Inverting recurrent neural networks for internal model control of nonlinear systems

In this paper, we show how a set of recently derived theoretical results for recurrent neural networks can be applied to the production of an internal model control system for a nonlinear plant. The results include determination of the relative order of a recurrent neural network and invertibility of such a network. A closed loop controller is produced without the need to retrain the neural network plant model. Stability of the closed-loop controller is also demonstrated.

A practical approach for training dynamic recurrent neural networks: use of Apriori information

Presents a technique for incorporating a priori knowledge from a state space system into a neural network training algorithm. The training algorithm considered is that of chemotaxis and the networks being trained are recurrent neural networks. Incorporation of the a priori knowledge ensures that the resultant network has behaviour similar to the system which it is modelling.

Internal model control of nonlinear systems through the inversion of recurrent neural networks

Recurrent neural networks can be used for both the identification and control of nonlinear systems. This paper takes a previously derived set of theoretical results about recurrent neural networks and applies them to the task of providing internal model control for a nonlinear plant. Using the theoretical results, we show how an inverse controller can be produced from a neural network model of the plant, without the need to train an additional network to perform the inverse control.

Dynamic input/output linearization using recurrent neural networks

A dynamic recurrent neural network (DRNN) is used to input/output linearize a control affine system in the globally linearizing control (GLC) structure. The network is trained as a part of a closed loop that involves a PI controller, the goal is to use the network, as a dynamic feedback, to cancel the nonlinear terms of the plant. The stability of the configuration is guarantee if the network and the plant are asymptotically stable and the linearizing input is bounded.

Identification of nonlinear systems with a dynamic recurrent neural network

Two approaches are presented to calculate the weights for a Dynamic Recurrent Neural Network (DRNN) in order to identify the input-output dynamics of a class of nonlinear systems. The number of states of the identified network is constrained to be the same as the number of states of the plant.

Relative order defines a topology for recurrent networks

This paper uses techniques from control theory in the analysis of trained recurrent neural networks. Differential geometry is used as a framework, which allows the concept of relative order to be applied to neural networks. Any system possessing finite relative order has a left-inverse. Any recurrent network with finite relative order also has an inverse, which is shown to be a recurrent network.

Phylogenetic systematics of Sternbergia (Amaryllidaceae) based on plastid and ITS sequence data

The phylogenetics of Sternbergia (Amaryllidaceae) were studied using DNA sequences of the plastid ndhF and matK genes and nuclear internal transcribed spacer (ITS) ribosomal region for 38, 37 and 32 ingroup and outgroup accessions, respectively. All members of Sternbergia were represented by at least one accession, except S. minoica and S. schubertii, with additional taxa from Narcissus and Pancratium serving as principal outgroups. Sternbergia was resolved and supported as sister to Narcissus and composed of two primary subclades: S. colchiciflora sister to S. vernalis, S. candida and S. clusiana, with this clade in turn sister to S. lutea and its allies in both Bayesian and bootstrap analyses. A clear relationship between the two vernal flowering members of the genus was recovered, supporting the hypothesis of a single origin of vernal flowering in Sternbergia. However, in the S. lutea complex, the DNA markers examined did not offer sufficient resolving power to separate taxa, providing some support for the idea that S. sicula and S. greuteriana are conspecific with S. lutea

Structural characterization of a new crystal form of the four-way Holliday junction formed by the DNA sequence d(CCGGTACCGG)2: sequence versus lattice?

DNA-strand exchange is a vital step in the recombination process, of which a key intermediate is the four-way DNA Holliday junction formed transiently in most living organisms. Here, the single-crystal structure at a resolution of 2.35 Å of such a DNA junction formed by d(CCGGTACCGG)2, which has crystallized in a more highly symmetrical packing mode to that previously observed for the same sequence, is presented. In this case, the structure is isomorphous to the mismatch sequence d(CCGGGACCGG)2, which reveals the roles of both lattice and DNA sequence in determining the junction geometry. The helices cross at the larger angle of 43.0° (the previously observed angle for this sequence was 41.4°) as a right-handed X. No metal cations were observed; the crystals were grown in the presence of only group I counter-cations.

Amyloid peptides incorporating a core sequence from the amyloid beta peptide and gamma amino acids: relating bioactivity to self-assembly

A series of heptapeptides comprising the core sequence Ab(16–20), KLVFF, of the amyloid b peptide coupled with paired N-terminal c-amino acids are investigated in terms of cytotoxicity reduction and binding to the full Ab peptide, both pointing to inhibition of fibrillisation for selected compounds. This is related to the self-assembly capacity of the heptapeptides.

Conformational modulation of sequence recognition in synthetic macromolecules

The different triplet sequences in high molecular weight aromatic copolyimides comprising pyromellitimide units ("I") flanked by either ether-ketone ("K") or ether-sulfone residues ("S") show different binding strengths for pyrene-based tweezer-molecules. Such molecules bind primarily to the diimide unit through complementary π-π-stacking and hydrogen bonding. However, as shown by the magnitudes of 1H NMR complexation shifts and tweezer-polymer binding constants, the triplet "SIS" binds tweezer-molecules more strongly than "KIS" which in turn bind such molecules more strongly than "KIK". Computational models for tweezer-polymer binding, together with single-crystal X-ray analyses of tweezer-complexes with macrocyclic ether-imides, reveal that the variations in binding strength between the different triplet sequences arise from the different conformational preferences of aromatic rings at diarylketone and diarylsulfone linkages. These preferences determine whether or not chain-folding and secondary π−π-stacking occurs between the arms of the tweezermolecule and the 4,4'-biphenylene units which flank the central diimide residue.

A genetic linkage map of an apple rootstock progeny anchored to the Malus genome sequence

An apple rootstock progeny raised from the cross between the very dwarfing ‘M.27’ and the more vigorous ‘M.116’ (‘M.M.106’ × ‘M.27’) was used for the construction of a linkage map comprising a total of 324 loci: 252 previously mapped SSRs, 71 newly characterised or previously unmapped SSR loci (including 36 amplified by 33 out of the 35 novel markers reported here), and the self-incompatibility locus. The map spanned the 17 linkage groups (LG) expected for apple covering a genetic distance of 1,229.5 cM, an estimated 91% of the Malus genome. Linkage groups were well populated and, although marker density ranged from 2.3 to 6.2 cM/SSR, just 15 gaps of more than 15 cM were observed. Moreover, only 17.5% of markers displayed segregation distortion and, unsurprisingly in a semi-compatible backcross, distortion was particularly pronounced surrounding the self-incompatibility locus (S) at the bottom of LG17. DNA sequences of 273 SSR markers and the S locus, representing a total of 314 loci in this investigation, were used to anchor to the ‘Golden Delicious’ genome sequence. More than 260 of these loci were located on the expected pseudo-chromosome on the ‘Golden Delicious’ genome or on its homeologous pseudo-chromosome. In total, 282.4 Mbp of sequence from 142 genome sequence scaffolds of the Malus genome were anchored to the ‘M.27’ × ‘M.116’ map, providing an interface between the marker data and the underlying genome sequence. This will be exploited for the identification of genes responsible for traits of agronomic importance such as dwarfing and water use efficiency.