Semantic segmentation via highly fused convolutional network with multiple soft cost functions

Semantic image segmentation is one of the most challenging tasks in computer vision. In this paper, we propose a highly fused convolutional network, which consists of three parts: downsampling, fused upsampling and multiple predictions. We adopt a VGG-net based downsampling structure, followed by multiple steps of upsampling. Feature maps in each pair of corresponding pooling layers and unpooling layers are combined. We also bring out multiple pre-outputs, each is generated from an unpooling layer by a one-step upsampling operation. Finally, we concatenate these pre-outputs to get the final output. As a result, our proposed network makes high use of the feature information by fusing and reusing features in low layers. In addition, when training our model, we add multiple soft cost functions on pre-outputs and the final output. In this way, we can reduce the loss reduction in backpropagation. We evaluate our model on three public segmentation datasets: CamVid, PASCAL VOC, and ADE20K. We achieve considerable segmentation performance on PASCAL VOC dataset and ADE20K dataset. Especially on CamVid dataset, we achieve state-of-the-art performance.     

The semantic pointer theory of emotion: Integrating physiology,appraisal, and construction

Emotion theory needs to explain the relationship of language and emotions, and the embodiment of emotions, by specifying the computational mechanisms underlying emotion generation in the brain. We used Chris Eliasmith’s Semantic Pointer Architecture to develop POEM, a computational model that explains numerous important phenomena concerning emotions, including how some stimuli generate immediate emotional reactions, how some emotional reactions depend on cognitive evaluations, how bodily states influence the generation of emotions, how some emotions depend on interactions between physiological inputs and cognitive appraisals, and how some emotional reactions concern syntactically complex representations. We contrast our theory with current alternatives, and discuss some possible applications to individual and social emotions.     

语义信息复杂度对二语者英语被动句加工的影响

采用英语所独有且非常重要的句法结构——被动语态作为语言材料,选取某些可以在句子中的词汇顺序基本没有变化的情况下转换为汉语句式（汉语“被”字句）的“直译型”被动句以及必须根据其意义进行句式转换方能形成相应的汉语句式（汉语主动句）的“意译型”被动句作为实验材料,探讨了在动词语义复杂条件下二语熟练度（包括晚期中等熟练者和高熟练者）对被动句加工过程的影响。结果表明：高熟练者正确率均高于中等熟练者,说明在动词语义复杂的情况下熟练度依然可以预测英语被动句实时加工的程度和水平。此时,语义违例的正确率最高、反应时最短,句法违例句则与之相反。双违例引发的N400效应显著,正确句和句法违例均未引发负向的N400;高熟练者P600的波幅明显高于中等熟练者,句法违例和双违例引发的P600效应最显著。行为指标支持语言间句法加工相似性效应——直译句反应快、正确率高。ERP数据则得出了与之相反的结果：直译句比意译句诱发出更大的N400效应,意译句则引发了较大的P600。上述结果进一步说明二语熟练度对语言间句法加工相似性效应具有很强的调节作用。     

形旁和声旁在形声字识别中的作用

形声字是由表示意义范畴的义符(形旁)和表示发音信息的声符(声旁)组合而成, 以往研究虽然肯定了声旁在形声字加工中的作用, 但对形旁的作用仍存在一定争议。结合行为和脑电技术, 采用汉字判断任务, 以形旁和声旁均为独体字的形声字作为实验材料, 通过操纵形旁频率和声旁频率, 本研究进一步探讨了形旁和声旁在形声字识别中的作用。行为结果(实验1)发现形旁频率并未引发反应时和错误率上的显著变化, 而声旁频率在反应时和错误率上均产生了显著的效应, 相对于低频声旁, 高频声旁条件下的反应时更长、错误率更高。脑电结果(实验2)发现, 相对于低频形旁, 高频形旁在前脑区引发了一个波幅更小的N400; 而声旁频率不仅可以在全部脑区引发N400的变化, 同时在左脑的前、中脑区引发了P200的变化。相对于低频声旁, 高频声旁所引发的P200波幅更小, N400波幅更大。两实验结果说明, 在形声字识别中, 形旁和声旁均可以产生作用, 但形旁的激活时间要晚于声旁, 且作用相对较弱。     

HOW RADICALLY CAN GOD BE RECONCEIVED BEFORE CEASING TO BE GOD? THE FOUR FACES OF PANENTHEISM

Panentheism has often been put forward as a means for bringing theology and science into dialogue, perhaps even resolving some of the major tensions between them. A variety of “faces” of panentheism are distinguished, including conservative, metaphysical, apophatic, and naturalist panentheisms. This series of increasingly radical panentheisms is explored, each one bringing its own core commitments, and each describing very different relationships between religion and science. We consider, for example, the diverse ways that the radical panentheisms construe emergent phenomena in the natural world. In the end, comparing the increasingly radical forms of panentheism yields a new understanding of the state of the religion/science dialogue today.     

Do individuals with autism process words in context? Evidence from language-mediated eye-movements

It is widely argued that people with autism have difficulty processing ambiguous linguistic information in context. To investigate this claim, we recorded the eye-movements of 24 adolescents with autism spectrum disorder and 24 language-matched peers as they monitored spoken sentences for words corresponding to objects on a computer display. Following a target word, participants looked more at a competitor object sharing the same onset than at phonologically unrelated objects. This effect was, however, mediated by the sentence context such that participants looked less at the phonological competitor if it was semantically incongruous with the preceding verb. Contrary to predictions, the two groups evidenced similar effects of context on eye-movements. Instead, across both groups, the effect of sentence context was reduced in individuals with relatively poor language skills. Implications for the weak central coherence account of autism are discussed.     

How young children judge whether a word is one they know: a dual criterion account

Word familiarity judgment may be important for word learning, yet little is known about how children make this judgment. We hypothesized that preschool-age children differ in the judgment criteria that they use and that this difference derives from individual differences in basic memory processes. Those who have superior phonological working memory, but who retrieve less semantic information than their peers, base the judgment on whether they recognize a word’s sound form. Those who show the opposite memory profile base the judgment on whether they retrieve a word’s meaning. The results of two studies of 3- and 4-year-olds were consistent with these claims. Among those performing poorly on one memory measure, judgment accuracy was directly related to performance on the other memory measure. These memory-judgment relations were also found to be highly specific. This is the first investigation to demonstrate the usefulness of an individual differences approach for identifying relations between linguistic judgment processes and basic memory processes during early childhood.     

The Role of Semantic Features in Verb Processing

The present study examined the general hypothesis that, as for nouns, stable representations of semantic knowledge relative to situations expressed by verbs are available and accessible in long term memory in normal people. Regular associations between verbs and past tenses in French adults allowed to abstract two superordinate semantic features in the representation of verb meaning: durativity and resultativity. A pilot study was designed to select appropriate items according to these features: durative, non-resultative verbs and non-durative, resultative verbs. An experimental study was then conducted to assess semantic priming in French adults with two visual semantic-decision tasks at a 200- and 100-ms SOA. In the durativity decision task, participants had to decide if the target referred to a durable or non-durable situation. In the resultativity decision task, they had to decide if it referred to a situation with a directly observable outcome or without any clear external outcome. Targets were preceded by similar, opposite, and neutral primes. Results showed that semantic priming can tap verb meaning at a 200- and 100-ms SOA, with the restriction that only the positive value of each feature benefited from priming, that is the durative and resultative values. Moreover, processing of durativity and resultativity is far from comparable since facilitation was shown on the former with similar and opposite priming, whereas it was shown on the latter only with similar priming. Overall, these findings support Le Ny’s (in: Saint-Dizier, Viegas (eds) Computational lexical semantics, 1995; Cahier de Recherche Linguistique LanDisCo 12:85–100, 1998; Comment l’esprit produit du sens, 2005) general hypothesis that classificatory properties of verbs could be interpreted as semantic features and the view that semantic priming can tap verb meaning, as noun meaning.     

SOME CORRELATIONS BETWEEN METHODS OF KNOWING AND THEOLOGICAL CONCEPTS IN ARTHUR PEACOCKE'S PERSONALISTIC PANENTHEISM AND NONPERSONAL NATURALISTIC THEISM

Differences in methods of knowing correlate with differences in concepts about what is known. This is an underlying issue in science and religion. It is seen, first, in Arthur Peacocke's reasoning about God as transcendent and personal, is based on an assumption of correlative thinking that like causes like. This contrasts with a notion of causation in empirical science, which explains the emergence of new phenomena as originating from temporally prior phenomena quite unlike that which emerges. The scientific understanding of causation is compatible with a naturalistic theism that holds a nonpersonal model of God as the creative process. However, focusing on the immanence of God, there is a second correlation between methods of knowing and concepts of God. Classical empiricism, used by science, correlates with God understood nonpersonally as the creative process. Radical empiricism, in which feelings and not only sense perceptions have cognitive import, opens up the possibility that one can experience Peacocke's personal, panentheistic God as pattern‐forming influence. I illustrate this second method‐concept correlation with a personal experience.     

The development and effectiveness of memory strategies in kindergarten and elementary school: Findings from the Würzburg and Göttingen longitudinal memory studies

This article reports findings of the Würzburg and Göttingen Longitudinal Memory Studies, which focused on children's verbal memory development. The studies started with 102 (German) kindergarten children in Würzburg and 86 second-graders in Göttingen who were tested on various memory measures, including sort-recall, memory capacity, metamemory, and verbal IQ. Assessments were repeated nine times, with adjacent measurement points separated by 6-month time intervals. The main goals of the studies concerned the assessment of utilization deficiencies during the course of strategy development, the analysis of factors influencing the effectiveness of strategy use, and the occurrence of multiple strategy use. Findings confirmed the outcome of previous longitudinal studies in that semantic organization strategies were not always accompanied by superior recall, but utilization deficiency problems were comparably rare. Strategy effectiveness was shown as being affected mainly by memory capacity and by children's tendency to engage in multiple strategy use. However, children did not start to apply multiple strategies effectively until the end of Grade 2.