Affinity Posters
Tiny Papers Poster Session 4
Krystal Maughan · Thomas F Burns
Halle B
Live content is unavailable. Log in and register to view live content
Schedule
Wed 7:30 a.m. - 9:30 a.m.
|
Cognitive resilience: Unraveling the proficiency of image-captioning models to interpret masked visual content
(
Poster
#308
)
>
link
Poster Location: Halle B #308 This study explores the ability of Image Captioning (IC) models to decode masked visual content sourced from diverse datasets. Our findings reveal the IC model's capability to generate captions from masked images, closely resembling the original content. Notably, even in the presence of masks, the model adeptly crafts descriptive textual information that goes beyond what is observable in the original image-generated captions. While the decoding performance of the IC model experiences a decline with an increase in the masked region's area, the model still performs well when important regions of the image are not masked at high coverage. The source code and all experimental results will be available on Github. |
Zhicheng Du · Xie Zhaotian · Huazhang Ying · Likun Zhang · Peiwu Qin 🔗 |
Wed 7:30 a.m. - 9:30 a.m.
|
CMFPN: Context Modeling Meets Feature Pyramid Network
(
Poster
#307
)
>
link
Poster Location: Halle B #307 Feature fusion is a powerful technique that enables predictors to access a semantically rich representation of an image. Feature Pyramid Networks (FPNs) are the most widely used models for fusing features. However, the context within the FPN layers is inconsistent, leading to false predictions. This article addresses the context inconsistency in FPN and proposes CMFPN, a new design that improves feature fusion by decoupling feature aggregation from context modeling. Experimental results, based on the COCO dataset, show that CMFPN effectively resolves the context issues and enhances the Average Precision (AP) results for both object detection and instance segmentation by $2.30\%$ and $1.7\%$, respectively. |
Faroq AL-Tam · Muhammad AL-Qurishi · Thariq Khalid · Riad Souissi 🔗 |
Wed 7:30 a.m. - 9:30 a.m.
|
A Framework for Policy Evaluation Enhancement by Diffusion Models
(
Poster
#306
)
>
link
Poster Location: Halle B #306 Reinforcement learning plays an important role in various fields, and has fast development due to advancements in policy evaluation and learning methods, which enjoys advantages of large data size. However, when data are limited, directly applying evaluation methods does not necessarily result in a good policy evaluation. In this work we provide a framework to generate synthetic data with diffusion models, to enhance policy evaluation, which is supported by experiments. |
Tao Ma · Xuzhi Yang 🔗 |
Wed 7:30 a.m. - 9:30 a.m.
|
Beyond Words: A Topological Exploration of Coherence in Text Documents
(
Poster
#305
)
>
link
Poster Location: Halle B #305 Coherence serves as a pivotal metric in evaluating the quality of a text. It quantifies how well the sentences within the text are connected and how well the text is structured and organized. It plays a vital role in various downstream Natural Language Processing tasks such as text summarization, question answering and machine translation among others. In this work, we explore the use of topological data analysis (TDA) techniques on attention graphs of text documents to model coherence. TDA techniques are known to capture structural information and patterns in data, making it suitable for modeling the $\textit{structure}$ and $\textit{flow}$ of a document, i.e. coherence. We validate our approach with experiments on the GCDC dataset, achieving state-of-the-art results with a simple MLP. |
Samyak Jain · Rishi Singhal · Sriram Krishna · Yaman Singla · Rajiv Ratn Shah 🔗 |
Wed 7:30 a.m. - 9:30 a.m.
|
When Does Second-Order Optimization Speed Up Training?
(
Poster
#304
)
>
link
Poster Location: Halle B #304 While numerous second-order optimization methods have been proposed to accelerate training in deep learning, they are seldom used in practice. This is partly due to a limited understanding of the conditions under which second-order optimization outperforms first-order optimization. This study aims to identify these conditions, particularly in terms of batch size and dataset size.We find empirically that second-order optimization outperforms first-order optimization when the batch size is large and the data set size is not too large. |
Satoki Ishikawa · Rio Yokota 🔗 |
Wed 7:30 a.m. - 9:30 a.m.
|
Lost in Translation: GANs' Inability to Generate Simple Probability Distributions
(
Poster
#303
)
>
link
Poster Location: Halle B #303 Since its inception, Generative Adversarial Networks (GAN) have marked a triumph in generative modeling. Its impeccable capacity to mimic observations from unknown probability distributions has positioned it as a widely used simulation tool. In typical applications, GANs find themselves simulating data rich in semantic information such as images or text out of random noise. As such, it is reasonable to expect that large parametric models such as GANs must be able to estimate standard theoretical probability densities with ease. In this paper, based on a series of disillusioning experimental findings, we show that GANs often fail to induce the simplest of statistical transformations between distributions. For example, starting with a standard Gaussian noise, GANs with 2-deep generators are unable to perform a positional translation. Supporting theoretical tests on generated data further corroborates our rather unsettling conclusions. |
Debanjan Dutta · Anish Chakrabarty · Swagatam Das 🔗 |