62 research outputs found
TouchPhoto: Enabling Independent Picture Taking and Understanding for Visually-Impaired Users
No full textThis paper presents TouchPhoto, which provides visualaudio- tactile assistive features to enable visually-impaired users to take and understand photographs independently. A user can take photographs under auditory guidance and record audio tags to aid later recall of the photographs' contents. For comprehension, the user can listen to audio tags embedded in a photograph while touching salient features, e.g., human faces, using an electrovibration display. We conducted two user studies with visually-impaired users, one for picture taking and the other for understanding and recall, in a two-month interval. They considered auditory assistance as very useful for taking and understanding photographs and tactile features as helpful but to a limited extent. © 2019 Copyright held by the owner/author(s).1
Arithmetic Transformers Can Length-Generalize in Both Operand Length and Count
No full textTransformers often struggle with length generalization, meaning they fail to generalize to sequences longer than those encountered during training. While arithmetic tasks are commonly used to study length generalization, certain tasks are considered notoriously difficult, e.g., multi-operand addition (requiring generalization over both the number of operands and their lengths) and multiplication (requiring generalization over both operand lengths). In this work, we achieve approximately 2–3× length generalization on both tasks, which is the first such achievement in arithmetic Transformers. We design task-specific scratchpads enabling the model to focus on a fixed number of tokens per each next-token prediction step, and apply multi-level versions of Position Coupling (Cho et al., 2024; McLeish et al., 2024) to let Transformers know the right position to attend to. On the theory side, we prove that a 1-layer Transformer using our method can solve multi-operand addition, up to operand length and operand count that are exponential in embedding dimension
Equal-Level Interaction: A Case Study for Improving User Experiences of Visually-Impaired and Sighted People in Group Activities
No full text1
Convergence and Implicit Bias of Gradient Descent on Continual Linear Classification
No full textWe study continual learning on multiple linear classification tasks by sequentially running gradient descent (GD) for a fixed budget of iterations per each given task. When all tasks are jointly linearly separable and are presented in a cyclic/random order, we show the directional convergence of the trained linear classifier to the joint (offline) max-margin solution. This is surprising because GD training on a single task is implicitly biased towards the individual max-margin solution for the task, and the direction of the joint max-margin solution can be largely different from these individual solutions. Additionally, when tasks are given in a cyclic order, we present a non-asymptotic analysis on cycle-averaged forgetting, revealing that (1) alignment between tasks is indeed closely tied to catastrophic forgetting and backward knowledge transfer and (2) the amount of forgetting vanishes to zero as the cycle repeats. Lastly, we analyze the case where the tasks are no longer jointly separable and show that the model trained in a cyclic order converges to the unique minimum of the joint loss function
TouchPhoto: Enabling Independent Picture-taking and Understanding of Photos for Visually-Impaired Users
No full text1
Fair Streaming Principal Component Analysis: Statistical and Algorithmic Viewpoint
No full textFair Principal Component Analysis (PCA) is a problem setting where we aim to perform PCA while making the resulting representation fair in that the projected distributions, conditional on the sensitive attributes, match one another. However, existing approaches to fair PCA have two main problems: theoretically, there has been no statistical foundation of fair PCA in terms of learnability; practically, limited memory prevents us from using existing approaches, as they explicitly rely on full access to the entire data. On the theoretical side, we rigorously formulate fair PCA using a new notion called probably approximately fair and optimal (PAFO) learnability. On the practical side, motivated by recent advances in streaming algorithms for addressing memory limitation, we propose a new setting called fair streaming PCA along with a memory-efficient algorithm, fair noisy power method (FNPM). We then provide its statistical guarantee in terms of PAFO-learnability, which is the first of its kind in fair PCA literature. We verify our algorithm in the CelebA dataset without any pre-processing; while the existing approaches are inapplicable due to memory limitations, by turning it into a streaming setting, we show that our algorithm performs fair PCA efficiently and effectively
Stereoselective Formal Hydroamidation of Si-Substituted Arylacetylenes with DIBAL-H and Isocyanates: Synthesis of (E)- and (Z)-alpha-Silyl-alpha,beta-unsaturated Amides
No full textAn efficient and stereoselective method for the synthesis of (E)- and (Z)-alpha-silyl-alpha,beta-unsaturated amides and its synthetic applications are presented herein. The solvent-controlled hydroaluminations of Si-substituted alkynes with DIBAL-H generate diastereomerically enriched alkenylaluminum reagents that are directly reacted with isocyanates at ambient temperature to afford alpha-silyl-alpha,beta-unsaturated amides in high yields with retained stereoselectivity. In particular, this process enables the synthesis of a broad range of (E)-alpha-silyl-alpha,beta-unsaturated amides, which are the less studied isomers. The synthetic utility of this method is highlighted by its short reaction time, ease of purification, easily accessible substrates and reagents, gram-scale synthesis, and the further transformations of C-Si bonds into C-H, C-X, and C-C bonds. © 2020 American Chemical Society.1
Perception of Electrostatic Friction Stimuli in Free Surface Exploration
No full textAssuming the use scenario of free exploration on tactile graphics for people with visual impairments, this study investigated how the users perceive electrostatic friction stimuli on contour-based graphical information. We designed and conducted two experiments with 16 participants (8 visually-impaired and 8 sighted). First, we obtained spatial gap detection thresholds between two lines rendered using the electrostatic display. Second, we investigated spatial numerosity judgement on rendered lines on the display. Results demonstrated that the visually-impaired and sighted participants had similar perceptual performance. We summarize the findings and present suggestions for tactile graphics on an electrostatic friction display.1
Oxygen activation on the interface between Pt nanoparticles and mesoporous defective TiO2 during CO oxidation
No full text© 2019 Author(s).Platinum-based heterogeneous catalysts are mostly used in various commercial chemical processes because of their high catalytic activity, influenced by the metal/oxide interaction. To design rational catalysts with high performance, it is crucial to understand the relationship between the metal-oxide interface and the reaction pathway. Here, we investigate the role of oxygen defect sites in the reaction mechanism for CO oxidation using Pt nanoparticles supported on mesoporous TiO2 catalysts with oxygen defects. We show an intrinsic correlation between the catalytic reactivity and the local properties of titania with oxygen defects (i.e., Ti3+ sites). In situ infrared spectroscopy observations of the Pt/mesoporous TiO2-x catalyst indicate that an oxygen molecule bond can be activated at the perimeter between the Pt and an oxygen vacancy in TiO2 by neighboring CO molecules on the Pt surface before CO oxidation begins. The proposed reaction pathways for O2 activation at the Pt/TiO2-x interface based on density functional theory confirm our experimental findings. We suggest that this provides valuable insight into the intrinsic origin of the metal/support interaction influenced by the presence of oxygen vacancies, which clarifies the pivotal role played by the support11sciescopu
PLASTIC: Improving Input and Label Plasticity for Sample Efficient Reinforcement Learning
No full textIn Reinforcement Learning (RL), enhancing sample efficiency is crucial, particularly in scenarios when data acquisition is costly and risky. In principle, off-policy RL algorithms can improve sample efficiency by allowing multiple updates per environment interaction. However, these multiple updates often lead the model to overfit to earlier interactions, which is referred to as the loss of plasticity. Our study investigates the underlying causes of this phenomenon by dividing plasticity into two aspects. Input plasticity, which denotes the model's adaptability to changing input data, and label plasticity, which denotes the model's adaptability to evolving input-output relationships. Synthetic experiments on the CIFAR-10 dataset reveal that finding smoother minima of loss landscape enhances input plasticity, whereas refined gradient propagation improves label plasticity. Leveraging these findings, we introduce the PLASTIC algorithm, which harmoniously combines techniques to address both concerns. With minimal architectural modifications, PLASTIC achieves competitive performance on benchmarks including Atari-100k and Deepmind Control Suite. This result emphasizes the importance of preserving the model's plasticity to elevate the sample efficiency in RL. The code is available at https://github.com/dojeon-ai/plastic
- …
