Epidural electrical stimulation (EES) targeting the dorsal roots of lumbosacral segments restored walking in people with spinal cord injury (SCI). However, EES was delivered with multielectrode paddle leads that were originally designed to target the dorsal column of the spinal cord. Here, we hypothesized that an arrangement of electrodes targeting the ensemble of dorsal roots involved in leg and trunk movements would result in superior efficacy, restoring more diverse motor activities after the most severe SCI. To test this hypothesis, we established a computational framework that informed the optimal arrangement of electrodes on a new paddle lead and guided its neurosurgical positioning. We also developed a software supporting the rapid configuration of activity-specific stimulation programs that reproduced the natural activation of motor neurons underlying each activity. We tested these neurotechnologies in three individuals with complete sensorimotor paralysis, as part of an ongoing clinical trial (clinicaltrials.gov, NCT02936453). Within a single day, activity-specific stimulation programs enabled the three individuals to stand, walk, cycle, swim, and control trunk movements. Neurorehabilitation mediated sufficient improvement to restore these activities in community settings, opening a realistic path to support everyday mobility with EES in people with SCI.

We approach the task of network congestion control in datacenters using Reinforcement Learning (RL). Successful congestion control algorithms can dramatically improve latency and overall network throughput. Until today, no such learning-based algorithms have shown practical potential in this domain. Evidently, the most popular recent deployments rely on rule-based heuristics that are tested on a predetermined set of benchmarks. Consequently, these heuristics do not generalize well to newly-seen scenarios. Contrarily, we devise an RL-based algorithm with the aim of generalizing to different configurations of real-world datacenter networks. We overcome challenges such as partial-observability, nonstationarity, and multi-objectiveness. We further propose a policy gradient algorithm that leverages the analytical structure of the reward function to approximate its derivative and improve stability. We show that this scheme outperforms alternative popular RL approaches, and generalizes to scenarios that were not seen during training. Our experiments, conducted on a realistic simulator that emulates communication networks' behavior, exhibit improved performance concurrently on the multiple considered metrics compared to the popular algorithms deployed today in real datacenters. Our algorithm is being productized to replace heuristics in some of the largest datacenters in the world.

THE most common manifestation of Listeria monocytogenes infection (listeriosis) in cattle is meningoencephalitis. The clinical signs are characteristic and result from lesions in the brainstem, pons or cerebellum (Cooper and Walker 1998). In sheep, a central nervous system form ofthe disease has been described in which lesions are restricted to the spinal cord. Affected animals have paralysis of one or more limbs, a normal mental status and no loss of cranial nerve function (Gates and others 1967, Seaman and others 1990). This short communication describes two Swiss Braunvieh heifers from different farms with myelitis caused by L monocytogenes infection.Heifer 1 was 18 months old and was presented with a three-day history of fever and ataxia. The heifer was in lateral recumbency and had been unable to rise for two days. The animal had an obvious opisthotonos, but was still chewing the cud. After being propped up, the heifer could hold itself in sternal position with a normal posture of the head and neck. The rectal temperature was 37-2°C and the heart rate was 48 bpm. The heifer was bright and alert; its feed and water intake were normal. There was no tail tone, and the anal and panniculus reflexes were severely reduced. Tetraplegia was noted when the heifer was lifted using a cow lift. The heifer had been treated with antibiotics and cortisone for two days before referral to the clinic.Heifer 2 was 12 months old and had a four-day history of hindlimb ataxia and fever. However, the animal was bright, alert and standing. The rectal temperature was 38-7°C and the heart rate was 84 bpm. Mental status and cranial nerve function were normal, but the heifer had mild, intermittent hindlimb ataxia and hypermetria of the hindlimbs when walking over a small obstacle. The heifer had an increased total leucocyte count. Before referral to the clinic it had received two days of antibiotic treatment. Both heifers had a decreased clotting time in the glutaraldehyde test, increased activity of aspartate aminotransferase, and increased concentration of creatinine kinase (Table 1).On the basis of the clinical findings, a diagnosis of spinal cord disease was made in both animals. In heifer 1, multiple lesions were assumed to be located between Cl and C5 and S1 and S3, and a diffuse inflammatory process of unclear aetiology was suspected. In heifer 2, the lesions were assumed to be located between T3 and L3. Possible causes included a vertebral abscess, parasitism or trauma. Weaver syndrome was also considered as a differential diagnosis in heifer 2, but was ruled out on the basis of the heifer's lineage (Braun and others 2003a, b, c).At the clinic, heifer 1 was treated with intravenous glucose and saline solution and 1 1 mg/kg flunixin (Fluniximin; Berna) intravenously once daily. However, the heifer's condition rapidly deteriorated and it was euthanased one day after admission. Postmortem examination revealed a severe, suppurative, necrotic myelitis with microgranulomas and microabscesses in the cervical spinal cord (Fig 1) and t...

We approach the task of network congestion control in datacenters using Reinforcement Learning (RL). Successful congestion control algorithms can dramatically improve latency and overall network throughput. Until today, no such learning-based algorithms have shown practical potential in this domain. Evidently, the most popular recent deployments rely on rule-based heuristics that are tested on a predetermined set of benchmarks. Consequently, these heuristics do not generalize well to newly-seen scenarios. Contrarily, we devise an RL-based algorithm with the aim of generalizing to different configurations of real-world datacenter networks. We overcome challenges such as partial-observability, non-stationarity, and multi-objectiveness. We further propose a policy gradient algorithm that leverages the analytical structure of the reward function to approximate its derivative and improve stability. We show that these challenges prevent standard RL algorithms from operating within this domain. Our experiments, conducted on a realistic simulator that emulates communication networks' behavior, show that our method exhibits improved performance concurrently on the multiple considered metrics compared to the popular algorithms deployed today in real datacenters. Our algorithm is being productized to replace heuristics in some of the largest datacenters in the world.