A confidence-based approach for improving keyword hypothesis scores

The task in keyword spotting (KWS) is to hypothesise times at which any of a set of key terms occurs in audio. An important aspect of such systems are the scores assigned to these hypotheses, the accuracy of which have a significant impact on performance. Estimating these scores may be formulated as a confidence estimation problem, where a measure of confidence is assigned to each key term hypothesis. In this work, a set of discriminative features is defined, and combined using a conditional random field (CRF) model for improved confidence estimation. An extension to this model to directly address the problem of score normalisation across key terms is also introduced. The implicit score normalisation which results from applying this approach to separate systems in a hybrid configuration yields further benefits. Results are presented which show notable improvements in KWS performance using the techniques presented in this work. © 2013 IEEE.

Small-scale turbulent fluctuations beyond Taylor's frozen-flow hypothesis

The space-time cross-correlation function C-T(r, tau) of local temperature fluctuations in turbulent Rayleigh-Benard convection is obtained from simultaneous two-point time series measurements. The obtained C-T(r, tau) is found to have the scaling form C-T(r(E), 0) with r(E)=[(r-U tau)(2)+ V-2 tau(2)](1/2), where U and V are two characteristic velocities associated with the mean and rms velocities of the flow. The experiment verifies the theory and demonstrates its applications to a class of turbulent flows in which the requirement of Taylor's frozen flow hypothesis is not met.

In vitro iron enrichment experiments in the Prydz Bay, the Southern Ocean: A test of the iron hypothesis

In "high nitrate, low chlorophyll" (HNLC) ocean regions, iron has been typically regarded as the limiting factor for phytoplankton production. This "iron hypothesis" needs to be tested in various oceanic environments to understand the role of iron in marine biological and biogeochemical processes. In this paper, three in vitro iron enrichment experiments were performed in Prydz Bay and at the Polar Front north of the Ross Sea, to study the role of iron on phytoplankton production. At the Polar Front of Ross Sea, iron addition significantly (P < 0.05, Student's t-test) stimulated phytoplankton growth. In Prydz Bay, however, both the iron treatments and the controls showed rapid phytoplankton growth, and no significant effect (P > 0.05, Student's t-test) as a consequence of iron addition was observed. These results confirmed the limiting role of iron in the Ross Sea and indicated that iron was not the primary factor limiting phytoplankton growth in Prydz Bay. Because the light environment for phytoplankton was enhanced in experimental bottles, light was assumed to be responsible for the rapid growth of phytoplankton in all treatments and to be the limiting factor controlling field phytoplankton growth in Prydz Bay. During the incubation experiments, nutrient consumption ratios also changed with the physiological status and the growth phases of phytoplankton cells. When phytoplankton growth was stimulated by iron addition, N was the first and Si was the last nutrient which absorption enhanced. The Si/N and Si/P consumption ratios of phytoplankton in the stationary and decay phases were significantly higher than those of rapidly growing phytoplankton. These findings were helpful for studies of the marine ecosystem and biogeochemistry in Prydz Bay, and were also valuable for biogeochemical studies of carbon and nutrients in various marine environments.

长白山火山成因—物理机制和化学演化

This thesis is one of the contributions to NSFC project, “The Changbaishan Volcanism and its Links to the Northeast Asia Tectonic System”. The thesis presents our most recent works on Changbaishan Volcanism, on two aspects as (1) the chemical evolution of the Cenozoic volcanism and the physical links of magma genesis, (2) the Holocene activity of the Tianchi Volcano and risks of potential eruption. 1. Chemical evolution and physical links to the Changbaishan Volcanism Physical links to the Chanbaishan Volcanism, just like origins of most of the volcanisms in eastern China, has long been an enigma. A large scale of volcanic activity has dominated many places of eastern China in Meso-Cenozoic. Activity of these volcanisms in eastern China covers several quite different blocks, covers variety of tectonics, and covers a period of time over 200 million years. Such a large-scale and long-lived volcanism in a continental area challenges our knowledge on dynamics of the Earth’s interior. Some works on “Diwa” hypothesis and “lithospheric thinning” hypothesis present possible links between volcanisms and dynamic evolutions of the earth’s interior, but still cannot interpret where are the sources of the heat and fluid, which are essential to the volcanisms. Based on the study of this thesis, we suggest that dynamics of the deep subduction of western Pacific Plate is the critical factor to the Changbaishan Volcanism and volcanisms in NE China, and maybe even essential to most of the Meso-Cenozoic volcanisms in eastern China. In NE China, stagnant slabs flatted in the mantle transition zone (MTZ, ca. 660 km deep) transport and release significant hydrous fluid to the upper mantle. Metamorphism of the deep-subducted slabs and hence a series of mineral phases play an important role in the water transport, exchange, restore, and release. Dehydrated fluid of the wet slab ascending from the MTZ fertile the upper mantle, and also provide upward heat flow which is essential to the magma genesis. Then magma and volcanism occur with the deep subduction from Mesozoic to mordern time in eastern China. To discribe the exact chemical characteristics of the deep subduction releated volcanics is very difficult, because few researches has contributed to the chemical behaviors of fluid and trace elments in the very deep interior of the Earth, such 660 km deep, 410km or 350km where the fluid may ascend and react. However we can still find some chemical characteristics of oceanic subduction. Basalts of the Changbaishan Volcanism have siginficant characteristics of potassium rich, and even can be called a potassic igneous province. If there are only two possible ways, recycled continentical crust or oceanic crust, to fertile the mantle potossium element as we know now, it’s easy to attribute this to the deep-subducted of the west Pacific Plate. To the eastern China, fluid inclusions in mantle xenoliths from the Cenozic basalts also reveal potassium-rich characteristics. This reveals that the same potassium feritle agents may occur in the mantle sources of eastern China. 2. Holocene activiy of the Tianchi Volcano As one of the large volcanic center and complex volcanic cone, the Tianchi Volcano is a dangerous active volcano, with several Holocene eruptions. Among these eruptions, the Millennium Eruption is regardede as one of the biggist eruptions in the world in the last 2000 years. To estimate the potential danger of volcanic eruption, we discuss two essential factors, as (i) volcanic history of Holocene eruptions, including volcanic geolgy, chronnology and chemistry, (ii) state, evolution and relationship of the magma chambers in mantle and crust beneath the Tianchi Volcano.

On the Relationship Between Generalization Error, Hypothesis Complexity, and Sample Complexity for Radial Basis Functions

In this paper, we bound the generalization error of a class of Radial Basis Function networks, for certain well defined function learning tasks, in terms of the number of parameters and number of examples. We show that the total generalization error is partly due to the insufficient representational capacity of the network (because of its finite size) and partly due to insufficient information about the target function (because of finite number of samples). We make several observations about generalization error which are valid irrespective of the approximation scheme. Our result also sheds light on ways to choose an appropriate network architecture for a particular problem.

Hypothesis Formation and Evaluation in Medical Diagnosis

This thesis describes some aspects of a computer system for doing medical diagnosis in the specialized field of kidney disease. Because such a system faces the spectre of combinatorial explosion, this discussion concentrates on heuristics which control the number of concurrent hypotheses and efficient "compiled" representations of medical knowledge. In particular, the differential diagnosis of hematuria (blood in the urine) is discussed in detail. A protocol of a simulated doctor/patient interaction is presented and analyzed to determine the crucial structures and processes involved in the diagnosis procedure. The data structure proposed for representing medical information revolves around elementary hypotheses which are activated when certain disposing of findings, activating hypotheses, evaluating hypotheses locally and combining hypotheses globally is examined for its heuristic implications. The thesis attempts to fit the problem of medical diagnosis into the framework of other Artifcial Intelligence problems and paradigms and in particular explores the notions of pure search vs. heuristic methods, linearity and interaction, local vs. global knowledge and the structure of hypotheses within the world of kidney disease.

Functional genomic hypothesis generation and experimentation by a robot scientist

King R. D., Whelan, K. E., Jones, F. M., Reiser, P. G. K., Bryant, C. H., Muggleton, S., Kell, D. B. and Oliver, S. G. (2004) Functional genomic hypothesis generation and experimentation by a robot scientist. Nature 427 (6971) p247-252

I am a Child. Hypothesis on Spectator Pedagogy

The paper consists of series of suggestions and historical references on the basis of which it would become possible to think and practice „spectator pedagogy” in performing arts. Contemporary performance practices can claim for new kind of political relevance by focusing on the way spectator´s corporeal experience changes during and through theatrical situation. Naive body produced by a performance is also most susceptible for thoroughgoing political and ecological change. This is the first outline by its author on this topic.

Distributed Hypothesis Testing, Attention Shifts and Transmitter Dynatmics During the Self-Organization of Brain Recognition Codes

BP (89-A-1204); Defense Advanced Research Projects Agency (90-0083); National Science Foundation (IRI-90-00530); Air Force Office of Scientific Research (90-0175, 90-0128); Army Research Office (DAAL-03-88-K0088)

Agglomeration effects in foreign direct investment and the pollution haven hypothesis

Does environmental regulation impair international competitiveness of pollution-intensive industries to the extent that they relocate to countries with less stringent regulation, turning those countries into "pollution havens"? We test this hypothesis using panel data on outward foreign direct investment (FDI) flows of various industries in the German manufacturing sector and account for several econometric issues that have been ignored in previous studies. Most importantly, we demonstrate that externalities associated with FDI agglomeration can bias estimates away from finding a pollution haven effect if omitted from the analysis. We include the stock of inward FDI as a proxy for agglomeration and employ a GMM estimator to control for endogenous time-varying determinants of FDI flows. Furthermore, we propose a difference estimator based on the least polluting industry to break the possible correlation between environmental regulatory stringency and unobservable attributes of FDI recipients in the cross-section. When accounting for these issues we find robust evidence of a pollution haven effect for the chemical industry. © 2008 Springer Science+Business Media B.V.

The induced innovation hypothesis and energy-saving technological change

We develop a methodology for testing Hicks's induced innovation hypothesis by estimating a product-characteristics model of energy-using consumer durables, augmenting the hypothesis to allow for the influence of government regulations. For the products we explored, the evidence suggests that (i) the rate of overall innovation was independent of energy prices and regulations; (ii) the direction of innovation was responsive to energy price changes for some products but not for others; (iii) energy price changes induced changes in the subset of technically feasible models that were offered for sale; (iv) this responsiveness increased substantially during the period after energy-efficiency product labeling was required; and (v) nonetheless, a sizable portion of efficiency improvements were autonomous.

The reappearance hypothesis revisited: recurrent involuntary memories after traumatic events and in everyday life.

Recurrent involuntary memories are autobiographical memories that come to mind with no preceding retrieval attempt and that are subjectively experienced as being repetitive. Clinically, they are classified as a symptom of posttraumatic stress disorder. The present work is the first to systematically examine recurrent involuntary memories outside clinical settings. Study 1 examines recurrent involuntary memories among survivors of the tsunami catastrophe in Southeast Asia in 2004. Study 2 examines recurrent involuntary memories in a large general population. Study 3 examines whether the contents of recurrent involuntary memories recorded in a diary study are duplicates of, or differ from, one another. We show that recurrent involuntary memories are not limited to clinical populations or to emotionally negative experiences; that they typically do not come to mind in a fixed and unchangeable form; and that they show the same pattern regarding accessibility as do autobiographical memories in general. We argue that recurrent involuntary memories after traumas and in everyday life can be explained in terms of general and well-established mechanisms of autobiographical memory.