The concept of in situ multisatellite altimetry calibration and validation in the absolute sense using ocean autonomous surface vehicles as global navigation satellite systems (GNSS) platforms is demonstrated through an experiment in the North Sea during 2016. A Wave Glider (WG) equipped with geodetic GNSS traveled to locations ranging from 21 to 78 km from the coast to be directly under four Jason-series tracks and two CryoSat-2 tracks. 5-Hz sea surface heights (SSHs) were estimated from precise point positioning (PPP) mode processing of GPS+GLONASS data, together with hourly zenith wet tropospheric delays (ZWDs), and used as reference values for altimetry satellite measured SSH, tropospheric delay, and significant wave height (SWH). SSH biases obtained were −30 to −8 mm for Jason-2 using geophysical data record (GDR)-D products, −40 to +1 mm for Jason-3 using GDR-F products, and −29 and +18 mm for CryoSat-2 using SAR mode GOP baseline C products. These biases are almost commensurate with results from previous studies in other regions that used GNSS buoys or onshore GNSS reference stations with geoid and tide extrapolation. The Jason-2 and Jason-3 microwave radiometer (MWR)-measured ZWDs differed, respectively, by −15 and −10 mm on average from those measured by the GNSS WG. Root-mean-square SWH differences of 2–6 cm were obtained between Jason-2/3 and the co-located GNSS WG, and equivalent differences of 19–21 cm for CryoSat-2.

We present and analyze a novel regularization technique based on enhancing our dataset with corrupted copies of our original data. The motivation is that since the learning algorithm lacks information about which parts of the data are reliable, it has to make more robust classification functions. Using this framework, we propose a simple addition to the gentle boosting algorithm which enables it to work with only a few examples. We test this new algorithm on a variety of datasets and show convincing results.

We give simple sufficient conditions for a code to be rectangular and show that large families of well-known nonlinear codes are rectangular. These include Hadamard (1893), Levenshtein (1964), Delsarte-Goethals (1975), Kerdock (1972), and Nordstrom-Robinson (1967) codes. Being rectangular, each of these codes has a unique minimal trellis that can be used for soft-decision maximum-likelihood decoding.