Search

The authors have requested that this preprint be removed from Research Square.

Many potential applications of artificial intelligence involve making real-time decisions in physical systems while interacting with humans. Automobile racing represents an extreme example of these conditions; drivers must execute complex tactical manoeuvres to pass or block opponents while operating their vehicles at their traction limits

The automatic composition of music with long-term structure is a central problem in music generation. Neural network-based models have been shown to perform relatively well in melody generation, but generating music with long-term structure is still a major challenge. This paper introduces a new approach for music modelling that combines recent advancements of transformer models with recurrent networks – the long-short term universal transformer (LSTUT), and compare its ability to predict music against current state-of-the-art music models. Our experiments are designed to push the boundaries of music models on considerably long music sequences – a crucial requirement for learning long-term structure effectively. Results show that the LSTUT outperforms all the other models and can potentially learn features related to music structure at different time scales. Overall, we show the importance of integrating both recurrence and attention in the architecture of music models, and their potential use in future automatic composition systems.

The pre-last glacial maximum paleolake sediments at Baumkirchen, western Austria, are well known in Alpine Quaternary stratigraphy as being the type locality of the Middle to Upper Würmian transition. Their location provides a rare opportunity to investigate the vegetation history of the interior of the Alps during the last glacial cycle. A recent renewed research effort involving new drilling revealed a 250-m-thick lacustrine sequence with an older, ca. Marine Oxygen Isotope Stage (MIS) 4 phase and a younger, mid- to late MIS 3 phase. Pollen analysis reveals generally poor preservation and very low pollen concentration due to very high sedimentation rates. On the basis of pollen percentages and influx rates, six pollen zones (PZ) were assigned. PZ1 and 2 correspond to the entire ca. MIS 4 section and are characterized by only scattered vegetation representing an extremely cold and dry climate. Two stadials and two interstadials were identified in the MIS 3 section. The interstadials are characterized by well-developed open vegetation with some stands of trees, with the upper PZ6 being better developed but still forest-free. On the basis of previous radiocarbon dating, this zone (PZ6) is correlated to Greenland Interstadial (GI) 7 and the lower interstadial (PZ4) tentatively to GI 8.

A detailed understanding of the development of coral reefs and their internal structure has important applications for predicting future reef trajectories in light of climate changes, petroleum exploration and reconstructing past environmental changes such as sea level. Numerical computer modelling provides a quantitative means of testing and understanding reef processes and better predicting past and future reef development. However, most existing carbonate stratigraphic forward models (SFMs) focus on the platform scale and several limitations in their simulation of important reef processes and reef-scale features (e.g. patch reefs, sand aprons) have been previously identified. We present the Reef Sedimentary Accretion Model (ReefSAM) – a new carbonate SFM designed to test new approaches to modelling processes important at the reef scale, with the goal of better simulating reef features and evolution at higher temporal and spatial resolutions. Specific developments include a sediment transport model which calculates deposition based on a range of grain sizes, additional controls on coralgal growth including wave energy and surface substrate specificity, and model outputs showing categorized paleo-conditions (composition, paleo-depth and paleo-energy) along with numerous quantitative measures for objective comparison of model results to real-world observational data, which is often more abundant for modern/Holocene reefs than buried ancient reefs. Simulations of the well-studied Holocene One Tree Reef (OTR) in the southern Great Barrier Reef (GBR) showed improvements upon previous modelling attempts in both quantitative measures (e.g. lagoon size or reef sea level catch-up time), and the ability to produce more realistic reef-scale morphological features (e.g. lagoonal patch-reefs, mostly sand-filled lagoon) and reef development characteristics (e.g. transition from mostly vertical aggradation during sea level catch-up to horizontal progradation afterwards). Controls on reef maturity/lagoonal filling identified in previous modelling experiments were corroborated. Sensitivity testing also revealed a very high sensitivity to reef drowning. These results indicate the new modelling approaches are viable and provide a way forward for future Holocene reef evolution model development ARC-DP1094001,DP120101793 Austrian Science Fund Universität Innsbruck

The Vinschgau is an inneralpine valley in the Southern Alps, whose steep flanks are composed of strongly sheared gneisses and schists affected by deep-seated gravitational slope deformations. Fractures generated by these movements host a groundwater system characterised by high amounts of dissolved solids, which reflect long residence times and water-rock interactions driven by sulphide oxidation. This, in combination with high air and soil temperatures on the south-facing slope (Sonnenberg) of the valley favoured the deposition of calcite and aragonite flowstone in shallow parts of this fractured, non-karstic aquifer. About 5% of these speleothems exhibit well-developed regular lamination, whose annual origin was confirmed by radiometric dating. A stable isotope, petrographic and trace element study combined with U-Th dating was undertaken to disentangle the processes controlling lamina formation. A succession of darker and lighter laminae forms distinct macroscopic couplets (0.2–2 mm wide) in three of the samples, while one sample comprises alternating white and translucent laminae. Microscopically, the darker and white laminae show a higher abundance of opaque particles, whose organic origin is confirmed by their strong epifluorescence. The crystal fabric is dominated by the fascicular-optic type. The calcite shows regular δ 18 O oscillations with an amplitude of up to 1.2‰. With the exception of one sample, δ 13 C lacks such a regular pattern, and only shows smaller variations which do not correlate with δ 18 O in a consistent way. All samples exhibit regular, low-amplitude Mg, Sr, Ba and U cycles. In three samples Mg shows oscillations similar to the δ 18 O cycles in their frequency, but opposite in phasing. δ 18 O and Mg oscillations are primarily attributed to surface temperature variations that are transmitted to the shallow subsurface by thermal conduction. Low-amplitude trace element cycles (Mg, Sr, Ba and U) are most probably driven by intra-annual variations in the degree of prior aragonite precipitation in the seepage path which relates to seasonal changes in evaporation.

The Baumkirchen clay pit near Innsbruck, western Austria, is a well-known site in Alpine Quaternary stratigraphy. Lacustrine sediments from the last glacial cycle from within the Alps provide a unique opportunity to investigate the regional palaeoclimate. Recent drilling has extended the known sequence to a total length of at least 250 m consisting of almost entirely well-laminated clayey silt. Luminescence dating identified two lake sequences, separated by a hiatus of ca. 7000–15 000 years. Lake phase 1 spans the period ca. 77–55 ka, i.e. from about Marine Isotope Stage (MIS) 5/4 to the MIS 4/3 transition. Lake phase 2 extends from mid- to late MIS 3 between ca. 45 and 33 ka. Down-core X-ray fluorescence core scanning confirmed the presence of the lake phases in the sediment composition, suggesting different sediment sources and/or transport mechanisms during these two intervals. A unique section of exotic, angular, silt matrix-supported gravel at the top of lake phase 1 is interpreted as ice-rafted debris. Luminescence dating constrains this layer to ca. 55 ka, thus providing the first evidence of a late MIS 4 or early MIS 3 ice advance confined to the interior of the Eastern Alps. A conceptual model of the sedimentary history of the valley is presented.

Robots are being deployed in an increasing variety of environments for longer periods of time. As the number of robots grows, they will increasingly need to interact with other robots. Additionally, the number of companies and research laboratories producing these robots is increasing, leading to the situation where these robots may not share a common communication or coordination protocol. While standards for coordination and communication may be created, we expect that robots will need to additionally reason intelligently about their teammates with limited information. This problem motivates the area of ad hoc teamwork in which an agent may potentially cooperate with a variety of teammates in order to achieve a shared goal. This article focuses on a limited version of the ad hoc teamwork problem in which an agent knows the environmental dynamics and has had past experiences with other teammates, though these experiences may not be representative of the current teammates. To tackle this problem, this article introduces a new general-purpose algorithm, PLASTIC, that reuses knowledge learned from previous teammates or provided by experts to quickly adapt to new teammates. This algorithm is instantiated in two forms: 1) PLASTIC-Model - which builds models of previous teammates' behaviors and plans behaviors online using these models and 2) PLASTIC-Policy - which learns policies for cooperating with previous teammates and selects among these policies online. We evaluate PLASTIC on two benchmark tasks: the pursuit domain and robot soccer in the RoboCup 2D simulation domain. Recognizing that a key requirement of ad hoc teamwork is adaptability to previously unseen agents, the tests use more than 40 previously unknown teams on the first task and 7 previously unknown teams on the second. While PLASTIC assumes that there is some degree of similarity between the current and past teammates' behaviors, no steps are taken in the experimental setup to make sure this assumption holds. The teammates were created by a variety of independent developers and were not designed to share any similarities. Nonetheless, the results show that PLASTIC was able to identify and exploit similarities between its current and past teammates' behaviors, allowing it to quickly adapt to new teammates.

This study seeks to reconstruct the music for Boethius’s final and most widely read work, On the Consolation of Philosophy. Although a handful of neumations for Boethius’s thirty-nine poems have long been known, the almost complete absence of surviving pitched versions of the melodies has hindered the task of reconstruction. Following a systematic study of neumed manuscripts dating from the ninth to the eleventh centuries and identification of underlying principles of melodic design, it is now possible to attempt informed reconstructions. Even so, the leap from scholarship to modern performance remains substantial, involving a preliminary need to recreate melodies through experimentation. Collaboration with members of the ensemble Sequentia over a four-year period (2014–17) provided an opportunity to explore ways in which creative practice might supplement scholarly knowledge, whether through posing new research questions, the formation and exploration of hypotheses, or recourse to memorized practices built up through sustained engagement with early medieval repertories and instruments. Taking a recovered mid eleventh-century leaf of the Cambridge Songs as a focus for investigation, songs from the first book of Boethius’s De consolatione philosophiae were newly reconstructed. Although debate continues to surround the status of results obtained through creative practice, comparison with the way other disciplines have proceeded under similar conditions reveals acceptance of experimentation as a mode of inquiry. This essay proposes that performance can fruitfully supplement philology in seeking to reconstruct lost songs from notated traces.

The Holocene reef in the outer Great Barrier Reef (GBR) represents an archetypal reef system, forming a thin veneer (10–30 m) built upon an older Pleistocene reef surface. The morphology, stratigraphy and maturity (degree of lagoonal sediment infilling) of the modern reef results from the complex interplay between biologic and abiologic processes (reef accretion, sediment erosion, transport and deposition), basement substrate, and Holocene sea level rise. Combining 3D forward stratigraphic modelling (CARBONATE-3D) with a re-analysis of published observational data, we quantitatively simulate the Holocene evolution of One Tree Reef (Southern GBR) as a well constrained, model system, and explore the main processes affecting reef growth in the GBR and elsewhere. We test the influence of different basement substrate surfaces, sea level curves, reef accretion rates, sediment erosion and transport parameters and assess their relative importance in controlling reef evolution—particularly growth histories, 3D internal structure and stratigraphy and reef maturity. Quantitative comparisons between our “best estimate” model output and the observed data confirm that we are able to simulate a 75% match of the main morphologic and growth characteristics of One Tree Reef. The range of parameters tested produced the full spectrum of reef maturities from unfilled “juvenile” buckets to planar “senile” reefs with completely sediment infilled lagoons. We conclude that the shape and depth of the basement substrate has the strongest influence—significantly impacting reef evolution and final maturity including the shape of the “bucket”, the size of the reef margins and internal reef structure. In contrast, variations in sea level, sediment production, erosion and transport mainly controlled the degree of lagoonal sediment filling. This study has implications for better understanding the past evolution of the GBR and other reefs but also lays the foundation for improved predictions of possible trajectories of modern reefs in general in the face of future environmental changes.

We propose to combine short-term block-based fuzzy support vector machine (FSVM) learning and long-term dynamic semantic clustering (DSC) learning to bridge the semantic gap in content-based image retrieval. The short-term learning addresses the small sample problem by incorporating additional image blocks to enlarge the training set. Specifically, it applies the nearest neighbor mechanism to choose additional similar blocks. A fuzzy metric is computed to measure the fidelity of the actual class information of the additional blocks. The FSVM is finally applied on the enlarged training set to learn a more accurate decision boundary for classifying images. The long-term learning addresses the large storage problem by building dynamic semantic clusters to remember the semantics learned during all query sessions. Specifically, it applies a cluster-image weighting algorithm to find the images most semantically related to the query. It then applies a DSC technique to adaptively learn and update the semantic categories. Our extensive experimental results demonstrate that the proposed short-term, long-term, and collaborative learning methods outperform their peer methods when the erroneous feedback resulting from the inherent subjectivity of judging relevance, user laziness, or maliciousness is involved. The collaborative learning system achieves better retrieval precision and requires significantly less storage space than its peers. © 2011 Wiley Periodicals, Inc. © 2011 Wiley Periodicals, Inc.

Chapter 5 introduced the algorithms used in this book for solving ad hoc teamwork problems. Before moving on to the empirical analysis of these algorithms in Chap. 7, it is useful to first investigate the theoretical attributes of PLASTIC. Our analysis focuses on whether the multi-armed bandit domain described in Sect. 3.2.1 is tractable for PLASTIC–Model. We chose to analyze the bandit domain because of its simplicity, which lends itself to more complete theoretical analysis. In addition, the bandit domain is interesting due to its use of communication, which is an important aspect of ad hoc teamwork that is not explored in the other domains. Note that we do not investigate the model learning aspect of PLASTIC–Model. Instead, we analyze whether the PLASTIC–Model can select from a set of known models (from \(\text {HandCodedKnowledge}\)) and plan its response to these models in polynomial time.