Accelerated Gradient Approach For Deep Neural Network-Based Adaptive Control of Unknown Nonlinear Systems.

Recent connections in the adaptive control literature to continuous-time analogs of Nesterov's accelerated gradient method have led to the development of new real-time adaptation laws based on accelerated gradient methods. However, previous results assume that the system's uncertainties are linear-in-the-parameters (LIP). To compensate for non-LIP uncertainties, our preliminary results developed a neural network (NN)-based accelerated gradient adaptive controller to achieve trajectory tracking for nonlinear systems; however, the development and analysis only considered single-hidden-layer NNs. In this article, a generalized deep NN (DNN) architecture with an arbitrary number of hidden layers is considered, and a new DNN-based accelerated gradient adaptation scheme is developed to generate estimates of all the DNN weights in real-time. A nonsmooth Lyapunov-based analysis is used to guarantee the developed accelerated gradient-based DNN adaptation design achieves global asymptotic tracking error convergence for general nonlinear control affine systems subject to unknown (non-LIP) drift dynamics and exogenous disturbances. A comprehensive set of simulation studies are conducted on a two-state nonlinear system, a robotic manipulator, and a complex 20-D nonlinear system to demonstrate the improved performance of the developed method. Our simulation studies demonstrate enhanced tracking and function approximation performance from both DNN architectures and accelerated gradient adaptation.

Recovery of upland acid grasslands after successful Pteridium aquilinum control: Long-term effectiveness of cutting, repeated herbicide treatment and bruising.

There is a clear need for the development of management strategies to control dominant, perennial weeds and restore semi-natural communities and an important part of this is to know how long control treatments take to be effective and how long they last after treatments stop. Here, we report the results from a 17-year long experiment where we compared the effects of five control treatments on dense Pteridium aquilinum (L. Kuhn) relative to an untreated experimental-control in Derbyshire, UK. The experiment was run in two phases. In Phase 1 (2005-2012) we controlled the P. aquilinum by cutting and bruising, both twice and thrice annually, and a herbicide treatment (asulam in year 1, followed by annual spot-re-treatment of all emergent fronds). In Phase 2 (2012-2021) all treatments were stopped, and the vegetation was allowed to develop naturally. Between 2005 and 2021 we monitored P. aquilinum performance annually and full plant species composition at intervals. Here, we concentrate on analysing the Phase 2 data where we used regression approaches to model individual species responses through time and unconstrained ordination to compare treatment effects on the entire species composition over both Phases. Remote sensing was also used to assess edge invasion in 2018. At the end of Phase 1, a good reduction of P. aquilinum and restoration of acid-grassland was achieved for the asulam and cutting treatments, but not for bruising. In Phase 2, P. aquilinum increased through time in all treated plots but the asulam and cutting ones maintained a much lower P. aquilinum performance for nine years on all measures assessed. There was a reduction in species richness and richness fluctuations, especially in graminoid species. However, multivariate analysis showed that the asulam and cutting treatments were stationed some distance from the untreated and bruising treatments with no apparent sign of reversions suggesting an Alternative Stable State had been created, at least over this nine-year period. P. aquilinum reinvasion was mainly from plot edges. The use of repeated P. aquilinum control treatments, either through an initial asulam spray with annual follow-up spot-spraying or cutting twice or thrice annually for eight years gave good P. aquilinum control and helped restore an acid-grassland community. Edge reinvasion was detected, and it is recommended that either whole-patch control be implemented or treatments should be continued around patch edges.

Change to ecosystem properties through changing the dominant species: Impact of Pteridium aquilinum-control and heathland restoration treatments on selected soil properties.

It is well known that soils are influenced by the plant species that grow in them. Here we consider the effects of management-induced changes to plant communities and their soils during restoration within a 20-year manipulative experiment where the aim was to change a late-successional community dominated by the weed, Pteridium aquilinum, to an earlier-successional grass-heath one. The ecological restoration treatments altered the above- and below-ground components of the community substantially. Untreated plots maintained a dense Pteridium cover with little understory vegetation, cutting treatments produce significant reductions of Pteridium, whereas herbicide (asulam) produced significant immediate reductions in Pteridium but regressed towards the untreated plots within 10 years. Thereafter, all asulam-treated plots were re-treated in year 11, and then were spot-sprayed annually. Both cutting and asulam treatments reduced frond density to almost zero and resulted in a grass-heath vegetation. There was also a massive change in biomass distribution, untreated plots had a large above-ground biomass/necromass that was much reduced where Pteridium was controlled. Below-ground in treated plots, there was a replacement of the substantive Pteridium rhizome mass with a much greater root mass of other species. The combined effects of Pteridium-control and restoration treatment, reduced soil total C and N as and available P concentrations, but increased soil pH and available N. Soil biological activity was also affected with a reduction in soil N mineralization rate, but an increased soil-root respiration. Multivariate analysis showed a clear trend along a pH/organic matter gradient, with movement along it correlated to management intensity from the untreated plots with low pH/high organic matter and treated plots with to a higher pH/lower organic matter in the sequence asulam treatment, cut once per year to cut twice per year. The role that these changed soil conditions might have in restricting Pteridium recovery are discussed.

Competing conservation goals, biodiversity or ecosystem services: element losses and species recruitment in a managed moorland-bracken model system.

Conservation management in Europe is often geared towards restoring semi-natural ecosystems, where the objective is to reverse succession and re-establish early-successional communities, to comply with national and international conservation targets. At the same time, it is increasingly recognised that ecosystems provide services that contribute to other, possibly conflicting policy requirements. Few attempts have been made to define these conflicts. Here, we assess some potential conflicts using a Calluna vulgaris-dominated moorland invaded by bracken (Pteridium aquilinum) as a model system, where the current policy is to reverse this process and restore moorland. We examined impacts of bracken control treatments on services (stocks and losses of C and mineral nutrients), litter turnover and biodiversity within a designed experiment over 7 years. Bracken litter was >2000 g m(-2) in untreated plots, and treatments reduced this quantity, and its element content, to varying degrees. Cutting twice per year was the most successful treatment in reducing bracken litter and its element content, increasing litter turnover, and increasing both mass and diversity of non-bracken vegetation. Diversity was greatest where bracken litter had been reduced, but species composition was also influenced by light sheep grazing. There was a significant loss of some chemical elements from bracken that could not be accounted for in other pools, and hence potentially lost from the system. In absolute terms large amounts of C and N were lost, but when expressed as a percentage of the total amount in the system, Mg was potentially more important with losses of almost a third of the Mg in the surface soil-vegetation system. There is, therefore, a potential dilemma between controlling a mid-successional invasive species for conservation policy objectives, especially when that species has evolved to sequester nutrients, and the negative effect of increasing environmental costs in terms of carbon accounting required, the potential input of nutrients to aquatic systems, and long-term nutrient loss. There is, therefore, a need to balance conservation goals against potential damage to biogeochemical structure and function.

Exploitation of astrocytes by glioma cells to facilitate invasiveness: a mechanism involving matrix metalloproteinase-2 and the urokinase-type plasminogen activator-plasmin cascade.

The presence of reactive astrocytes around glioma cells in the CNS suggests the possibility that these two cell types could be interacting. We addressed whether glioma cells use the astrocyte environment to modulate matrix metalloproteinase-2 (MMP-2), a proteolytic enzyme implicated in the invasiveness of glioma cells. We found that astrocytes in culture produce significant amounts of the pro-form of MMP-2 but undetectable levels of active MMP-2. However, after coculture with the U251N glioma line, astrocyte pro-MMP-2 was converted to the active form. The mechanism of pro-MMP-2 activation in glioma-astrocyte coculture was investigated and was found to involve the urokinase-type plasminogen activator (uPA)-plasmin cascade whereby uPA bound to uPA receptor (uPAR), leading to the conversion of plasminogen to plasmin. The latter cleaved pro-MMP-2 to generate its active form. Furthermore, key components (i.e., uPAR, uPA, and pro-MMP-2) were contributed principally by astrocytes, whereas the U251N glioma cells provided plasminogen. In correspondence with this biochemical cascade, the transmigration of U251N cells through Boyden invasion chambers coated with an extracellular matrix barrier was increased significantly in the presence of astrocytes, and this was inhibited by agents that disrupted the uPA-plasmin cascade. Finally, using resected human glioblastoma specimens, we found that tumor cells, but not astrocytes, expressed plasminogen in situ. We conclude that glioma cells exploit their astrocyte environment to activate MMP-2 and that this leads to the increased invasiveness of glioma cells.

Verapamil: a selective antagonist of constrictor substances in dog coronary artery: implications for variant angina.

1. Canine circumflex coronary artery ring segments were contracted in vitro by ergometrine, serotonin, phenylephrine, noradrenaline (with propranolol) and a thromboxane A2 analogue, U46619. 2. Ergometrine was classified as a serotonin agonist since concentration-response curves were competitively inhibited by methysergide but not by alpha-adrenoceptor antagonists. 3. Glyceryl trinitrate (IC50 18.6 nmol/l) relaxed the coronary rings precontracted with serotonin, phenylephrine or U46619. In contrast (+/-)-verapamil (0.1-10 mumol/l) was more effective against serotonin than phenylephrine or noradrenaline and was almost inactive against U46619. 4. In a blood perfused left anterior descending coronary artery preparation external diameter was measured by sonomicrometry. Serotonin, U46619 and ergometrine infusions (i.a.) decreased diameter by up to 18% without causing spasm (zero lumen diameter). Lowering the perfusion pressure from 90 to 60 mmHg increased the fall in diameter during serotonin infusions. 5. The negative inotropic potency of verapamil against noradrenaline induced beta-adrenergic stimulation in guinea-pig left atria was compared with the vasodilator potency of verapamil in noradrenaline constricted dog coronary artery rings. Verapamil was eighteen times more potent in cardiac muscle than in coronary smooth muscle. 6. This apparent tissue selectivity of verapamil was confirmed in anaesthetized dogs where plasma concentrations of verapamil 50-150 ng/ml (in the therapeutic range) lowered blood pressure and heart rate and increased P-R interval without greatly reducing the constrictor response to serotonin in the coronary artery. 7. These studies suggest that inhibition of constrictor responses in large coronary vessels may not be an important site of action of verapamil in patients with variant angina.