Code-specific learning rules improve action selection by populations of spiking neurons

Population coding is widely regarded as a key mechanism for achieving reliable behavioral decisions. We previously introduced reinforcement learning for population-based decision making by spiking neurons. Here we generalize population reinforcement learning to spike-based plasticity rules that take account of the postsynaptic neural code. We consider spike/no-spike, spike count and spike latency codes. The multi-valued and continuous-valued features in the postsynaptic code allow for a generalization of binary decision making to multi-valued decision making and continuous-valued action selection. We show that code-specific learning rules speed up learning both for the discrete classification and the continuous regression tasks. The suggested learning rules also speed up with increasing population size as opposed to standard reinforcement learning rules. Continuous action selection is further shown to explain realistic learning speeds in the Morris water maze. Finally, we introduce the concept of action perturbation as opposed to the classical weight- or node-perturbation as an exploration mechanism underlying reinforcement learning. Exploration in the action space greatly increases the speed of learning as compared to exploration in the neuron or weight space.

Growth Codes: Intermediate Performance Analysis and Application to Video

Growth codes are a subclass of Rateless codes that have found interesting applications in data dissemination problems. Compared to other Rateless and conventional channel codes, Growth codes show improved intermediate performance which is particularly useful in applications where partial data presents some utility. In this paper, we investigate the asymptotic performance of Growth codes using the Wormald method, which was proposed for studying the Peeling Decoder of LDPC and LDGM codes. Compared to previous works, the Wormald differential equations are set on nodes' perspective which enables a numerical solution to the computation of the expected asymptotic decoding performance of Growth codes. Our framework is appropriate for any class of Rateless codes that does not include a precoding step. We further study the performance of Growth codes with moderate and large size codeblocks through simulations and we use the generalized logistic function to model the decoding probability. We then exploit the decoding probability model in an illustrative application of Growth codes to error resilient video transmission. The video transmission problem is cast as a joint source and channel rate allocation problem that is shown to be convex with respect to the channel rate. This illustrative application permits to highlight the main advantage of Growth codes, namely improved performance in the intermediate loss region.

Sparse computation for large-scale binary classification

Well-known data mining algorithms rely on inputs in the form of pairwise similarities between objects. For large datasets it is computationally impossible to perform all pairwise comparisons. We therefore propose a novel approach that uses approximate Principal Component Analysis to efficiently identify groups of similar objects. The effectiveness of the approach is demonstrated in the context of binary classification using the supervised normalized cut as a classifier. For large datasets from the UCI repository, the approach significantly improves run times with minimal loss in accuracy.

Cancer intelligence acquired (CIA): tumor glycosylation and sialylation codes dismantling antitumor defense

Aberrant glycosylation is a key feature of malignant transformation and reflects epigenetic and genetic anomalies among the multitude of molecules involved in glycan biosynthesis. Although glycan biosynthesis is not template bound, altered tumor glycosylation is not random, but associated with common glycosylation patterns. Evidence suggests that acquisition of distinct glycosylation patterns evolves from a ‘microevolutionary’ process conferring advantages in terms of tumor growth, tumor dissemination, and immune escape. Such glycosylation modifications also involve xeno- and hypersialylation. Xeno-autoantigens such as Neu5Gc-gangliosides provide potential targets for immunotherapy. Hypersialylation may display ‘enhanced self’ to escape immunosurveillance and involves several not mutually exclusive inhibitory pathways that all rely on protein–glycan interactions. A better understanding of tumor ‘glycan codes’ as deciphered by lectins, such as siglecs, selectins, C-type lectins and galectins, may lead to novel treatment strategies, not only in cancer, but also in autoimmune disease or transplantation.

RC-NDN: Raptor Codes Enabled Named Data Networking

Information-centric networking (ICN) has been proposed to cope with the drawbacks of the Internet Protocol, namely scalability and security. The majority of research efforts in ICN have focused on routing and caching in wired networks, while little attention has been paid to optimizing the communication and caching efficiency in wireless networks. In this work, we study the application of Raptor codes to Named Data Networking (NDN), which is a popular ICN architecture, in order to minimize the number of transmitted messages and accelerate content retrieval times. We propose RC-NDN, which is a NDN compatible Raptor codes architecture. In contrast to other coding-based NDN solutions that employ network codes, RC-NDN considers security architectures inherent to NDN. Moreover, different from existing network coding based solutions for NDN, RC-NDN does not require significant computational resources, which renders it appropriate for low cost networks. We evaluate RC-NDN in mobile scenarios with high mobility. Evaluations show that RC-NDN outperforms the original NDN significantly. RC-NDN is particularly efficient in dense environments, where retrieval times can be reduced by 83% and the number of Data transmissions by 84.5% compared to NDN.

CODE’s new ultra-rapid orbit and ERP products for the IGS

The International GNSS Service (IGS) issues four sets of so-called ultra-rapid products per day, which are based on the contributions of the IGS Analysis Centers. The traditional (“old”) ultra-rapid orbit and earth rotation parameters (ERP) solution of the Center for Orbit Determination in Europe (CODE) was based on the output of three consecutive 3-day long-arc rapid solutions. Information from the IERS Bulletin A was required to generate the predicted part of the old CODE ultra-rapid product. The current (“new”) product, activated in November 2013, is based on the output of exactly one multi-day solution. A priori information from the IERS Bulletin A is no longer required for generating and predicting the orbits and ERPs. This article discusses the transition from the old to the new CODE ultra-rapid orbit and ERP products and the associated improvement in reliability and performance. All solutions used in this article were generated with the development version of the Bernese GNSS Software. The package was slightly extended to meet the needs of the new CODE ultra-rapid generation.

CODE’s new solar radiation pressure model for GNSS orbit determination

The Empirical CODE Orbit Model (ECOM) of the Center for Orbit Determination in Europe (CODE), which was developed in the early 1990s, is widely used in the International GNSS Service (IGS) community. For a rather long time, spurious spectral lines are known to exist in geophysical parameters, in particular in the Earth Rotation Parameters (ERPs) and in the estimated geocenter coordinates, which could recently be attributed to the ECOM. These effects grew creepingly with the increasing influence of the GLONASS system in recent years in the CODE analysis, which is based on a rigorous combination of GPS and GLONASS since May 2003. In a first step we show that the problems associated with the ECOM are to the largest extent caused by the GLONASS, which was reaching full deployment by the end of 2011. GPS-only, GLONASS-only, and combined GPS/GLONASS solutions using the observations in the years 2009–2011 of a global network of 92 combined GPS/GLONASS receivers were analyzed for this purpose. In a second step we review direct solar radiation pressure (SRP) models for GNSS satellites. We demonstrate that only even-order short-period harmonic perturbations acting along the direction Sun-satellite occur for GPS and GLONASS satellites, and only odd-order perturbations acting along the direction perpendicular to both, the vector Sun-satellite and the spacecraft’s solar panel axis. Based on this insight we assess in the third step the performance of four candidate orbit models for the future ECOM. The geocenter coordinates, the ERP differences w. r. t. the IERS 08 C04 series of ERPs, the misclosures for the midnight epochs of the daily orbital arcs, and scale parameters of Helmert transformations for station coordinates serve as quality criteria. The old and updated ECOM are validated in addition with satellite laser ranging (SLR) observations and by comparing the orbits to those of the IGS and other analysis centers. Based on all tests, we present a new extended ECOM which substantially reduces the spurious signals in the geocenter coordinate z (by about a factor of 2–6), reduces the orbit misclosures at the day boundaries by about 10 %, slightly improves the consistency of the estimated ERPs with those of the IERS 08 C04 Earth rotation series, and substantially reduces the systematics in the SLR validation of the GNSS orbits.

Index Tracking Using Data-Mining Techniques and Mixed-Binary Linear Programming

Index tracking has become one of the most common strategies in asset management. The index-tracking problem consists of constructing a portfolio that replicates the future performance of an index by including only a subset of the index constituents in the portfolio. Finding the most representative subset is challenging when the number of stocks in the index is large. We introduce a new three-stage approach that at first identifies promising subsets by employing data-mining techniques, then determines the stock weights in the subsets using mixed-binary linear programming, and finally evaluates the subsets based on cross validation. The best subset is returned as the tracking portfolio. Our approach outperforms state-of-the-art methods in terms of out-of-sample performance and running times.

Binary recombinase systems for high-resolution conditional mutagenesis.

Conditional mutagenesis using Cre recombinase expressed from tissue specific promoters facilitates analyses of gene function and cell lineage tracing. Here, we describe two novel dual-promoter-driven conditional mutagenesis systems designed for greater accuracy and optimal efficiency of recombination. Co-Driver employs a recombinase cascade of Dre and Dre-respondent Cre, which processes loxP-flanked alleles only when both recombinases are expressed in a predetermined temporal sequence. This unique property makes Co-Driver ideal for sequential lineage tracing studies aimed at unraveling the relationships between cellular precursors and mature cell types. Co-InCre was designed for highly efficient intersectional conditional transgenesis. It relies on highly active trans-splicing inteins and promoters with simultaneous transcriptional activity to reconstitute Cre recombinase from two inactive precursor fragments. By generating native Cre, Co-InCre attains recombination rates that exceed all other binary SSR systems evaluated in this study. Both Co-Driver and Co-InCre significantly extend the utility of existing Cre-responsive alleles.