Republic of Developer Agenda

--- tags: Republic of Developer --- Republic of Developer Agenda == 如有任何疑問，請寄信連繫我們：info@republicofdeveloper.dev ## 2021年主要研討議題 - [One-Shot/Few-Shot Learning](#One-ShotFew-Shot-Learning) - [Theory of Generalization and Optimization in Deep Learning (2021)](#Theory-of-Generalization-and-Optimization-in-Deep-Learning (2021)) - [Neural-Symbolic AI](#Neural-Symbolic-AI) - [Domain Adaption](#Domain-Adaption) - [Inverse Reinforcement Learning](#Inverse-Reinforcement-Learning) - [Knowledge Graph and its Applications for NLP](#Knowledge-Graph-and-its-Applications-for-NLP) ## 經典論文導讀（2019） | 日期 | 主題 | 類別 | 講者 | 資源 | | ---------- | --------------------------------- | --------------------- | ---------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | 2019.01.04 | LeNet - Part I | CNN | Kevin | [論文](http://yann.lecun.com/exdb/publis/pdf/lecun-98.pdf), [demo](http://yann.lecun.com/exdb/lenet/index.html), [講義](https://drive.google.com/file/d/1Jb_hnUzlusNbCNtoqMC3eMMDsCv17SUS/view?usp=sharing) | | 2019.01.11 | LeNet - Part II: Back Propagation | CNN | Kevin | [論文](http://yann.lecun.com/exdb/publis/pdf/lecun-98.pdf), [demo](http://yann.lecun.com/exdb/lenet/index.html), [講義](https://drive.google.com/file/d/1VijV3xarXe3Gts-tW4ouyqqLgFEraVUs/view?usp=sharing), [錄影](https://drive.google.com/file/d/1SyfB1hxQchjZzPQGl70ST1OmUKOOl6y-/view?usp=sharing) | | 2019.01.18 | LeNet - Part III | CNN | Kevin | [論文](http://yann.lecun.com/exdb/publis/pdf/lecun-98.pdf), [demo](http://yann.lecun.com/exdb/lenet/index.html), [講義](https://drive.google.com/file/d/1cgXbzt9wPS_NWhi0dELdayOSdoXRo_cd/view?usp=sharing) | | 2019.01.25 | AlexNet | CNN | 林俊宇 | [論文](http://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf), [講義](https://drive.google.com/file/d/1bAvEenRF0Cehmrzd2r28GBpY_K2r0I1F/view?usp=sharing), [錄影](https://drive.google.com/file/d/1iI59IRW1tdXjvmuqqqbN5MQcXKvDzH_j/view?usp=sharing) | | 2019.02.15 | ZFNet | CNN | 蘇嘉冠 | [論文](https://arxiv.org/pdf/1311.2901.pdf), [講義](https://drive.google.com/open?id=1CawxJn_US0-eW1DYkP3F7bo21_IFFmUu-Bkvtnvo0Zo), [錄影](https://drive.google.com/open?id=1ogxiuDBqzEBYeBo4NHHfVZye1Bt95qSp) | | 2019.02.22 | Network in Network | CNN | 君諦 | [論文](https://arxiv.org/pdf/1312.4400.pdf), [講義](https://drive.google.com/open?id=1np00ohZAi4q-NAOeM4P_CSozxh_0d2GB), [錄影](https://drive.google.com/file/d/1Q_ortO7ACVmXVAg38Ykh02J2AOjf7gIq/view?usp=sharing) | | 2019.03.08 | GoogLeNet | CNN | Wesley | [論文](http://openaccess.thecvf.com/content_cvpr_2015/papers/Szegedy_Going_Deeper_With_2015_CVPR_paper.pdf), [講義](https://drive.google.com/file/d/1ttXsCFcQSh5qUkHocZGqjE_ttNC7m40r/view?usp=sharing), [錄影](https://drive.google.com/open?id=1wsMjmbj1CRYT_j9nrf7IS9mQ_Z-r3Vb8) | | 2019.03.22 | VGGNet | CNN | Kevin | [論文](https://arxiv.org/pdf/1409.1556/), [講義](https://drive.google.com/file/d/1EyicnZAVclLjHXGHNfbsL-Y2KbW07z-O/view?usp=sharing), [錄影](https://drive.google.com/file/d/1cRxfP677L3eNKzTnIL1y0nZvROvdydqz/view?usp=sharing) | | 2019.03.29 | ResNet | CNN | 君諦 | [論文](https://arxiv.org/pdf/1512.03385.pdf), [講義](https://drive.google.com/open?id=1RRno3KxkY-5-glyOUYVwxC87BXO_OOjR), [錄影](https://drive.google.com/file/d/1oLGCldWthNgSQvF0HdPNs56y9mITFonR/view?usp=sharing) | | 2019.04.12 | SqueezeNet | CNN | Kevin | [論文](https://arxiv.org/pdf/1602.07360.pdf), [code](https://github.com/DeepScale/SqueezeNet), [錄影](https://drive.google.com/open?id=1Bi8t_yzxzFPjB0t7DqanptqPShihDCwe) | | 2019.04.19 | 實作經驗討論 | 其他 | | | | 2019.04.26 | RNN / LSTM | NLP | Erik Apostol | [textbook](http://www.deeplearningbook.org/contents/rnn.html), [論文](http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.676.4320&rep=rep1&type=pdf), [code](https://github.com/pecu/PyTorch_CSX/tree/master/04_RNN), [講義](https://drive.google.com/open?id=1Uij38tSkqX-La1IgtZmg6vy2cFk2vDYt), [錄影](https://drive.google.com/open?id=1T2RQF_tkEK0lAVX3mWpkI-0qyf3gIxYr) | | 2019.05.03 | NLP Basic | NLP | 君諦 | [code](https://github.com/pecu/PyTorch_CSX/tree/master/06_Natural_Language_Processing), [講義](https://drive.google.com/open?id=1PCPV21ZOwLjn3vXYfYXyteSk6XWepSJc), [錄影](https://drive.google.com/open?id=1SgYQYPpqHsIjTJ-T58CyURzmCnDdoeNy) | | 2019.05.10 | Seq2Seq | NLP | Jerry | [論文](http://papers.nips.cc/paper/5346-sequence-to-sequence-learning-with-neural-networks.pdf), [code](https://github.com/d06521005/NLP_Competition/blob/master/basic_seq2seq.ipynb), [講義](https://drive.google.com/open?id=1iof7fBmE6C77_GBuoMdluMXZhhrrRtXE) | | 2019.05.17 | Seq2Seq with attention | NLP | 陳明達 | [論文](https://arxiv.org/pdf/1409.0473.pdf), [code](https://github.com/pecu/PyTorch_CSX/tree/master/09_Attention_seq2seq), [講義](https://drive.google.com/file/d/1msFVb4yXoyrrMpNPYjRj1-W92h5MQBN8/view?usp=sharing) | | 2019.05.24 | FCN | Semantic Segmentation | 蘇嘉冠 | [論文](https://www.cv-foundation.org/openaccess/content_cvpr_2015/papers/Long_Fully_Convolutional_Networks_2015_CVPR_paper.pdf), [講義](https://drive.google.com/open?id=1MA6zWYC-kbVTkraUM5yK7O8GkbHmzTPnu830J2nDOdo), [錄影](https://drive.google.com/open?id=1OP28hwyCUmP6JBqFSz-uh0c8oqqjr1LH) | | 2019.05.31 | YOLOv1 | Object Detection | Isa, Elisa Chang | [論文](https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Redmon_You_Only_Look_CVPR_2016_paper.pdf), [講義(part I)](https://drive.google.com/file/d/1S1TDUSVaMYYXIt4Bm7VX807KBBLpHpET/view?usp=sharing), [講義(part II)](https://drive.google.com/file/d/1zZiyM3P5QTya65HsVEohnnoXWw3oiSGj/view?usp=sharing)), [錄影(part I)](https://drive.google.com/file/d/12aUup5Vpxht4cr5Dv4QDObV1Fdxh7lid/view?usp=sharing), [錄影(part II)](https://drive.google.com/open?id=1F-PNOe5Q6PMVJSjXGyhE2A7gnzKi1JoB), [錄影(part III)](https://drive.google.com/open?id=16eFpoA6v0Gs_XFWtZpCwSa-KTT7ZPAcJ) | | 2019.06.14 | SSD | Object Detection | 博儒 | [論文](https://arxiv.org/pdf/1512.02325.pdf), [講義](https://drive.google.com/open?id=1htWNky8dpUux-kZFa6pyeJXVfKPLTfxI), [錄影](https://drive.google.com/open?id=1FfbKsPhsTp6SjFV2NttYqh1cMDWpPbTf) | | 2019.07.05 | YOLOv2 | Object Detection | Jerry | [論文](http://openaccess.thecvf.com/content_cvpr_2017/papers/Redmon_YOLO9000_Better_Faster_CVPR_2017_paper.pdf), [講義](https://docs.google.com/presentation/d/1-0K6p82mehzYxDLffXU5BHMezePJRZg8zh9Hgcfptu0/edit?usp=drivesdk), [錄影](https://drive.google.com/open?id=1O_RxYoiqi1DBXLUOKgUdjxFqlyHhItDY) | | 2019.06.28 | FPN | Object Detection | Jason Lee | [論文](http://openaccess.thecvf.com/content_cvpr_2017/papers/Lin_Feature_Pyramid_Networks_CVPR_2017_paper.pdf), [講義](https://drive.google.com/file/d/1-jUNzEBqef07mWYWF36i65TA8dEP37z4/view?usp=sharing), [錄影](https://drive.google.com/open?id=1F7LBc3h8mFvyZnmmtZ1DDmuaLik4mb0h) | | 2019.07.26 | RetinaNet | Object Detection | 坤賢 | [論文](https://arxiv.org/pdf/1708.02002.pdf), [講義](https://docs.google.com/presentation/d/1rWb8PFjtUALMkFQmNhnwJf1eeFhgcbO7lNw74AszJBI/edit?usp=sharing), [錄影](https://drive.google.com/open?id=1-BCpN3Qa1ZdNKk1kz59cIzPqz2jMA7Jb) | | 2019.08.16 | YOLOv3 | Object Detection | Jerry | [論文](https://pjreddie.com/media/files/papers/YOLOv3.pdf), [講義](https://drive.google.com/open?id=1HN_b7ZqVThDLxG44Os8LZwtSrdF_gYTTmZ1JKnRrl1U), [錄音](https://drive.google.com/open?id=1TbdSJPiMAT0oM2cMEf9ILGrZxB6E9-EO) | | 2019.08.23 | Faster R-CNN | Object Detection | CMIND | [論文](http://papers.nips.cc/paper/5638-faster-r-cnn-towards-real-time-object-detection-with-region-proposal-networks.pdf), [講義](https://drive.google.com/open?id=1SC33WlIT264Ry8cJmqeFyqNZDGMJeQCUAA21a9JOkGM) | | 2019.09.27 | Dropout | Regularization | 蘇嘉冠 | [論文](http://www.jmlr.org/papers/volume15/srivastava14a/srivastava14a.pdf), [講義](https://drive.google.com/open?id=1cOB7b42aA6y1qUODm7IOJVp16-Ks4bbn231ADJ_40eU), [錄影](https://drive.google.com/open?id=1ohiJcFKAsM_E0nchyp8xv7eZQobvkriF) | | 2019.10.25 | Batch Normalization | Normalization | 小七 | [論文](http://proceedings.mlr.press/v37/ioffe15.pdf), [講義](https://drive.google.com/open?id=1UIE-72AF2dtmcymZ8QOjMypn_2obb3pB), [錄影](https://drive.google.com/file/d/1wNpk1oV4Qq8UGl2Etz3_rFOMOQl-clB1/view?usp=sharing) | | 2019.10.25 | Layer Normalization | Normalization | 博儒 | [論文](https://arxiv.org/pdf/1607.06450.pdf), [講義](https://drive.google.com/open?id=1T_-mnex7PPk6l6jgq0H1KrBdk_AxSV4p), [錄影](https://drive.google.com/file/d/1IIaOIVtRtcAmgSI3RIPYxRkS0FJxoay6/view?usp=sharing) | | 2019.11.01 | Momentum | Optimization | 郭瑞申 | [論文](http://proceedings.mlr.press/v28/sutskever13.pdf), 講義, [錄影(PART I)](https://drive.google.com/open?id=1j4nc6I37V9IB4vA1aRGvqja0CcPx3i3R), [錄影(PART II)](https://drive.google.com/open?id=1BWbyqHR9ej-1M4BXz0TiWyR3S-nAMWpt) | | 2019.11.08 | Adam | Optimization | Bill | [論文](https://arxiv.org/pdf/1412.6980.pdf), [講義](https://drive.google.com/open?id=1wjeWjXry_6h2ZloVHOODilvf48rnpNVj), [錄影(PART I)](https://drive.google.com/open?id=1_DVovjOj_cftxpP-Ab0kog7wG-PNrjiw), [錄影(PART II)](https://drive.google.com/open?id=1ATaRJRiHC4GqCzr3JH9wqsxCn7uaK8hr) | | 2019.11.29 | ConvS2S | NLP | Yan | [論文](https://arxiv.org/pdf/1705.03122.pdf), [講義](https://docs.google.com/presentation/d/1qgB5rPSXfUnnBbq0cG54lWvnVemKci4Fnj3JWAyDcdg/edit?usp=sharing), 錄影 ([1](https://drive.google.com/open?id=1c1yIzFaxfmMcFa10dioaGyIR3quTKWkX), [2](https://drive.google.com/open?id=1dbLQtq0oQmw84FTqyffEYk4AQ8nADzIS)) | | 2019.12.13 | Transformer | NLP | 坤賢 | [論文](https://papers.nips.cc/paper/7181-attention-is-all-you-need.pdf), [講義](https://drive.google.com/open?id=1GIBjZzaQ-YnWfyok4TxgqvZX6elELxCbp5DaArMO988), [錄影](https://drive.google.com/open?id=1z5WgRTfdj5TkKyQNeEy8kU22QnV1XeAD) | | 2019.12.20 | BERT | NLP | 坤賢 | [論文](https://arxiv.org/pdf/1810.04805.pdf), [講義](https://drive.google.com/open?id=19ZjbmDknt9PlV0_KKKKns1tVkVsNkpRLFoxlPe4QlZQ), [錄影](https://drive.google.com/open?id=1-JEg07wpVKnhtGXuCIwVy5Usuu7r6EkU) | ## Generative Adversarial Network (2020) ### 參考資料 - [Generative Adversarial Network (GAN), 2018 by Hung-Yi Lee](https://www.youtube.com/watch?v=DQNNMiAP5lw&list=PLJV_el3uVTsMq6JEFPW35BCiOQTsoqwNw) ([Notes](https://hackmd.io/2cIDrq28THycN5W_usEESg)) - [Must-Read Papers on GANs](https://towardsdatascience.com/must-read-papers-on-gans-b665bbae3317) - [GAN paper list and review](https://spark-in.me/post/gan-paper-review) - [Advanced GANs - Exploring Normalization Techniques for GAN training: Self-Attention and Spectral Norm](https://sthalles.github.io/advanced_gans/) | 日期 | 主題 | 類別 | 講者 | 資源 | 簡介 | |-----|-----|------|-----|-----|-----| | 2020.02.21 | GAN導讀 | GAN Basics | 蘇嘉冠 | [講義](https://docs.google.com/presentation/d/1NdC8W9GUKqhB1d-ed_l34-3-VCarGn1x63vCL8Ipetw/edit#slide=id.g33c28c5f61_0_31), 錄影([1](https://drive.google.com/open?id=1tW3-hArYqDyLPRCBDnPUiXgp8c6Re9E9), [2](https://drive.google.com/open?id=1bLQ4FnR7dekeLbHLAUa9IS7Vj_GJE52w)) | GAN論文系列的導讀，入門必看! | | X | GAN (2014) | GAN Basics | X | [論文](https://arxiv.org/pdf/1406.2661), 其他同導讀 | GAN的第一篇論文，經典中的經典！必讀！ | | 2020.07.10 | DCGAN (2015) | GAN Basics | 蘇嘉冠 | [論文](https://arxiv.org/pdf/1511.06434.pdf), [講義](https://docs.google.com/presentation/d/1OorjBcYG6KhfnRLEQ-UGKBJVnBTvhQKW3FUwLWwjHN4/edit?usp=sharing), [code](https://github.com/SuJiaKuan/anime_faces_dcgan), [錄影](https://www.facebook.com/ROD455287/videos/1109211666129111) | 講述CNN如何用在GAN裡面 | | 2020.07.24 | Improved Techniques for Training GANs (2016) | GAN Basics | 坤賢 | [論文](https://arxiv.org/pdf/1606.03498.pdf), [講義](https://docs.google.com/presentation/d/1F6iGbljsUhRasFaRWcw1TP_7fCWtufbs7i8odX45Ugw/edit?usp=sharing), [錄影](https://www.facebook.com/ROD455287/videos/285565442675572) | 提出許多使GAN訓練能穩定收斂的技巧 | | 自行閱讀 | Conditional GAN (2014) | Conditional GANs | X | [論文](https://arxiv.org/pdf/1411.1784), 其他同導讀 | 讓GAN能輸入可控制的標記，根據標記來生成內容 | | 自行閱讀 | Generative Adversarial Text to Image Synthesis (2016) | Conditional GANs | X | [論文](https://arxiv.org/pdf/1605.05396.pdf) | 文字生成圖片！ | | 2020.08.07 | Pix2Pix (2016) | Conditional GANs | 勇伯 | [論文](https://arxiv.org/pdf/1611.07004.pdf), [講義](https://drive.google.com/file/d/14t9LVVbKAf-h0XYZbvfxOR0gqh_cIldh/view?usp=sharing), [論文](https://www.facebook.com/ROD455287/videos/1190192111358323) | 可訓練影像的對映關係，如灰階影像轉彩階、線條圖生成真實圖 | | 2020.08.21 | StackGAN (2016) | Conditional GANs | 博儒 | [論文](https://arxiv.org/pdf/1612.03242), [講義](https://drive.google.com/file/d/1VY6BT8XwJltybgo9UnRNCmdM9whHh2y0/view?usp=sharing), [錄影](https://www.facebook.com/ROD455287/videos/616825712307601) | 文字生成圖片，利用兩階段的GAN生成高畫質的圖片 | | 2020.08.28 | CycleGAN (2017) | Unsupervised Conditional GANs | 君諦 | [論文](https://arxiv.org/pdf/1703.10593.pdf), [講義](https://docs.google.com/presentation/d/10sSnoIHa8h1PqIU3HQMr7J4t7-MANZAaat34HSz2H5Q/edit?usp=sharing), [錄影](https://www.facebook.com/ROD455287/videos/343179913728666) | 不用成對的標記資料，便可以在讓兩個不同domain的圖片互轉 | | 2020.09.04 | StarGAN (2017) | Unsupervised Conditional GANs | 廖婉丞 | [論文](https://arxiv.org/pdf/1711.09020.pdf), [講義](https://drive.google.com/file/d/1VY6BT8XwJltybgo9UnRNCmdM9whHh2y0/view?usp=sharing), [錄影](https://www.facebook.com/ROD455287/videos/734965470680426) | 不用成對的標記資料，便可以在讓多個不同domain的圖片互轉 | | 2020.09.11 | WGAN (2017) | Advanced GANs | 羅宇昇 | [論文](https://arxiv.org/pdf/1701.07875.pdf), [錄影](https://www.facebook.com/106405837440910/videos/999156030522230) | 改良GAN的loss定義方式，用Earth Mover’s Distance (又稱Wasserstein Distance)來定義 | | 自行閱讀 | Improved WGAN / WGAN-GP (2017) | Advanced GANs | X | [論文](https://arxiv.org/pdf/1704.00028.pdf) | WGAN再改良版 | | 自行閱讀 | SN-GAN (2018) | Advanced GANs | X | [論文 ](https://arxiv.org/pdf/1802.05957.pdf) | 提出一種叫做spectral normalization的weight normalization的方法在GAN的訓練上 | | 自行閱讀 | Progressive Growing GAN (2017) | Advanced GANs | X | [論文](https://arxiv.org/pdf/1710.10196.pdf) | 從小圖生成開始訓練，再慢慢加大尺寸繼續訓練 | | 自行閱讀 | SA-GAN (2018) | Advanced GANs | X | [論文](https://arxiv.org/pdf/1805.08318.pdf), [講義](https://docs.google.com/presentation/d/1vqLqWp-xL-FbzQHtB6-mw99nt_X45XVz2dUs20R5fUg/edit?usp=sharing) | 還記得Transformer的Self-Attention嗎？把它用在GAN裡面 | | 2020.11.13 | BigGAN (2018) | Advanced GANs | 蘇嘉冠 | [論文](https://arxiv.org/pdf/1809.11096.pdf), [講義](https://docs.google.com/presentation/d/1IxCAmFC0XSS_JAt2sjE92baX2_BjT5i57o6WAhDPhus/edit?usp=sharing), [錄影](https://drive.google.com/file/d/1sSi6DFqmMzgGtDrRxzwjJiNku2zLCMeB/view?usp=sharing) | 將各種方法集結起來，例如Self-Attention, Spectral Normalization, Conditional GAN | | 2020.11.27 | StyleGAN (2019) | Advanced GANs | 郭瑞申 | [論文](https://arxiv.org/pdf/1812.04948.pdf) | 將所謂的Adaptive Instance Normalization (AdaIN)引入 | ## One-Shot/Few-Shot Learning ### 參考資料 - [從 CVPR 2019 一覽小樣本學習研究進展](https://www.chainnews.com/articles/996640556602.htm) - [CVPR19-Few-shot](https://zhuanlan.zhihu.com/p/67402889) | 日期 | 主題 | 類別 | 講者 | 資源 | 簡介 | |-----|-----|------|-----|-----|-----| | |2015-Siamese Neural Networks for One-shot Image Recognition |Wilson Ho | [論文](https://www.cs.cmu.edu/~rsalakhu/papers/oneshot1.pdf) |通過孿生網路學圖片特徵，然後在不重新學習直接複用網路輸出的特徵 | | |2016-Learning to learn by gradient descent by gradient descent |Wilson Ho | [論文](https://arxiv.org/pdf/1606.04474.pdf) | | ||2016-One-shot Learning with Memory-Augmented Neural Networks||[論文](https://arxiv.org/pdf/1605.06065.pdf)| 記憶增強神經網絡具有快速吸收新數據知識的能力，並且能利用這些吸收了的數據，在少量樣本的基礎上做出準確的預測 | ||2016-Low-shot Visual Recognition by Shrinking and Hallucinating Features||[論文](https://arxiv.org/pdf/1606.02819.pdf)|Low-shot也叫做lifelong learning，一般分為base category和novel category，其中novel category每類只有少量樣本（K-shot），希望模型在測試集上的base category和novel category都表示很好 | ||2017-Optimization as a model for few-shot learning|Wilson Ho|[論文](https://openreview.net/pdf?id=rJY0-Kcll)| | ||2017-Matching Networks for One Shot Learning|Wilson Ho|[論文](https://arxiv.org/pdf/1606.04080.pdf)| 基於小樣本去學習歸類（或者別的任務），並且這個訓練好的模型不需要經過調整，也可以用在對訓練過程中未出現過的類別進行歸類 | ||2017-One-Shot Imitation Learning||[論文](https://arxiv.org/pdf/1703.07326.pdf)|一眼模仿學習，可以說是機器人學習的一個比較終極的問題，最理想的情況就是我們人教機器人一個任務，我們稍微演示一下，機器人就能學會！一旦機器人具備這樣的模仿學習能力，機器人就具備了非常強大的通用性，也非常類似我們人類的學習過程，可以說是機器人智能的一大進步 | ||2017-Prototypical Networks for Few-shot Learning|Wilson Ho|[論文](https://arxiv.org/pdf/1703.05175v2.pdf)|對於分類問題，原型網絡將其看做在語義空間中尋找每一類的原型中心。學習一個度量函數，該度量函數可以通過少量的幾個樣本找到所屬類別在該度量空間的原型中心。測試時，用支持集中的樣本來計算新的類別的聚類中心，再利用最近鄰分類器的思路進行預測 | ||2018-Learning to Compare_Relation Network for Few-Shot Learning|Wilson Ho|[論文](https://openaccess.thecvf.com/content_cvpr_2018/papers/Sung_Learning_to_Compare_CVPR_2018_paper.pdf)| | ||2017-Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks |Wilson Ho|[論文](https://arxiv.org/pdf/1703.03400v3.pdf)|MAML| ||2018-Zero-Shot Object Detection||[論文](https://arxiv.org/pdf/1811.11507v2.pdf)| | ||2018-One-Shot Instance Segmentation||[論文](https://arxiv.org/pdf/1811.11507v2.pdf)|本論文提出SiameseMaskRCNN，給一個之前沒有見過的新的類別的物體, 然後要在一些query images中分割出這個類別的所有物體| ||2018-Dynamic Few-Shot Visual Learning without Forgetting||[論文](http://openaccess.thecvf.com/content_cvpr_2018/papers/Gidaris_Dynamic_Few-Shot_Visual_CVPR_2018_paper.pdf)|| ||2018-Meta-Transfer Learning for Few-Shot Learning||[論文](https://zpascal.net/cvpr2019/Sun_Meta-Transfer_Learning_for_Few-Shot_Learning_CVPR_2019_paper.pdf)| | ||2018-Deep Reinforcement One-Shot Learning for Artificially Intelligent Classification Systems||[論文](https://arxiv.org/pdf/1808.01527.pdf)| | ||2018-On First-Order Meta-Learning Algorithms|Wilson Ho|[論文](https://arxiv.org/pdf/1803.02999.pdf)|Reptile| ||2019-Image Deformation Meta-Networks for One-Shot Learning||[論文](https://arxiv.org/pdf/1905.11641.pdf)| | ||2019-Meta-Transfer Learning for Few-Shot Learning||[論文](https://arxiv.org/pdf/1812.02391.pdf)| | ||2019-Revisiting Local Descriptor based Image-to-Class Measure for Few-shot Learning||[論文](https://arxiv.org/pdf/1903.12290.pdf)| | ||2019-LCC: Learning to Customize and Combine Neural Networks for Few-Shot Learning||[論文](https://arxiv.org/pdf/1904.08479.pdf)|| ||2019-Finding Task-Relevant Features for Few-Shot Learning by Category Traversal||[論文](https://arxiv.org/pdf/1905.11116.pdf)| | ||2019-Generalizing from a Few Examples_A Survey on Few-Shot Learning|Wilson Ho|[論文](https://arxiv.org/pdf/1904.05046.pdf)| | ## Stanford cs231n (2019) 本系列以 [Stanford cs231n](http://cs231n.stanford.edu/2017/) 為內容參考 | 日期 | 主題 | 講者 | 資源 | |-----|-----|-----|-----| | 2019.03.29 | Lecture 1 - Course Introduction | Scott Lin | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture1.pdf), [錄影](https://drive.google.com/file/d/119mPkkzlUL8XNVthK8qmXKQ_UYCLZB06/view?usp=sharing) | | 2019.04.12 | Lecture 2 - Image Classification | Jason Lee | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture2.pdf), [錄影](https://drive.google.com/open?id=1gMNcoBMNBqKZ_nC9WrjotZrjc7mXFxcE) | | 2019.04.19 | Lecture 3 - Loss Functions and Optimization | 嘉冠 | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture3.pdf), [錄影](https://drive.google.com/open?id=1ODh7UlwK6uRayQOCUfjjH5MfDGsl2l7O) | | 2019.04.26 | Lecture 4 - Introduction to Neural Networks | Scott Lin | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture4.pdf), [錄影](https://drive.google.com/file/d/1GOb2eQLqIYjkxZvbhmLK7hiY8MDpDfPV/view?usp=sharing) | | 2019.05.03 | Lecture 5 - Convolutional Neural Networks | Jason Lee | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture5.pdf), [錄影](https://drive.google.com/open?id=1an5xf24Djkzqx2O9Tzuk-1DedRb8CYFZ) | | 2019.05.10 | Lecture 6 - Training Neural Networks, part I | 李詩欽 | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture6.pdf), [錄影](https://drive.google.com/open?id=1Lrc88fkXgueiKhX3E7-OJPX4rJLyNPMC) | | 2019.05.17 | Lecture 7 - Training Neural Networks, part II | 翁堉珊 | [講義(講者)](https://drive.google.com/open?id=1Si3Av-dPWegbZ7QT4uDredV7ppAbGAdh) [講義(官方)](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture7.pdf) | | 2019.05.24 | Lecture 8 - Deep Learning Software | 林志龍 | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture8.pdf), [錄影](https://drive.google.com/open?id=1zXNjqIJXWOzSfZWEQw3c1nIOCQnVF_oI) | | 2019.06.14 | Lecture 9 - CNN Architectures | Vincent | [講義(官方)](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture9.pdf), [錄影](https://drive.google.com/open?id=1hZk2PHaxbBG6FGrjJBEp8Uor089072Mr) | | 2019.06.21 | Lecture 10 - Recurrent Neural Networks | 蘇嘉冠 | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture10.pdf), [錄影](https://drive.google.com/open?id=1SRk8E1ftvowpSgpp85v6xHWsIb4BOkoI) | | 2019.06.28 | Lecture 11 - Detection and Segmentation | 博儒 | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture11.pdf), [錄影](https://drive.google.com/open?id=1jGxNTYY9cSZJqUnwAnIGmiied-yznNMl) | | 2019.07.05 | Lecture 12 - Visualizing and Understanding | Wilson | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture12.pdf), [錄影](https://drive.google.com/open?id=1cLLHSFaYfzksq1FhVX1THytT9GZKjop6) | | 2019.07.26 | Lecture 13 - Generative Models | Wilson | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture13.pdf), [錄影](https://drive.google.com/open?id=1t_-mIkIZn50DcPGn1mxAkSQUzvvJEfSY) | | 2019.08.02 | Lecture 14 - Deep Reinforcement Learning | 翁堉珊 | [講義(官方)](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture14.pdf), [講義(講者)](https://drive.google.com/a/g2.nctu.edu.tw/file/d/1CrCjk-v6mRuhYsIq8_tzZo0k7TtfbgsO/view?usp=drivesdk) | | 2019.08.23 | Guest Lecture - Efficient Methods and Hardware for Deep Learning | 郭瑞申 | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture15.pdf) | | 2019.09.06 | Guest Lecture - Adversarial Examples and Adversarial Training | 君諦 | [講義](http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture16.pdf) | ## Stanford cs224n (2019~2020) 本系列以 [Stanford cs224n](https://web.stanford.edu/class/archive/cs/cs224n/cs224n.1194/) 為內容參考 | 日期 | 主題 | 講者 | 資源 | |-----|-----|-----|-----| | 2019.10.18 | Lecture 1: Introduction to NLP and Deep Learning | 嘉冠 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture01-wordvecs1.pdf), [補充](https://hackmd.io/wARI4jJ9TJuKzh3m1l8cmA), [錄影](https://drive.google.com/open?id=1lplCLBAaWv9xf0iKiwSi_AFYKHIO1SIi) | | 2019.11.01 | Lecture 2: Word Vector Representations: word2vec | Yan | [講義](https://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture02-wordvecs2.pdf), [補充](https://hackmd.io/RApheMj-RuGO_vW1w10iFg), [錄影](https://drive.google.com/open?id=1DKCasRnyJdW46zxJ8z88k4H8ChywO-tf) | | 2019.11.08 | Lecture 3: Word Window Classification, Neural Networks, and Matrix Calculus | Patrick | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture03-neuralnets.pdf), [錄影(PART I)](https://drive.google.com/open?id=19FDVQz054XIK63gSVQmen52wfcHk9wtx), [錄影(PART II)](https://drive.google.com/open?id=19Kn2aesHf_igSwNXyE2uyRPJDRDMW9au) | | 自行閱讀 | Lecture 4: Backpropagation and Computation Graphs | 自行閱讀 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture04-backprop.pdf) | | 2019.11.22 | Lecture 5: Linguistic Structure: Dependency Parsing | 勇伯 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture05-dep-parsing.pdf), 錄影 ([1](https://drive.google.com/open?id=1vE0kR0YxZLZI7T23d-fKZDt86vcHEcOh), [2](https://drive.google.com/open?id=1LmYY8GA805L1YZdjXbpSvzgqguQmJrwb), [3](https://drive.google.com/open?id=1ht-mX1sZ4jlJ24aYnx5rATJkrAAgcBl7), [4](https://drive.google.com/open?id=1k5qE-bAWs3UXo6FnYCz5kxj9ySzSNrl5)), [講者補充包(code,pdf等)](https://drive.google.com/open?id=1maznKSquPcHHSlHtjDmbxJO-SbzTtzl3) | | 2019.11.29 | Lecture 6: The probability of a sentence? Recurrent Neural Networks and Language Models | 博儒 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture06-rnnlm.pdf), 錄影 ([1](https://drive.google.com/open?id=1uz7S4kk-vTuChB1lN7LskTWAvktGPSSE), [2](https://drive.google.com/open?id=168MlzjBI5UY3poW3vpM_6K8Pkgkc3ATr)) | | 2019.12.06 | Lecture 7: Vanishing Gradients and Fancy RNNs | 博鈞 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture07-fancy-rnn.pdf), 錄影([1](https://drive.google.com/open?id=1o3ZkIt9aIuEMWB6TbFNi_fx0roxGzWv7), [2](https://drive.google.com/open?id=1QiDl4Z8VcEym63-07APnZMvMRCfn_Xm0), [3](https://drive.google.com/open?id=18vNQ17btLvpgaOPhe9GMOvXV_MmvGkl9)), [講者補充](https://docs.google.com/document/d/1pUczhr-WDe4-wlI9HjuebKqfGE9zCBL35ScyxJH1qC4/edit?usp=sharing) | | 2019.12.13 | Lecture 8: Machine Translation, Seq2Seq and Attention | 君諦 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture08-nmt.pdf), [錄影](https://drive.google.com/open?id=1CO6cQHa9hbLgFs-3xfCqF6ZbcSXTURPl) | | 2020.01.03 | Lecture 9: Practical Tips for Final Projects | 博儒 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture09-final-projects.pdf), 錄影 ([1](https://drive.google.com/open?id=1cyx3prNRnxjxsU5WWfzXtfkXGG5PYO2d), [2](https://drive.google.com/open?id=1s_mS0t7M6IMMMwCSvWW87qkwdadka6Ax)) | | 2020.01.03 | Lecture 10: Question Answering and the Default Final Project | 嘉冠 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture10-QA.pdf), [補充](https://hackmd.io/Qf_Q8YClQfie9TfuPb1L2A), [錄影](https://drive.google.com/open?id=1UWjqVArb6PPzmNLnGSPCVMSPB5oTO58R) | | 2020.01.17 | Lecture 11: ConvNets for NLP | 勇伯 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture11-convnets.pdf), 錄影([1](https://drive.google.com/open?id=1T0if3AaOh1gh9ZiPXnnee5F7fBNj39cz), [2](https://drive.google.com/open?id=1T12xAEtxgu6awQPXjl7EmC1wwGsZx61r), [3](https://drive.google.com/open?id=1T1u_iJqaYYBOJNHPMnTEkGsBX6-5BvI4), [4](https://drive.google.com/open?id=1T4VFqdLHOYn9X-RpZJDKBOuownF271Sd)), [講者補充包](https://drive.google.com/open?id=1lrwNUqoBfxexLt1Wndg42SrYAwjCA44a) | | 2020.02.14 | Lecture 12: Information from parts of words: Subword Models | 郭瑞申 | [講義(講者)](https://drive.google.com/file/d/1OD2rWy5FQ7Z-bY54b_Zautc662W7Luqy/view?usp=sharing), [講義(官方)](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture12-subwords.pdf), [錄影](https://drive.google.com/open?id=1wNWp-_UnUFvCrsWggovVM3bwYWLbEyxL) | | 2020.03.13 | Lecture 13: Modeling contexts of use: Contextual Representations and Pretraining | 博鈞 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture13-contextual-representations.pdf) , [錄影](https://drive.google.com/open?id=1NN7edvqw2o7XnFw_lbS2ek7ICDEDFmTb) | | 2020.03.20 | Lecture 14: Transformers and Self-Attention For Generative Models | 郭瑞申 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture14-transformers.pdf) , [錄影](https://drive.google.com/open?id=18nYuOYsDR3qjPLf8dvMFUVSy3xIJr1EI)| | 2020.04.10 | Lecture 15: Natural Language Generation | 勇伯 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture15-nlg.pdf), [講者補充包](https://drive.google.com/drive/u/0/folders/1XR7fJy5Gtb_7XYcNlc8DNsLZodr_Jm3Y), [錄影](https://drive.google.com/file/d/15PSYUm5Lq61oTfADAimJA9GrTSXYyBqQ/view?usp=sharing) | | 2020.04.17 | Lecture 16: Reference in Language and Coreference Resolution | 嘉冠 | [講義](http://web.stanford.edu/class/cs224n/slides/cs224n-2019-lecture16-coref.pdf), [錄影](https://drive.google.com/open?id=1XkA-KBiMcwWl8rJVrdCOa7RkT5C9Yd8u) ,[講者補充包](https://hackmd.io/L-8FCrVVRY-PYu3sJsYVTQ) | | 2020.05.08 | Lecture 17: Multitask Learning: A general model for NLP? | 博鈞 | [講義](https://web.stanford.edu/class/archive/cs/cs224n/cs224n.1194/slides/cs224n-2019-lecture17-multitask.pdf), [錄影](https://drive.google.com/file/d/1uG0JosIFsE5zUKNrgjcweQs8FQU7AULm/view?usp=sharing) | | 2020.05.15 | Lecture 18: Constituency Parsing and Tree Recursive Neural Networks | Vivian Ou | [講義](https://web.stanford.edu/class/archive/cs/cs224n/cs224n.1194/slides/cs224n-2019-lecture18-TreeRNNs.pdf), [錄影](https://drive.google.com/file/d/1UHL8WGlUrrZxVWEvusCHOGNtSEQhDB_U/view?usp=sharing) | | 2020.05.22 | Lecture 19: Safety, Bias, and Fairness | 嘉冠 | [講義](https://web.stanford.edu/class/archive/cs/cs224n/cs224n.1194/slides/cs224n-2019-lecture19-bias.pdf), [錄影](https://drive.google.com/file/d/1cjMgNsKWgVGFRUveBSagRrlk1t3nze4Z/view?usp=sharing) | | 2020.07.03 | Lecture 20: Future of NLP + Deep Learning | 郭瑞申 | [講義](https://web.stanford.edu/class/archive/cs/cs224n/cs224n.1194/slides/cs224n-2019-lecture20-future.pdf) | ## Theory of Generalization and Optimization in Deep Learning (2021) | 日期 | 年份 | 主題 | 講者 | 資源 | |-----|-----|------|-----|-----| | 2021.01.08 | 2019 |Fine-Grained Analysis of Optimization and Generalization for Over-parameterized Neural Networks| Mark Chang | [論文](https://arxiv.org/abs/1901.08584), [講義](https://drive.google.com/file/d/1eANnXhNYoelblzeXbZolFLxalvPe97xJ/view)| | 2021.01.22 | 2019 |Gradient Descent Provably Optimizes Over-parameterized Neural Networks | Mark Chang | [論文](https://arxiv.org/abs/1810.02054) | | 2021.02.05 | 2018 |Neural Tangent Kernel: Convergence and Generalization in Neural Networks | Mark Chang | [論文](https://arxiv.org/abs/1806.07572) | ## Neural-Symbolic AI | 日期 | 年份 | 主題 | 講者 | 資源 | |-----|-----|------|-----|-----| | | | | | | ## Domain Adaption | 日期 | 年份 | 主題 | 講者 | 資源 | |-----|-----|------|-----|-----| | | 2020 | A survey on domain adaptation theory: learning bounds and theoretical guarantees | Mark Chang| [論文](https://arxiv.org/abs/2004.11829) | ## Inverse Reinforcement Learning | 日期 | 年份 | 主題 | 講者 | 資源 | |-----|-----|------|-----|-----| |2021.05.07|2004 |Reinforcement Learning & Optimal Control| 坤賢 | | |2021.05.21 <br />2021.06.04|2008 |1. Apprenticeship Learning IRL<br />2. Maximum Entropy IRL | 坤賢 |[論文1](https://ai.stanford.edu/~ang/papers/icml04-apprentice.pdf) <br /> [論文2](https://www.aaai.org/Papers/AAAI/2008/AAAI08-227.pdf) | |2021.07.02|2016 |Guided Cost Learning | 郭瑞申 |[論文](https://arxiv.org/pdf/1603.00448.pdf) | |2021.07.30|2016 | A Connection between Generative Adversarial Networks, Inverse Reinforcement Learning, and Energy-Based Models| 郭瑞申 |[論文](https://arxiv.org/abs/1611.03852) | ## Knowledge Graph and its Applications for NLP | 日期 | 年份 | 主題 | 講者 | 資源 | |-----|-----|------|-----|-----| | | | | | | ## 參與者技術經驗交流 | 日期 | 主題 | 講者 | 資源 | |-----|-----|-----|-----| | 2019.01.11 | ECG Classification with CNN | 君諦 | [講義](https://drive.google.com/file/d/1ZpKkYhZY2Vc2KgmMgJTjkp72JofDPiiN/view?usp=sharing), [錄影](https://drive.google.com/file/d/1duHqikWpO95PBstYRps6oYUeth9jaBlq/view?usp=sharing) | | 2019.03.22 | Introduction to Deep Learning | Wilson Ho | [講義](https://docs.google.com/presentation/d/16j8Rk-ibugdgkfkovX4QTPRumUZoLr0Y924kkf3P6o4/edit?usp=sharing) | | 2019.04.19 | computer vision and how to create own image dataset | Elisa Chang | | | 2019.06.21 | AI視力檢測 | Scott Lin | | | 2019.07.19 | Chatbot開發經驗分享 | Wilson Ho | [講義](https://drive.google.com/open?id=1xLKxOvt_-G-vokTbwzasgl0ry-52Tm4I) ,[錄影1](https://drive.google.com/open?id=1xQagyrumnkWi9h-cy5aJCACKuhTQifaY) ,[錄影2](https://drive.google.com/open?id=1-R4LsfeB00Ytuhqe0hMO580rJs83zqbI)| | 2019.08.30 | Intel OpenVINO | 王宗業(Intel平台研發經理) | | | 2019.12.27 | 圖解物件偵測之Why, How, What - From OverFeat to WSMA-Seg | Wilson Ho | 錄影([1](https://drive.google.com/open?id=1wO05ll_mrF8Rt8eKYXX05s4sQyqHoeLn), [2](https://drive.google.com/open?id=16IMbG4zA72Gc431tLPxFrdU_qEfh0Vt6)), | 2020.06.05 | 圖解一階段物件偵測 - Part1 | Wilson Ho |[錄影](https://www.youtube.com/playlist?list=PLANbacZNzD9FOcLenvcfgE7R4QdHgOXSq) | | 2020.06.12 | 圖解一階段物件偵測 - Part2 | Wilson Ho |[錄影](https://www.youtube.com/playlist?list=PLANbacZNzD9FOcLenvcfgE7R4QdHgOXSq)| | 2020.06.19 | 圖解一階段物件偵測 - Part3 | Wilson Ho |[錄影](https://www.youtube.com/playlist?list=PLANbacZNzD9FOcLenvcfgE7R4QdHgOXSq)| | 2020.07.17 | 2020 科技大擂台比賽經驗分享 | CMIND, 冠廷, 嘉冠 | [講義](https://docs.google.com/presentation/d/1N9b6ggDz7Y70Kr3pP8lv8_dEPSFk-fe4hwALdXhieCw/edit?usp=sharing), [錄影](https://www.facebook.com/ROD455287/videos/1109211666129111) | | 2020.08.14 |Self-Constructing Fuzzy Neural Network and Differential Evolution Algorithm for Implementation of Parameter Estimation | 張明弘 | [錄影](https://www.facebook.com/ROD455287/videos/631225654458622) | | 2020.09.25 | What's Wrong with Deep Learning? Lessons from Gary Marcus | 嘉冠 | [講義](https://docs.google.com/presentation/d/170zATsCMogfxhfGK0ozKMAa59N9iBnNRv4jXZBzN9ns/edit?usp=sharing), [錄影](https://www.facebook.com/106405837440910/videos/376486893368060) | | 2020.10.16 |Risk Bound for Interpolated Models| Mark Chang | [論文](https://arxiv.org/abs/1806.05161), [講義](https://drive.google.com/file/d/1EtmB4zOgBpmKb-LyMKP6zsaoYEmAXIS2/view?usp=sharing), [錄影](https://www.facebook.com/106405837440910/videos/776200606551052) | | 2020.11.06 | Optimization and Generalization for Over-parameterized Neural Networks | Mark Chang | [論文](https://arxiv.org/abs/1901.08584), [講義](https://drive.google.com/file/d/12RGzGd89JwUZ_QMVi9M7SYpbmbewJoTR/view?usp=sharing) | | 2020.12.04 | 圖解 Meta-Learning for Few-Shot Learning | Wilson Ho | [論文](https://arxiv.org/abs/1904.05046) | | 2021.02.19 | 2020 Study on Reinforcement Learning: Drivers, Restraints & Opportunities | 謝其宏 |[論文](https://arxiv.org/pdf/1904.12901.pdf), [講義](https://docs.google.com/presentation/d/1O6SJf4CCjHLwn_41Wp0q8xsq9HIgNLj1CS-UUWtsR30/edit?usp=sharing), [錄影](https://www.facebook.com/ROD455287/videos/330427088373163) | | 2021.03.12 | Graph Neural Network : Introductions Models and Applications Part I | Berlin Cho |[講義](), [錄影1](https://www.facebook.com/ROD455287/videos/903213517178883) [錄影2](https://www.facebook.com/ROD455287/videos/2764701487174861)| | 2021.03.26 | Big Bird: Transformers for Longer Sequences | 冠廷 |[論文](https://arxiv.org/abs/2007.14062), [講義](), [錄影](https://www.facebook.com/ROD455287/videos/284224516540896) | | 2021.04.09 | Graph Neural Network : Introductions Models and Applications Part II | Berlin Cho |[講義](), [錄影]() | ## 參考資源 - [Republic of Developer ](https://drive.google.com/open?id=13b3jsf6A8NBDREOFcX4M90_33SMhLYE_) - [Deep Learning Papers Reading Roadmap](https://github.com/floodsung/Deep-Learning-Papers-Reading-Roadmap) - [Browse State-of-the-Art](https://paperswithcode.com/sota?fbclid=IwAR3GIXjZkjwoX8tYDr_QclrfWAe8IMUZQ-hJq-V8KmAns4QIZbu4qPESAKk)