Attention please

Temporal DifferenceOn-policy 와 Off-policy에 대해 들어가기 전 TD(Temporal Difference)에 대해 다시 한번 짚고 넘어가보도록 하겠습니다. $$Q(s_t, a_t) \leftarrow Q(s_t, a_t) + \alpha \left( R_t + \gamma Q(s_{t+1}, a_{t+1}) - Q(s_t, a_t) \right)$$ 위 수식은 TD를 보여주고 있으며, $\alpha$는 학습률을 나타내며, TD의 가장 핵심이 되는 $ Q(s_{t+1}, a_{t+1}) $ 는 다음 상태-행동의 예측 value값인 TD target이 됩니다. $$Q(s_t, a_t) \leftarrow Q(s_t, a_t) + \alpha \left( \underbrace..

이번에 리뷰할 논문은 Mind with Eyes: from Language Reasoning to Multimodal Reasoning 입니다.https://arxiv.org/abs/2503.18071 Mind with Eyes: from Language Reasoning to Multimodal ReasoningLanguage models have recently advanced into the realm of reasoning, yet it is through multimodal reasoning that we can fully unlock the potential to achieve more comprehensive, human-like cognitive capabilities. This surve..

이번에 리뷰할 논문은 VadCLIP: Adapting Vision-Language Models for Weakly Supervised Video Anomaly Detection 입니다. https://arxiv.org/abs/2308.11681 VadCLIP: Adapting Vision-Language Models for Weakly Supervised Video Anomaly DetectionThe recent contrastive language-image pre-training (CLIP) model has shown great success in a wide range of image-level tasks, revealing remarkable ability for learning powerfu..

이번에 리뷰할 논문은 Imagine while Reasoning in Space: Multimodal Visualization-of-Thought 입니다. https://arxiv.org/abs/2501.07542 Imagine while Reasoning in Space: Multimodal Visualization-of-ThoughtChain-of-Thought (CoT) prompting has proven highly effective for enhancing complex reasoning in Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs). Yet, it struggles in complex spatial r..

이번에 리뷰할 논문은 SAMScore: A Semantic Structural Similarity Metric for Image Translation Evaluation 입니다. https://paperswithcode.com/paper/samscore-a-semantic-structural-similarity/review/ Papers with Code - Paper tables with annotated results for SAMScore: A Semantic Structural Similarity Metric for Image Translati Paper tables with annotated results for SAMScore: A Semantic Structural Similarity Met..

다양한 convolution 기법들 original convolution dilated convolution separable & depthwise & pointwise convolution depthwise separable convolution 이번 글에서는 Transposed Convolution에 대해 알아보자. 이 convolution 기법은 Deconvolution 이라는 이름으로 많이 쓰이지만 이는 잘못된 표현이다. Deconvolution은 이전의 convolution 작업을 되돌려 output을 원본 input으로 되돌리는 것이다. Transposed Convolution 역시 같은 spatial dimension을 만들어낸다는 점에서 Deconvolution과 유사하지만 Transpos..