a. Radon-Sobolev) kernel. The result behaves according to intuition and is in line with more classical methods.•Estimation of the Value at Risk with generative neural networks•No a priori assumptions on the distribution of the returns•Good practical behavior.

Sickle-cell disease stroke throughout life: a retrospective study in an adult referral center.

Strokes are one of the most severe complications of sickle-cell disease. Most studies have been restricted to children with sickle-cell disease. To better understand the characteristics and follow-up of strokes occurring from childhood to adulthood, we undertook a retrospective cohort study of 69 stroke patients among the 2,875 patients consulting at the French Adult Sickle-Cell Disease Referral Center. Between 1970 and 2008, they had experienced 104 strokes: 80 ischemic, 22 hemorrhagic, and 2 intracranial sinus thromboses. Coma and/or fatal outcomes underscored the severity of strokes in sickle-cell disease patients.Hemorrhagic strokes occurred mostly in adults and carried a higher risk of death than ischemic stroke. The mechanisms underlying sickle-cell disease associated strokes were reevaluated and etiologies were determined for first stroke and recurrences, in childhood and adulthood. Sickle-cell disease vasculopathy concerned only SS patients and remains their most frequent stroke etiology. Cardioembolism, vaso-occlusive crisis and triggering factors were other etiologies identified in adults. Recurrences occurred in 19 SS patients only after a first ischemic stroke. SC patients' strokes occurred in adulthood and were associated with cardiovascular risk factors. Our findings provide novel information about cerebrovascular pathologies throughout the lives of sickle-cell disease patients and suggest the need for different diagnostic and therapeutic management approaches in those different settings.

At-risk phenotype of neurofibromatose-1 patients: a multicentre case-control study.

OBJECTIVES: To assess associations between subcutaneous neurofibromas (SC-NFs) and internal neurofibromas in patients with neurofibromatosis type 1 (NF-1) and to determine whether the association between SC-NFs and peripheral neuropathy was ascribable to internal neurofibromas. PATIENTS AND METHODS: Prospective multicentre case-control study. Between 2005 and 2008, 110 NF-1 adults having two or more SC-NFs were individually matched for age, sex and hospital with 110 controls who had no SC-NF. Patients underwent standardized MRI of the spinal cord, nerve roots and sciatic nerves and an electrophysiological study. Analyses used adjusted multinomial logistic regression (ORa) to estimate the risk of the presence of internal neurofibromas or peripheral neuropathies associated with patients presented 2 to 9 SC-NFs, at least 10 SC-NFs as compared to patients without any (referential category). RESULTS: Cases had a mean age of 41 (± 13) years; 85 (80%) had two to nine SC-NFs and 21 (19%) at least ten SC-NFs. SC-NFs were more strongly associated with internal neurofibromas in patients with ten or more SC-NFs than in patients with fewer NF-SCs (e.g., sciatic nerve, aOR = 29.1 [8.5 to 100] vs. 4.3 [2.1 to 9.0]). The association with SC-NFs was stronger for diffuse, intradural, and > 3 cm internal neurofibromas than with other internal neurofibromas. Axonal neuropathy with slowed conduction velocities (SCV) was more strongly associated with having at least ten SC-NFs (aOR = 29.9, 5.5 to 162.3) than with having fewer SC-NFs (aOR = 4.4, 0.9 to 22.0). Bivariate analyses showed that the association between axonal neuropathy with SCV and sciatic neurofibromas was mediated by the association between SC-NFs and sciatic neurofibromas. CONCLUSION: The at-risk phenotype of NF-1 patients (i.e. NF-1 patients with SC-NFs) is ascribable to associations linking SC-NFs to internal neurofibromas at risk for malignant transformation and to axonal neuropathies with slowed conduction velocities. Axonal neuropathies with SCV are particularly common in patients with at least ten SC-NFs.