Catalogue Search | MBRL
Are you sure you want to remove the book from the shelf?
{{itemTitle}}
4,372
result(s) for
"Promises"
Sort by:
Context Autoencoder for Self-supervised Representation Learning
We present a novel masked image modeling (MIM) approach, context autoencoder (CAE), for self-supervised representation pretraining. We pretrain an encoder by making predictions in the encoded representation space. The pretraining tasks include two tasks: masked representation prediction—predict the representations for the masked patches, and masked patch reconstruction—reconstruct the masked patches. The network is an encoder–regressor–decoder architecture: the encoder takes the visible patches as input; the regressor predicts the representations of the masked patches, which are expected to be aligned with the representations computed from the encoder, using the representations of visible patches and the positions of visible and masked patches; the decoder reconstructs the masked patches from the predicted encoded representations. The CAE design encourages the separation of learning the encoder (representation) from completing the pertaining tasks: masked representation prediction and masked patch reconstruction tasks, and making predictions in the encoded representation space empirically shows the benefit to representation learning. We demonstrate the effectiveness of our CAE through superior transfer performance in downstream tasks: semantic segmentation, object detection and instance segmentation, and classification. The code will be available at https://github.com/Atten4Vis/CAE.
Journal Article
Improving Research-Based Knowledge of College Promise Programs
Also known as \"free tuition\" and \"free college\" programs, college promise programs are an emerging approach for increasing higher education attainment of people in particular places. To maximize the effectiveness of their efforts and investments, program leaders and policymakers need research-based evidence to inform program design, implementation, and evaluation. With the goal of addressing this knowledge need, this volume presents a collection of research studies that examine several categories and variations of college promise programs. These theoretically grounded empirical investigations use varied data sources and analytic techniques to examine the effects of college promise programs that have different design features and operate in different places. Individually and collectively, the results of these studies have implications for the design and implementation of promise programs if these programs are to create meaningful improvements in attainment for people from underserved groups. The authors' efforts also provide a useful foundation for the next generation of college promise research.
eBook
The bioeconomy and its untenable growth promises: reality checks from research
This paper starts out from the observation that recent official bioeconomy strategies and policy concepts are markedly more moderate in their promises of economic growth compared to the high-flying expectations of a ‘biotech revolution’ promoted around the turn of the millennium. We argue that this stepwise process of moderation is partly due to a series of ‘reality checks’ to which various strands of research on the bioeconomy have (willingly or unwillingly) subjected these promises, forcing governments to move away from visions exposed as unrealistic and to adopt more humble ones. We identify four such ‘reality checks’, originating from research on (a) bioeconomy discourses and knowledges, (b) contestation and power dynamics among actors and competing interests in bioeconomy politics and policymaking, as well as on (c) the economic and (d) biophysical dimensions of existing bio-based economies. In conclusion, we argue that bioeconomy research should adopt a broader perspective that considers transitions toward bio-based processes and resources as but one element in a comprehensive social–ecological transformation of current modes of production and living, and that understanding the dynamics of societal conflict around that transformation is crucial for assessing the social possibility of bioeconomy visions.
Journal Article
CLIP-guided Prototype Modulating for Few-shot Action Recognition
Learning from large-scale contrastive language-image pre-training like CLIP has shown remarkable success in a wide range of downstream tasks recently, but it is still under-explored on the challenging few-shot action recognition (FSAR) task. In this work, we aim to transfer the powerful multimodal knowledge of CLIP to alleviate the inaccurate prototype estimation issue due to data scarcity, which is a critical problem in low-shot regimes. To this end, we present a CLIP-guided prototype modulating framework called CLIP-FSAR, which consists of two key components: a video-text contrastive objective and a prototype modulation. Specifically, the former bridges the task discrepancy between CLIP and the few-shot video task by contrasting videos and corresponding class text descriptions. The latter leverages the transferable textual concepts from CLIP to adaptively refine visual prototypes with a temporal Transformer. By this means, CLIP-FSAR can take full advantage of the rich semantic priors in CLIP to obtain reliable prototypes and achieve accurate few-shot classification. Extensive experiments on five commonly used benchmarks demonstrate the effectiveness of our proposed method, and CLIP-FSAR significantly outperforms existing state-of-the-art methods under various settings. The source code and models are publicly available at https://github.com/alibaba-mmai-research/CLIP-FSAR.
Journal Article
The lion and the mouse : narrated by the timid but truthful mouse
In this humorous retelling of the classic fable told by a mouse, Catnip is the bold mouse who is caught by the lion (she really should not have tried to jump over him)--but it is her timid twin sister Bitsy who finds the courage to fulfill her sister's promise, and chews through the net the lion gets caught in.
Book
Backing Out or Backing In? Commitment and Consistency in Audience Costs Theory
Audience costs theory posits that domestic publics punish leaders for making an external threat and then backing down. One key mechanism driving this punishment involves the value the public places on consistency between their leaders' statements and actions. If true, this mechanism should operate not only when leaders fail to implement threats, but also when they fail to honor promises to stay out of a conflict. We use a survey experiment to examine domestic responses to the president's decision to \"back down\" from public threats and \"back into\" foreign conflicts. We find the president loses support in both cases, but suffers more for \"backing out\" than \"backing in. \" These differential consequences are partially explained by asymmetries in the public's treatment of new information. Our findings strongly suggest that concerns over consistency undergird audience costs theory and that punishment for inconsistency will be incurred, regardless of the leader's initial policy course.
Journal Article