Vision transformers use [CLS] token to predict image classes. Their explainability visualization has been studied using relevant information from the [CLS] token or focusing on attention scores during self-attention. However, such visualization is challenging because of the dependence of the interpretability of a vision transformer on skip connections and attention operators, the instability of non-linearities in the learning process, and the limited reflection of self-attention scores on relevance. We argue that the output patch embeddings in a vision transformer preserve the image information of each patch location, which can facilitate the prediction of an image class. In this paper, we propose ICEv2 (ICEv2: $${{{\underline{\varvec{I}}}}}$$ nterpretability, $${{{\underline{\varvec{C}}}}}$$ omprehensiveness, and $${{{\underline{\varvec{E}}}}}$$ xplainability in Vision Transformer), an explainability visualization method that addresses the limitations of ICE (i.e., high dependence of hyperparameters on performance and the inability to preserve the model’s properties) by minimizing the number of training encoder layers, redesigning the MLP layer, and optimizing hyperparameters along with various model size. Overall, ICEv2 shows higher efficiency, performance, robustness, and scalability than ICE. On the ImageNet-Segmentation dataset, ICEv2 outperformed all explainability visualization methods in all cases depending on the model size. On the Pascal VOC dataset, ICEv2 outperformed both self-supervised and supervised methods on Jaccard similarity. In the unsupervised single object discovery, where untrained classes are present in the images, ICEv2 effectively distinguished between foreground and background, showing performance comparable to the previous state-of-the-art. Lastly, ICEv2 can be trained with significantly lower training computational complexity.

Large language model (LLM)-based AI for code model (e.g., Copilot) demonstrates the potential of using AI in specialized domains such as software engineering. While previous research has focused on fine-tuning models with additional data and computational cost to construct models optimized for specific domains, our research focuses on prompt engineering methods that maximize the performance of existing models. We conducted a quantitative and qualitative user study using the AI for code model and identified two limitations that hinder the recommendation performance of the model. We propose two methods to address these limitations through effective prompt engineering. Finally, we identified the potential for the use of our proposed methods to be utilized and discussed the direction of future research for the effective use of the LLM.

Vision transformers use [CLS] tokens to predict image classes. Their explainability visualization has been studied using relevant information from [CLS] tokens or focusing on attention scores during self-attention. Such visualization, however, is challenging because of the dependence of the structure of a vision transformer on skip connections and attention operators, the instability of non-linearities in the learning process, and the limited reflection of self-attention scores on relevance. We argue that the output vectors for each input patch token in a vision transformer retain the image information of each patch location, which can facilitate the prediction of an image class. In this paper, we propose ICE (Adversarial Normalization: I Can visualize Everything), a novel method that enables a model to directly predict a class for each patch in an image; thus, advancing the effective visualization of the explainability of a vision transformer. Our method distinguishes background from foreground regions by predicting background classes for patches that do not determine image classes. We used the DeiT-S model, the most representative model employed in studies, on the explainability visualization of vision transformers. On the ImageNet-Segmentation dataset, ICE outperformed all explainability visualization methods for four cases depending on the model size. We also conducted quantitative and qualitative analyses on the tasks of weakly-supervised object localization and unsupervised object discovery. On the CUB-200-2011 and PASCALVOC07/12 datasets, ICE achieved comparable performance to the state-of-the-art methods. We incorporated ICE into the encoder of DeiT-S and improved efficiency by 44.01% on the ImageNet dataset over that achieved by the original DeiT-S model. We showed performance on the accuracy and efficiency comparable to EViT, the state-of-the-art pruning model, demonstrating the effectiveness of ICE. The code is available at https://github.com/Hanyang-HCC-Lab/ICE.

One of the significant social problems emerging in modern society is mental illness, and a growing number of people are seeking psychological help. Art therapy is a technique that can alleviate psychological and emotional conflicts through creation. However, the expression of a drawing varies by individuals, and the subjective judgments made by art therapists raise the need to secure an objective assessment. In this paper, we present M2C (Multimodal classification with 2-stage Co-attention), a deep learning model that predicts stress from art therapy psychological test data. M2C employs a co-attention mechanism that combines two modalities-drawings and post-questionnaire answers-to complement the weaknesses of each, which corresponds to therapists' psychometric diagnostic processes. The results of the experiment show that M2C yielded higher performance than other state-of-the-art single- or multi-modal models, demonstrating the effectiveness of the co-attention approach that reflects the diagnosis process.

In this paper, we evaluate the voice service capacity of the TDD OFDMA system based on IEEE 802.16d/e. Voice traffic must be transmitted in real time with very small delay when we use packet transmission networks. Voice over IP (VoIP) services can be provided in this wireless packet network as Unsolicited Granted Service (UGS) or Real Time Polling Service (rtPS) class. As UGS, a fixed amount of bandwidth can be allocated to transmit the voice traffic for the duration of a call. On the other hand, as rtPS, bandwidth is allocated variably as needed to transmit the talk spurts using polling process. In this analysis, we also take into account the delay limit as one of the main Quality of Service (QoS) requirements of the voice services. The capacities of VoIP services for the classes with different delay limits are derived and compared in this paper. Simulation results show that the voice capacity is larger when we treat the VoIP service as rtPS than that as UGS.

In this paper, we study the coverage and performance improvements that can be achieved by using smart antenna systems in the CDMA2000 1X network. We present a simulation model and an evaluation method for the smart antenna systems in the CDMA systems. We propose a method that can show the impact of smart antenna systems on the E/sub C//I/sub 0/ ratio. We consider the data service users as well as the voice service users. We derive the E/sub C//I/sub 0/ and data throughput. The simulation results of the cell coverage area and radius are also presented as a function of the antenna types and beam width. The simulation results show that the performance improvements are remarkable if the smart antenna systems are used for data service users. From the simulation results we discuss the impact of smart antenna systems on the cell coverage area, the average E/sub C//I/sub 0/ ratio, data throughput and so forth.