DEVELOPMENT OF INTEGRATED BIOPROCESSING TECHNOLOGIES FOR THE PRODUCTION OF PECTIC OLIGOSACCHARIDES (POS) FROM AGRO-PROCESSING RESIDUES

This research deals with the production of pectic oligosaccharides (POS) from agro-industrial residues, with specific focus on development of continuous cross flow enzyme membrane reactor. Pectic oligosaccharides have recently gained attention due to their prebiotic activity. Lack of information on the continuous production of POS from agro-industrial residues formed the basis for the present study. Four residues i.e sugar beet pulp, onion hulls, pressed pumpkin cake and berry pomace were taken to study their pectin content. Based on the presence of higher galacturonic acid and arabinose (both homogalacturonan and rhamnogalacturonan) in sugar beet pulp and galacturonic acid (only homogalacturonan) in onion hulls, further optimization of different extraction methods of pectin (causing minimum damage to pectic chain) from these residues were done. The most suitable extractant for sugar beet pulp and onion hulls were nitric acid and sodium hexametaphosphate respectively. Further the experiments on the continuous production of POS from sugar beet pulp in an enzyme membrane reactor was initiated. Several optimization experiments indicated the optimum enzyme (Viscozyme) as well as feed concentration (25 g/L) to be used for producing POS from sugar beet pulp in an enzyme membrane reactor. The results highlighted that steady state POS production with volumetric and specific productivity of 22g/L/h and 11 g/gE/h respectively could be achieved by continuous cross flow filtration of sugar beet pulp pectic extract over 10 kDa membrane at residence time of 20 min. The POS yield of about 80% could be achieved using above conditions. Also, in this thesis preliminary experiments on the production and characterization of POS from onion hulls were conducted. The results revelaed that the most suitable enzyme for POS production from onion hulls is endo-polygalacturonase M2. The POS produced from onion hulls were present in the form of DP1 -DP10 in substituted as well as unsubstituted forms. This study clearly demonstrates that continuous production of POS from pectin rich sources can be achieved by using cross flow continuous enzyme membrane reactor.

Curvature-driven smoothing: a learning algorithm for feed-forward networks

The performance of feed-forward neural networks in real applications can be often be improved significantly if use is made of a-priori information. For interpolation problems this prior knowledge frequently includes smoothness requirements on the network mapping, and can be imposed by the addition to the error function of suitable regularization terms. The new error function, however, now depends on the derivatives of the network mapping, and so the standard back-propagation algorithm cannot be applied. In this paper, we derive a computationally efficient learning algorithm, for a feed-forward network of arbitrary topology, which can be used to minimize the new error function. Networks having a single hidden layer, for which the learning algorithm simplifies, are treated as a special case.

Regularization and complexity control in feed-forward networks

In this paper we consider four alternative approaches to complexity control in feed-forward networks based respectively on architecture selection, regularization, early stopping, and training with noise. We show that there are close similarities between these approaches and we argue that, for most practical applications, the technique of regularization should be the method of choice.

Topographic mappings and feed-forward neural networks

This thesis is a study of the generation of topographic mappings - dimension reducing transformations of data that preserve some element of geometric structure - with feed-forward neural networks. As an alternative to established methods, a transformational variant of Sammon's method is proposed, where the projection is effected by a radial basis function neural network. This approach is related to the statistical field of multidimensional scaling, and from that the concept of a 'subjective metric' is defined, which permits the exploitation of additional prior knowledge concerning the data in the mapping process. This then enables the generation of more appropriate feature spaces for the purposes of enhanced visualisation or subsequent classification. A comparison with established methods for feature extraction is given for data taken from the 1992 Research Assessment Exercise for higher educational institutions in the United Kingdom. This is a difficult high-dimensional dataset, and illustrates well the benefit of the new topographic technique. A generalisation of the proposed model is considered for implementation of the classical multidimensional scaling (¸mds}) routine. This is related to Oja's principal subspace neural network, whose learning rule is shown to descend the error surface of the proposed ¸mds model. Some of the technical issues concerning the design and training of topographic neural networks are investigated. It is shown that neural network models can be less sensitive to entrapment in the sub-optimal global minima that badly affect the standard Sammon algorithm, and tend to exhibit good generalisation as a result of implicit weight decay in the training process. It is further argued that for ideal structure retention, the network transformation should be perfectly smooth for all inter-data directions in input space. Finally, there is a critique of optimisation techniques for topographic mappings, and a new training algorithm is proposed. A convergence proof is given, and the method is shown to produce lower-error mappings more rapidly than previous algorithms.