Search

Drawing tests like the Rey Complex Figure Test (RCFT) are widely used to assess cognitive functions such as visuospatial skills and memory, making them valuable tools for detecting mild cognitive impairment (MCI). Despite their utility, existing predictive models based on these tests often suffer from limitations like small sample sizes and lack of external validation, which undermine their reliability. We developed a multi-stream deep learning framework that integrates two distinct processing streams: a multi-head self-attention based spatial stream using raw RCFT images and a scoring stream employing a previously developed automated scoring system. Our model was trained on data from 1,740 subjects in the Korean cohort and validated on an external hospital dataset of 222 subjects from Korea. The proposed multi-stream model demonstrated superior performance over baseline models (AUC = 0.872, Accuracy = 0.781) in external validation. The integration of both spatial and scoring streams enables the model to capture intricate visual details from the raw images while also incorporating structured scoring data, which together enhance its ability to detect subtle cognitive impairments. This dual approach not only improves predictive accuracy but also increases the robustness of the model, making it more reliable in diverse clinical settings. Our model has practical implications for clinical settings, where it could serve as a cost-effective tool for early MCI screening., Comment: 20 pages, 3 figures, 2 tables

We prove several new results on the Hamming weight of bounded uniform and small-bias distributions. We exhibit bounded-uniform distributions whose weight is anti-concentrated, matching existing concentration inequalities. This construction relies on a recent result in approximation theory due to Erd\'eyi (Acta Arithmetica 2016). In particular, we match the classical tail bounds, generalizing a result by Bun and Steinke (RANDOM 2015). Also, we improve on a construction by Benjamini, Gurel-Gurevich, and Peled (2012). We give a generic transformation that converts any bounded uniform distribution to a small-bias distribution that almost preserves its weight distribution. Applying this transformation in conjunction with the above results and others, we construct small-bias distributions with various weight restrictions. In particular, we match the concentration that follows from that of bounded uniformity and the generic closeness of small-bias and bounded-uniform distributions, answering a question by Bun and Steinke (RANDOM 2015). Moreover, these distributions are supported on only a constant number of Hamming weights. We further extend the anti-concentration constructions to small-bias distributions perturbed with noise, a class that has received much attention recently in derandomization. Our results imply (but are not implied by) a recent result of the authors (CCC 2024), and are based on different techniques. In particular, we prove that the standard Gaussian distribution is far from any mixture of Gaussians with bounded variance.

The goal of trace reconstruction is to reconstruct an unknown $n$-bit string $x$ given only independent random traces of $x$, where a random trace of $x$ is obtained by passing $x$ through a deletion channel. A Statistical Query (SQ) algorithm for trace reconstruction is an algorithm which can only access statistical information about the distribution of random traces of $x$ rather than individual traces themselves. Such an algorithm is said to be $\ell$-local if each of its statistical queries corresponds to an $\ell$-junta function over some block of $\ell$ consecutive bits in the trace. Since several -- but not all -- known algorithms for trace reconstruction fall under the local statistical query paradigm, it is interesting to understand the abilities and limitations of local SQ algorithms for trace reconstruction. In this paper we establish nearly-matching upper and lower bounds on local Statistical Query algorithms for both worst-case and average-case trace reconstruction. For the worst-case problem, we show that there is an $\tilde{O}(n^{1/5})$-local SQ algorithm that makes all its queries with tolerance $\tau \geq 2^{-\tilde{O}(n^{1/5})}$, and also that any $\tilde{O}(n^{1/5})$-local SQ algorithm must make some query with tolerance $\tau \leq 2^{-\tilde{\Omega}(n^{1/5})}$. For the average-case problem, we show that there is an $O(\log n)$-local SQ algorithm that makes all its queries with tolerance $\tau \geq 1/\mathrm{poly}(n)$, and also that any $O(\log n)$-local SQ algorithm must make some query with tolerance $\tau \leq 1/\mathrm{poly}(n).$, Comment: RANDOM 2024

Harmful suicide content on the Internet is a significant risk factor inducing suicidal thoughts and behaviors among vulnerable populations. Despite global efforts, existing resources are insufficient, specifically in high-risk regions like the Republic of Korea. Current research mainly focuses on understanding negative effects of such content or suicide risk in individuals, rather than on automatically detecting the harmfulness of content. To fill this gap, we introduce a harmful suicide content detection task for classifying online suicide content into five harmfulness levels. We develop a multi-modal benchmark and a task description document in collaboration with medical professionals, and leverage large language models (LLMs) to explore efficient methods for moderating such content. Our contributions include proposing a novel detection task, a multi-modal Korean benchmark with expert annotations, and suggesting strategies using LLMs to detect illegal and harmful content. Owing to the potential harm involved, we publicize our implementations and benchmark, incorporating an ethical verification process., Comment: 30 pages, 7 figures

The tetrahedron algebra $\boxtimes$ is an infinite-dimensional Lie algebra defined by generators $\{x_{ij} \mid i, j \in \{0, 1, 2, 3\}, i \neq j\}$ and some relations, including the Dolan-Grady relations. These twelve generators are called standard. We introduce a type of element in $\boxtimes$ that "looks like" a standard generator. For mutually distinct $h, i, j, k \in \{0, 1, 2, 3\}$, consider the standard generator $x_{ij}$ of $\boxtimes$. An element $\xi \in \boxtimes$ is called $x_{ij}$-like whenever both (i) $\xi$ commutes with $x_{ij}$; (ii) $\xi$ and $x_{hk}$ satisfy a Dolan-Grady relation. Pick mutually distinct $i,j,k \in \{0,1,2,3\}$. In our main result, we find an attractive basis for $\boxtimes$ with the property that every basis element is either $x_{ij}$-like or $x_{jk}$-like or $x_{ki}$-like. We discuss this basis from multiple points of view., Comment: 30 pages

In recent interactive segmentation algorithms, previous probability maps are used as network input to help predictions in the current segmentation round. However, despite the utilization of previous masks, useful information contained in the probability maps is not well propagated to the current predictions. In this paper, to overcome this limitation, we propose a novel and effective algorithm for click-based interactive image segmentation, called MFP, which attempts to make full use of probability maps. We first modulate previous probability maps to enhance their representations of user-specified objects. Then, we feed the modulated probability maps as additional input to the segmentation network. We implement the proposed MFP algorithm based on the ResNet-34, HRNet-18, and ViT-B backbones and assess the performance extensively on various datasets. It is demonstrated that MFP meaningfully outperforms the existing algorithms using identical backbones. The source codes are available at https://github.com/cwlee00/MFP., Comment: Accepted to CVPR 2024

Recent experiments have pointed to the formation of a new state of matter, the electron quadrupling condensate in Ba$_{1-x}$K$_x$Fe$_2$As$_2$ . The state spontaneously breaks time-reversal symmetry and is sandwiched between two critical points, separating it from the superconducting and normal metal states. The adjacent two critical points make acoustic effects a promising tool to study such states because of their sensitivity to symmetry-breaking transitions. We report a theory of the acoustic effects of systems with an electron quadrupling phase and new ultrasound velocity measurements of Ba$_{1-x}$K$_x$Fe$_2$As$_2$ single crystals. The presented theory for the electron quadrupling state gives the same type of singularities that are observed in experiment., Comment: 11 pages, 6 figures

Graph neural networks have been shown successful in recent years. While different GNN architectures and training systems have been developed, GNN training on large-scale real-world graphs still remains challenging. Existing distributed systems load the entire graph in memory for graph partitioning, requiring a huge memory space to process large graphs and thus hindering GNN training on such large graphs using commodity workstations. In this paper, we propose CATGNN, a cost-efficient and scalable distributed GNN training system which focuses on scaling GNN training to billion-scale or larger graphs under limited computational resources. Among other features, it takes a stream of edges as input, instead of loading the entire graph in memory, for partitioning. We also propose a novel streaming partitioning algorithm named SPRING for distributed GNN training. We verify the correctness and effectiveness of CATGNN with SPRING on 16 open datasets. In particular, we demonstrate that CATGNN can handle the largest publicly available dataset with limited memory, which would have been infeasible without increasing the memory space. SPRING also outperforms state-of-the-art partitioning algorithms significantly, with a 50% reduction in replication factor on average.

Let $\Gamma$ be a graph with diameter at least two. Then $\Gamma$ is said to be $1$-homogeneous (in the sense of Nomura) whenever for every pair of adjacent vertices $x$ and $y$ in $\Gamma$, the distance partition of the vertex set of $\Gamma$ with respect to both $x$ and $y$ is equitable, and the parameters corresponding to equitable partitions are independent of the choice of $x$ and $y$. Assume $\Gamma$ is $1$-homogeneous distance-regular with intersection number $a_1>0$ and $D\geqslant 5$. Define $b=b_1/(\theta_1+1)$, where $b_1$ is the intersection number and $\theta_1$ is the second largest eigenvalue of $\Gamma$. We show that if intersection number $c_2\geqslant 2$, then $b\geqslant 1$ and one of the following (i)--(vi) holds: (i) $\Gamma$ is a regular near $2D$-gon, (ii) $\Gamma$ is a Johnson graph $J(2D,D)$, (iii) $\Gamma$ is a halved $\ell$-cube where $\ell \in \{2D,2D+1\}$, (iv) $\Gamma$ is a folded Johnson graph $\bar{J}(4D,2D)$, (v) $\Gamma$ is a folded halved $(4D)$-cube, (vi) the valency of $\Gamma$ is bounded by a function of $b$. Using this result, we characterize $1$-homogeneous graphs with classical parameters and $a_1>0$, as well as tight distance-regular graphs., Comment: 19 pages

We present a mathematical framework for modeling two-player noncooperative games in which one player (the defender) is uncertain of the costs of the game and the second player's (the attacker's) intention but can preemptively allocate information-gathering resources to reduce this uncertainty. We obtain the defender's decisions by solving a two-stage problem. In Stage 1, the defender allocates information-gathering resources, and in Stage 2, the information-gathering resources output a signal that informs the defender about the costs of the game and the attacker's intent, and then both players play a noncooperative game. We provide a gradient-based algorithm to solve the two-stage game and apply this framework to a tower-defense game which can be interpreted as a variant of a Colonel Blotto game with smooth payoff functions and uncertainty over battlefield valuations. Finally, we analyze how optimal decisions shift with changes in information-gathering allocations and perturbations in the cost functions., Comment: https://github.com/CLeARoboticsLab/GamesVoI.jl

We grew (Hg,Re)Ba$_2$Ca$_2$Cu$_3$O$_{8+\delta}$ ((Hg,Re)1223) single crystals with good reproducibility via the single-step flux method using monoxides as raw materials. A double-sealing method using a thick-walled quartz tube and a stainless-steel container was adopted for explosion protection. The maximum crystal size was approximately 1 mm x 1 mm in the ab plane and 0.04 mm in thickness. The crystal was square-shaped, reflecting the tetragonal crystal structure of (Hg,Re)1223. Magnetic susceptibility measurements indicated a critical temperature of 130 K. The in-plane resistivity exhibited a linear temperature dependence, indicating that the sample was close to optimal doping level. The out-of-plane resistivity was also measured, and the anisotropy parameter was 250-650 at 300 K., Comment: 21 pages, 9 figures, 2 tables

In this paper, we propose a method to improve the accuracy of speech emotion recognition (SER) by using vision transformer (ViT) to attend to the correlation of frequency (y-axis) with time (x-axis) in spectrogram and transferring positional information between ViT through knowledge transfer. The proposed method has the following originality i) We use vertically segmented patches of log-Mel spectrogram to analyze the correlation of frequencies over time. This type of patch allows us to correlate the most relevant frequencies for a particular emotion with the time they were uttered. ii) We propose the use of image coordinate encoding, an absolute positional encoding suitable for ViT. By normalizing the x, y coordinates of the image to -1 to 1 and concatenating them to the image, we can effectively provide valid absolute positional information for ViT. iii) Through feature map matching, the locality and location information of the teacher network is effectively transmitted to the student network. Teacher network is a ViT that contains locality of convolutional stem and absolute position information through image coordinate encoding, and student network is a structure that lacks positional encoding in the basic ViT structure. In feature map matching stage, we train through the mean absolute error (L1 loss) to minimize the difference between the feature maps of the two networks. To validate the proposed method, three emotion datasets (SAVEE, EmoDB, and CREMA-D) consisting of speech were converted into log-Mel spectrograms for comparison experiments. The experimental results show that the proposed method significantly outperforms the state-of-the-art methods in terms of weighted accuracy while requiring significantly fewer floating point operations (FLOPs). Overall, the proposed method offers an promising solution for SER by providing improved efficiency and performance.

Classical structural-based visual localization methods offer high accuracy but face trade-offs in terms of storage, speed, and privacy. A recent innovation, keypoint scene coordinate regression (KSCR) named D2S addresses these issues by leveraging graph attention networks to enhance keypoint relationships and predict their 3D coordinates using a simple multilayer perceptron (MLP). Camera pose is then determined via PnP+RANSAC, using established 2D-3D correspondences. While KSCR achieves competitive results, rivaling state-of-the-art image-retrieval methods like HLoc across multiple benchmarks, its performance is hindered when data samples are limited due to the deep learning model's reliance on extensive data. This paper proposes a solution to this challenge by introducing a pipeline for keypoint descriptor synthesis using Neural Radiance Field (NeRF). By generating novel poses and feeding them into a trained NeRF model to create new views, our approach enhances the KSCR's generalization capabilities in data-scarce environments. The proposed system could significantly improve localization accuracy by up to 50% and cost only a fraction of time for data synthesis. Furthermore, its modular design allows for the integration of multiple NeRFs, offering a versatile and efficient solution for visual localization. The implementation is publicly available at: https://github.com/ais-lab/DescriptorSynthesis4Feat2Map.

In this paper, we propose a radar odometry structure that directly utilizes radar velocity measurements for dead reckoning while maintaining its ability to update estimations within the Kalman filter framework. Specifically, we employ the Doppler velocity obtained by a 4D Frequency Modulated Continuous Wave (FMCW) radar in conjunction with gyroscope data to calculate poses. This approach helps mitigate high drift resulting from accelerometer biases and double integration. Instead, tilt angles measured by gravitational force are utilized alongside relative distance measurements from radar scan matching for the filter's measurement update. Additionally, to further enhance the system's accuracy, we estimate and compensate for the radar velocity scale factor. The performance of the proposed method is verified through five real-world open-source datasets. The results demonstrate that our approach reduces position error by 47% and rotation error by 52% on average compared to the state-of-the-art radar-inertial fusion method in terms of absolute trajectory error., Comment: 9 pages, 5 figures, 1 table, conference

Recent advancements in visual localization and mapping have demonstrated considerable success in integrating point and line features. However, expanding the localization framework to include additional mapping components frequently results in increased demand for memory and computational resources dedicated to matching tasks. In this study, we show how a lightweight neural network can learn to represent both 3D point and line features, and exhibit leading pose accuracy by harnessing the power of multiple learned mappings. Specifically, we utilize a single transformer block to encode line features, effectively transforming them into distinctive point-like descriptors. Subsequently, we treat these point and line descriptor sets as distinct yet interconnected feature sets. Through the integration of self- and cross-attention within several graph layers, our method effectively refines each feature before regressing 3D maps using two simple MLPs. In comprehensive experiments, our indoor localization findings surpass those of Hloc and Limap across both point-based and line-assisted configurations. Moreover, in outdoor scenarios, our method secures a significant lead, marking the most considerable enhancement over state-of-the-art learning-based methodologies. The source code and demo videos of this work are publicly available at: https://thpjp.github.io/pl2map/

Existing works on long-term open-domain dialogues focus on evaluating model responses within contexts spanning no more than five chat sessions. Despite advancements in long-context large language models (LLMs) and retrieval augmented generation (RAG) techniques, their efficacy in very long-term dialogues remains unexplored. To address this research gap, we introduce a machine-human pipeline to generate high-quality, very long-term dialogues by leveraging LLM-based agent architectures and grounding their dialogues on personas and temporal event graphs. Moreover, we equip each agent with the capability of sharing and reacting to images. The generated conversations are verified and edited by human annotators for long-range consistency and grounding to the event graphs. Using this pipeline, we collect LoCoMo, a dataset of very long-term conversations, each encompassing 300 turns and 9K tokens on avg., over up to 35 sessions. Based on LoCoMo, we present a comprehensive evaluation benchmark to measure long-term memory in models, encompassing question answering, event summarization, and multi-modal dialogue generation tasks. Our experimental results indicate that LLMs exhibit challenges in understanding lengthy conversations and comprehending long-range temporal and causal dynamics within dialogues. Employing strategies like long-context LLMs or RAG can offer improvements but these models still substantially lag behind human performance., Comment: 19 pages; Project page: https://snap-research.github.io/locomo/

The burgeoning field of on-device AI communication, where devices exchange information directly through embedded foundation models, such as language models (LMs), requires robust, efficient, and generalizable communication frameworks. However, integrating these frameworks with existing wireless systems and effectively managing noise and bit errors pose significant challenges. In this work, we introduce a practical ondevice AI communication framework, integrated with physical layer (PHY) communication functions, demonstrated through its performance on a link-level simulator. Our framework incorporates end-to-end training with channel noise to enhance resilience, incorporates vector quantized variational autoencoders (VQ-VAE) for efficient and robust communication, and utilizes pre-trained encoder-decoder transformers for improved generalization capabilities. Simulations, across various communication scenarios, reveal that our framework achieves a 50% reduction in transmission size while demonstrating substantial generalization ability and noise robustness under standardized 3GPP channel models.

Automatic Speech Recognition (ASR) still face challenges when recognizing time-variant rare-phrases. Contextual biasing (CB) modules bias ASR model towards such contextually-relevant phrases. During training, a list of biasing phrases are selected from a large pool of phrases following a sampling strategy. In this work we firstly analyse different sampling strategies to provide insights into the training of CB for ASR with correlation plots between the bias embeddings among various training stages. Secondly, we introduce a neighbourhood attention (NA) that localizes self attention (SA) to the nearest neighbouring frames to further refine the CB output. The results show that this proposed approach provides on average a 25.84% relative WER improvement on LibriSpeech sets and rare-word evaluation compared to the baseline., Comment: Accepted for IEEE ASRU 2023

On-device Automatic Speech Recognition (ASR) models trained on speech data of a large population might underperform for individuals unseen during training. This is due to a domain shift between user data and the original training data, differed by user's speaking characteristics and environmental acoustic conditions. ASR personalisation is a solution that aims to exploit user data to improve model robustness. The majority of ASR personalisation methods assume labelled user data for supervision. Personalisation without any labelled data is challenging due to limited data size and poor quality of recorded audio samples. This work addresses unsupervised personalisation by developing a novel consistency based training method via pseudo-labelling. Our method achieves a relative Word Error Rate Reduction (WERR) of 17.3% on unlabelled training data and 8.1% on held-out data compared to a pre-trained model, and outperforms the current state-of-the art methods., Comment: Accepted for IEEE ASRU 2023