Search for: All records

critical to reveal a blackbox model’s decision-making process from raw data to prediction. In this article, we use two real datasets, the MNIST handwritten digits and MIT-BIH Electrocardiogram (ECG) signals, to motivate key characteristics of discriminative features, namely adaptiveness, predictive importance and effectiveness. Then, we develop a localization framework based on adversarial attacks to effectively localize discriminative features. In contrast to existing heuristic methods, we also provide a statistically guaranteed interpretability of the localized features by measuring a generalized partial R2. We apply the proposed method to the MNIST dataset and the MIT-BIH dataset with a convolutional auto-encoder. In the first, the compact image regions localized by the proposed method are visually appealing. Similarly, in the second, the identified ECG features are biologically plausible and consistent with cardiac electrophysiological principles while locating subtle anomalies in a QRS complex that may not be discernible by the naked eye. Overall, the proposed method compares favorably with state-of-the-art competitors. Accompanying this paper is a Python library dnn-locate that implements the proposed approach. 

Objective. Dynamic positron emission tomography (PET) imaging, which can provide information on dynamic changes in physiological metabolism, is now widely used in clinical diagnosis and cancer treatment. However, the reconstruction from dynamic data is extremely challenging due to the limited counts received in individual frame, especially in ultra short frames. Recently, the unrolled modelbased deep learning methods have shown inspiring results for low-count PET image reconstruction with good interpretability. Nevertheless, the existing model-based deep learning methods mainly focus on the spatial correlations while ignore the temporal domain. Approach. In this paper, inspired by the learned primal dual (LPD) algorithm, we propose the spatio-temporal primal dual network (STPDnet) for dynamic low-count PET image reconstruction. Both spatial and temporal correlations are encoded by 3D convolution operators. The physical projection of PET is embedded in the iterative learning process of the network, which provides the physical constraints and enhances interpretability. Main results. The experiments of both simulation data and real rat scan data have shown that the proposed method can achieve substantial noise reduction in both temporal and spatial domains and outperform the maximum likelihood expectation maximization, spatio-temporal kernel method, LPD and FBPnet. Significance. Experimental results show STPDnet better reconstruction performance in the low count situation, which makes the proposed method particularly suitable in whole-body dynamic imaging and parametric PET imaging that require extreme short frames and usually suffer from high level of noise. 

Statistical inference of directed relations given some unspecified interventions (i.e., the intervention targets are unknown) is challenging. In this article, we test hypothesized directed relations with unspecified interventions. First, we derive conditions to yield an identifiable model. Unlike classical inference, testing directed relations requires identifying the ancestors and relevant interventions of hypothesis-specific primary variables. To this end, we propose a peeling algorithm based on nodewise regressions to establish a topological order of primary variables. Moreover, we prove that the peeling algorithm yields a consistent estimator in low-order polynomial time. Second, we propose a likelihood ratio test integrated with a data perturbation scheme to account for the uncertainty of identifying ancestors and interventions. Also, we show that the distribution of a data perturbation test statistic converges to the target distribution. Numerical examples demonstrate the utility and effectiveness of the proposed methods, including an application to infer gene regulatory networks. The R implementation is available at https://github.com/chunlinli/intdag. 

The past ∼200 million years of Earth's geomagnetic field behavior have been recorded within oceanic basalts, many of which are only accessible via scientific ocean drilling. Obtaining the best possible paleomagnetic measurements from such valuable samples requires an a priori understanding of their magnetic mineralogies when choosing the most appropriate protocol for stepwise demagnetization experiments (either alternating field or thermal). Here, we present a quick, and non‐destructive method that utilizes the amplitude‐dependence of magnetic susceptibility to screen submarine basalts prior to choosing a demagnetization protocol, whenever conducting a pilot study or other detailed rock‐magnetic characterization is not possible. We demonstrate this method using samples acquired during International Ocean Discovery Program Expedition 391. Our approach is rooted in the observation that amplitude‐dependent magnetic susceptibility is observed in basalt samples whose dominant magnetic carrier is multidomain titanomagnetite (∼TM_60–65, (Ti_0.60–0.65Fe_0.35–0.40)Fe₂O₄). Samples with low Ti contents within titanomagnetite or samples that have experienced a high degree of oxidative weathering do not display appreciable amplitude dependence. Due to their low Curie temperatures, basalts that possess amplitude‐dependence should ideally be demagnetized either using alternating fields or via finely‐spaced thermal demagnetization heating steps below 300°C. Our screening method can enhance the success rate of paleomagnetic studies of oceanic basalt samples.

 

Electric Vehicle (EV) charging recommendation that both accommodates user preference and adapts to the ever-changing external environment arises as a cost-effective strategy to alleviate the range anxiety of private EV drivers. Previous studies focus on centralized strategies to achieve optimized resource allocation, particularly useful for privacy-indifferent taxi fleets and fixed-route public transits. However, private EV driver seeks a more personalized and resource-aware charging recommendation that is tailor-made to accommodate the user preference (when and where to charge) yet sufficiently adaptive to the spatiotemporal mismatch between charging supply and demand. Here we propose a novel Regularized Actor-Critic (RAC) charging recommendation approach that would allow each EV driver to strike an optimal balance between the user preference (historical charging pattern) and the external reward (driving distance and wait time). Experimental results on two real-world datasets demonstrate the unique features and superior performance of our approach to the competing methods. 

Human intervention and climate change jointly influence the functions and dynamics of marine ecosystems. Studying the impacts of human and climate on ecosystem dynamics is challenging. Unlike experimental studies, research on natural systems is not amendable at the scale of time, space, and biology. With confounding factors well balanced for two adjacent subtropical estuaries except urbanized disturbances, we conducted ecosystem modelling using indirect reasoning by exclusion to quantify the relative impacts of human disruption on estuarine ecosystems under climate variability. One major finding of this study is that the human intervention tends to magnify species fluctuations, complicate the species interaction network, and enhance species interaction strength combined with disclosed downscaling climate effects (indexed as North Atlantic Oscillation and Atlantic Multi-decadal Oscillation) on estuarine hydrology and biological communities. In addition, functional groups appeared to respond more diversely to external forcing in company with human interventions. While human perturbation was shown to destabilize the estuarine ecosystems, making them vulnerable to environmental variability under climate change, buffering effects of species diversity and trophic interaction tend to underpin the ecosystem functions. The findings of this study contribute to the holistic assessment and strategic management of estuarine ecosystems subjected to human and natural disturbances in the face of climate change.

 

Hotspot tracks (chains of seamounts, ridges, and other volcanic structures) provide important records of plate motions, as well as mantle geodynamics, magma flux, and mantle source compositions. The Tristan-Gough-Walvis Ridge (TGW) hotspot track, extending from the active volcanic islands of Tristan da Cunha and Gough through a province of guyots and then along Walvis Ridge to the Etendeka flood basalt province, forms one of the most prominent and complex global hotspot tracks. The TGW hotspot track displays a tight linear age progression in which ages increase from the islands to the flood basalts (covering ~135 My). Unlike Pacific tracks, which are often simple, nearly linear chains of seamounts, the TGW track is alternately a steep-sided narrow ridge, an oceanic plateau, subparallel linear ridges and chains of seamounts (most are flat-topped guyots). The track displays isotopic zonation over the last ~70 My. The zonation appears near the middle of the track just before it splits into two to three chains of ridge- and guyot-type seamounts. Walvis Ridge, forming the older part of the track, is also overprinted with age-progressive late-stage volcanism, which was emplaced ~30–40 My after the initial eruptions and has a distinct isotopic composition. The plan for Expedition 391 was to drill at six sites, three along Walvis Ridge and three in the seamounts of the Guyot Province, to collect igneous rocks to better understand the formation of volcanic edifices, the temporal and geochemical evolution of the hotspot, and the variation in paleolatitudes at which the volcanic edifices formed. After a delay of 18 days to address a shipboard Coronavirus (COVID-19) outbreak, Expedition 391 proceeded to drill at four of the proposed sites: three sites on Walvis Ridge around Valdivia Bank, an ocean plateau within the ridge, and one site on the lower flank of a guyot in the Center track of the Guyot Province, a ridge located between the Tristan subtrack (which extends from the end of Walvis Ridge to the islands of Tristan da Cunha) and the Gough subtrack (which extends from Walvis Ridge to Gough Island). The first hole was drilled at Site U1575, located on a low portion of the northeastern Walvis Ridge just north of Valdivia Bank. At this location, 209.9 m of sediments and 122.4 m of igneous basement were cored. The sediments ranged in age from Late Pleistocene (~0.43–1.24 Ma) to Late Cretaceous (Campanian; 72–78 Ma). The igneous basement comprised 10 submarine lava units consisting of pillow, lobate, sheet, and massive lava flows, the thickest of which was ~21 m. Most lavas are tholeiitic, but some alkalic basalts were recovered. A portion of the igneous succession consists of low-Ti basalts, which are unusual because they appear in the Etendeka flood basalts but have not been previously found on Walvis Ridge. Two holes were drilled at Site U1576 on the west flank of Valdivia Bank. The first of these holes was terminated because a bit jammed shortly after entering the igneous basement. Hole U1576A recovered a remarkable ~380 m thick sedimentary section consisting mostly of chalk covering a nearly complete sequence from Late Pleistocene (~0.43–1.24 Ma) to Late Cretaceous (Campanian; ~79–81.38 Ma). These sediments display short and long cyclic color changes that imply astronomically forced and longer term paleoenvironmental changes. The igneous basement recovered in Hole U1576B yielded 11 submarine lava units (total thickness = ~65 m). The flows range from pillows to massive flows with compositions varying from tholeiitic basalt to basaltic andesite, only the second occurrence of the latter composition recovered from the TGW track thus far. These units are separated by seven sedimentary chalk units that range 0.1–11.6 m in thickness, implying a long-term interplay of sedimentation and lava eruptions. These intercalated sediments revealed Upper Cretaceous (Campanian) ages of ~77–79 Ma for the upper two interbeds and ~79–81.38 Ma for the lower beds. Coring at Site U1577, on the extreme eastern flank of Valdivia Bank, penetrated a 154.8 m thick sedimentary section ranging from the Paleocene (Thanetian; ~58.8 Ma) to Upper Cretaceous (Campanian; ~81.43–83.20 Ma). Igneous basement coring progressed only 39.1 m below the sediment/basalt contact, recovering three massive submarine tholeiitic basalt lava flows that are 4.1, 15.5, and >19.1 m thick, respectively. Paleomagnetic data from Sites U1577 and U1576 indicate that the former volcanic basement formed just before the end of the Cretaceous Normal Superchron and the latter during Chron 33r, shortly afterward. Biostratigraphic and paleomagnetic data suggest that Valdivia Bank becomes younger from east to west. Site U1578, located on a Center track guyot, provided a long and varied igneous section. After coring through 184.3 m of pelagic carbonate sediments mainly consisting of Eocene and Paleocene chalk (~55.64–63.5 Ma), Hole U1578A cored 302.1 m of igneous basement. Basement lavas are largely pillows but are interspersed with sheet and massive flows. Lava compositions are mostly alkalic basalts with some hawaiite. Several intervals contain abundant olivine (some fresh), and some of the pillow stacks consist of basalt with remarkably high Ti content. The igneous sequence is interrupted by 10 sedimentary interbeds consisting of chalk and volcaniclastics and ranging 0.46–10.19 m in thickness. Investigations of toothpick samples from the intercalated sediments were examined, each revealing the same age range of ~63.5–64.81 Ma (lower Paleocene; Danian). Paleomagnetic data display a change in basement magnetic polarity ~100 m above the base of the hole. Combining magnetic stratigraphy with biostratigraphic data, the igneous section is inferred to span >1 My. Nearly 7 months after Expedition 391, JOIDES Resolution transited from Cape Town to the north Atlantic. During this transit (Expedition 397T), 7.9 days of ship time were used to drill two holes (U1584A and U1585A) at sites on the Gough and Tristan tracks that had been omitted because of COVID-19–related time loss on the earlier cruise. For both, coring was begun only a short distance above the igneous basement to save time. The 75.2 m thick section drilled in Hole U1584A contains two sedimentary units: clay-rich carbonate sediments overlie a pumice-dominated volcaniclastic deposit containing basalt fragments. Because the goal was to core basalt and the base of the volcaniclastic deposit was not imaged in the seismic profile, the hole was terminated early to save operation time for the next site. In Hole U1585A, coring penetrated a 273.5 m thick sediment section overlying an 81.2 m thick pile of massive basalt flows. The sediment section is divided into four units: The uppermost unit consists of nannofossil chalk; The two intermediate units contain alternating chalk and volcaniclastic sediments containing several breccia units; and The lowermost unit consists of volcanic breccia containing juvenile blocks, bombs, and accretionary lapilli. This thick sedimentary section documents a transition from shallow-water volcanism to open-ocean sedimentation as the seamount subsided. The thick underlying basalt section is made up of four sparsely to highly phyric massive flows, the thickest of which is >43 m thick. Samples of these units are mostly basalt with a few trachybasalts and one trachyandesite. Although the igneous penetration was less than planned, coring during Expeditions 391 and 397T obtained samples that clearly will lead to an improved understanding of the evolution of the TGW hotspot and its track. Reasonable recovery of fresh basalt in some holes provides ample samples for geochemical, geochronologic, and paleomagnetic studies. Good recovery of Late Cretaceous and early Cenozoic chalk successions provides samples for paleoenvironmental study. 

The strategy for International Ocean Discovery Program (IODP) Expedition 391 (Walvis Ridge Hotspot) was to drill at three general locations on Walvis Ridge and one in Guyot Province, providing an age transect along the Tristan-Gough-Walvis (TGW) hotspot track. Site U1575 (proposed Site FR-1B), located on the lower Walvis Ridge between Valdivia Bank and Frio Ridge (Figure F1), is the easternmost and presumably the oldest site. Both hotspot models and the age progression of Homrighausen et al. (2019) predict an age of ~100 Ma (Figures F2, F3). Site U1575 is thus an important sample of the early TGW track shortly after it transitioned from the continental flood basalt to the submarine hotspot track setting.