With the tremendous growth of social networks, there has been a growth in the amount of new data that is being created every minute on these networking sites. The notion of community in this social networking world has caught lots of attention. Studying Twitter is useful for understanding how people use new communication technologies to form social connections and maintain existing ones. We analysed how geo-tagged tweets in Twitter can be used to identify useful user features and behavior as well as identify landmarks/places of interests. We also analysed several clustering algorithms and proposed different similarity measures to detect communities.
Slide deck for talk at IETF#92 (Dallas, March 2015) at the IETF Light-Weight Implementation Guidance (lwig) working group about the performance of cryptographic algorithms on ARM processors.
Can ChatGPT be compatible with the GDPR? Discuss.Lilian Edwards
Since the Italian Garantie became the first DP authority in the world to even temporarily ban ChatGPT, debate has broken out as to whether generative AI models can comply with data protection laws, not just in the GDPR but around the world. The use of personal data for training requires a legal basis which is hard to find, special category data raises special problems (duh) and the model itself may be considered personal data due to inversion attacks and data leakage in outputs. Hallucination presents seemingly insuperable problems as to accuracy and rectification. Even though Open AI have temporarily satisfied the Garantie, further disputes still seem likely to eventually reach the courts. In this talk I will attempt to throw the entirety of DP law against the wall of large language and image models and even, jut for fun, raise the spectre of whether AI models can libel
Arduino based 74-series integrated circuits testing system at gate level IJECEIAES
The goal of this research article is to build and implement a low-cost, user-friendly 74-series logic integrated circuits (ICs) tester that is independent of a computer. Depending on the truth table of the gates and the IC configuration, the logic IC tester will be able to test the operation of the 74 series logic gates (AND, OR, NOR, NAND, XOR) of those ICs. It is feasible to test a range of logic ICs with higher pin widths thanks to the proposed system’s usage of an Arduino Mega platform module as a microcontroller, which provides the ability to connect 54 programmed logic inputs or outputs. The versatility offered by this design and the use of a personal computer allow for the reprograming and updating of the logic IC functional tester. Any 74-series ICs testing outcome will be shown on liquid crystal display (LCD) at the gate level. The logic IC functional tester was successfully constructed and operates flawlessly.
This document provides an overview of SQL (Structured Query Language). It covers SQL basics, data types, constraints, operators, functions, clauses, queries, normalization and more. Several SQL examples are also included to demonstrate concepts like SELECT statements, WHERE clauses, operators, and more. The document appears to be part of a training or tutorial on SQL and relational database concepts.
Can ChatGPT be compatible with the GDPR? Discuss.Lilian Edwards
Since the Italian Garantie became the first DP authority in the world to even temporarily ban ChatGPT, debate has broken out as to whether generative AI models can comply with data protection laws, not just in the GDPR but around the world. The use of personal data for training requires a legal basis which is hard to find, special category data raises special problems (duh) and the model itself may be considered personal data due to inversion attacks and data leakage in outputs. Hallucination presents seemingly insuperable problems as to accuracy and rectification. Even though Open AI have temporarily satisfied the Garantie, further disputes still seem likely to eventually reach the courts. In this talk I will attempt to throw the entirety of DP law against the wall of large language and image models and even, jut for fun, raise the spectre of whether AI models can libel
Arduino based 74-series integrated circuits testing system at gate level IJECEIAES
The goal of this research article is to build and implement a low-cost, user-friendly 74-series logic integrated circuits (ICs) tester that is independent of a computer. Depending on the truth table of the gates and the IC configuration, the logic IC tester will be able to test the operation of the 74 series logic gates (AND, OR, NOR, NAND, XOR) of those ICs. It is feasible to test a range of logic ICs with higher pin widths thanks to the proposed system’s usage of an Arduino Mega platform module as a microcontroller, which provides the ability to connect 54 programmed logic inputs or outputs. The versatility offered by this design and the use of a personal computer allow for the reprograming and updating of the logic IC functional tester. Any 74-series ICs testing outcome will be shown on liquid crystal display (LCD) at the gate level. The logic IC functional tester was successfully constructed and operates flawlessly.
This document provides an overview of SQL (Structured Query Language). It covers SQL basics, data types, constraints, operators, functions, clauses, queries, normalization and more. Several SQL examples are also included to demonstrate concepts like SELECT statements, WHERE clauses, operators, and more. The document appears to be part of a training or tutorial on SQL and relational database concepts.
本スライドは、弊社の梅本により弊社内の技術勉強会で使用されたものです。
近年注目を集めるアーキテクチャーである「Transformer」の解説スライドとなっております。
"Arithmer Seminar" is weekly held, where professionals from within and outside our company give lectures on their respective expertise.
The slides are made by the lecturer from outside our company, and shared here with his/her permission.
Arithmer株式会社は東京大学大学院数理科学研究科発の数学の会社です。私達は現代数学を応用して、様々な分野のソリューションに、新しい高度AIシステムを導入しています。AIをいかに上手に使って仕事を効率化するか、そして人々の役に立つ結果を生み出すのか、それを考えるのが私たちの仕事です。
Arithmer began at the University of Tokyo Graduate School of Mathematical Sciences. Today, our research of modern mathematics and AI systems has the capability of providing solutions when dealing with tough complex issues. At Arithmer we believe it is our job to realize the functions of AI through improving work efficiency and producing more useful results for society.
Seeing Unseens with Machine Learning -- 見えていないものを見出す機械学習Tatsuya Shirakawa
Deep Learningを筆頭に、データから意味やパターンを抽出する機械学習は、いまや誰もが使えるツールになりつつあります。
本セッションでは、AIブームわく最中、機械学習がなぜ大事なのか、どんな使い方をするのが重要になっていくかについて展望しつつ、「見えていなかったものを見出す」というネクストフロンティアになるであろう機械学習の方向性についてお話します。
This document summarizes a paper titled "DeepI2P: Image-to-Point Cloud Registration via Deep Classification". The paper proposes a method for estimating the camera pose within a point cloud map using a deep learning model. The model first classifies whether points in the point cloud fall within the camera's frustum or image grid. It then performs pose optimization to estimate the camera pose by minimizing the projection error of inlier points onto the image. The method achieves more accurate camera pose estimation compared to existing techniques based on feature matching or depth estimation. It provides a new approach for camera localization using point cloud maps without requiring cross-modal feature learning.
2020/10/10に開催された第4回全日本コンピュータビジョン勉強会「人に関する認識・理解論文読み会」発表資料です。
以下の2本を読みました
Harmonious Attention Network for Person Re-identification. (CVPR2018)
Weekly Supervised Person Re-Identification (CVPR2019)
2018/10/20コンピュータビジョン勉強会@関東「ECCV読み会2018」発表資料
Yew, Z. J., & Lee, G. H. (2018). 3DFeat-Net: Weakly Supervised Local 3D Features for Point Cloud Registration. European Conference on Computer Vision.
6. 紹介する論文
7
State Space Models for Event Cameras
Nikola Zubic, Mathias Gehrig, Davide Scaramuzza
Robotics and Perception Group, University of Zurich, Switzerland
イベントカメラを用いた物体検出等を行う従来手法は、学習測度の
問題や、学習時と異なる周波数に対応できない問題を、状態空間
モデル(SSM)を導入することで解決
8. Related Work
9
Gehrig, M., & Scaramuzza, D. (2023). RecurrentVisionTransformers for
Object Detection with Event Cameras. Proceedings of the IEEE Computer
Society Conference on ComputerVision and Pattern Recognition (CVPR)
VisionTransformer + LSTMを用いて、イベントカメラから物体検出
9. Related Work
10
Gehrig, M., & Scaramuzza, D. (2023). RecurrentVisionTransformers for
Object Detection with Event Cameras. Proceedings of the IEEE Computer
Society Conference on ComputerVision and Pattern Recognition (CVPR)
VisionTransformer + LSTMを用いて、イベントカメラから物体検出
𝒆𝑘 = (𝑥𝑘, 𝑦𝑘, 𝑡𝑘, 𝑝𝑘)
画素の
座標
発生
時刻
変化方向
(正/負)
(2𝑇, 𝐻, 𝑊)とすることで、
画像として処理
10. Related Work
11
Gehrig, M., & Scaramuzza, D. (2023). RecurrentVisionTransformers for
Object Detection with Event Cameras. Proceedings of the IEEE Computer
Society Conference on ComputerVision and Pattern Recognition (CVPR)
VisionTransformer + LSTMを用いて、イベントカメラから物体検出
空間内での
Local Window
Self-Attention
空間内での
Dilated
Attention
13. HiPPO
14
Gu,A., Dao,T., Ermon, S., Rudra,A., & Ré, C. (2020). HiPPO: Recurrent
memory with optimal polynomial projections.Advances in Neural Information
Processing Systems (NeurIPS).
系列データを関数(直交多項式の和)で近似することで、理論上無限長の
系列を扱うことができる。
14. HiPPO
15
Gu,A., Dao,T., Ermon, S., Rudra,A., & Ré, C. (2020). HiPPO: Recurrent
memory with optimal polynomial projections.Advances in Neural Information
Processing Systems (NeurIPS).
系列データを関数(直交多項式の和)で近似することで、理論上無限長の
系列を扱うことができる。
任意の関数𝑓(𝑡)
からサンプリングし
た系列データ𝑓𝑖
系列データをN個の
直交多項式へ投影
𝜇(𝑡𝑖)
: 測度(=データへの重み)
基底関数(直交多項
式)の係数𝑐(𝑡)
係数𝑐(𝑡)を逐次的に更新
離散化
𝐴𝑘はサンプリング間隔∆t
に依存しない(学習時と推
論時に異なる周波数に対
応)
15. LSSL
16
Gu,A., Johnson, I., Goel, K., Saab, K., Dao,T., Rudra,A., & Ré, C. (2021). Combining Recurrent,
Convolutional, and Continuous-time Models with Linear State-Space Layers. Advances in Neural
Information Processing Systems (NeurIPS)
HiPPOを状態空間モデルへ拡張
逐次処理を畳み込み処理として並列化
16. LSSL
17
Gu,A., Johnson, I., Goel, K., Saab, K., Dao,T., Rudra,A., & Ré, C. (2021). Combining Recurrent,
Convolutional, and Continuous-time Models with Linear State-Space Layers. Advances in Neural
Information Processing Systems (NeurIPS)
HiPPOを状態空間モデルへ拡張
逐次処理を畳み込み処理として並列化
状態空間モデル
ሶ
𝑥 𝑡 = 𝐀𝑥 𝑡 + 𝐁𝑢 𝑡
𝑦 𝑡 = 𝐂𝑥 𝑡 + 𝐃𝑢 𝑡
入力
内部状態
出力
HiPPO
ሶ
𝑐 𝑡 = 𝐀𝑐 𝑡 + 𝐁𝑓 𝑡
17. LSSL
18
Gu,A., Johnson, I., Goel, K., Saab, K., Dao,T., Rudra,A., & Ré, C. (2021). Combining Recurrent,
Convolutional, and Continuous-time Models with Linear State-Space Layers. Advances in Neural
Information Processing Systems (NeurIPS)
HiPPOを状態空間モデルへ拡張
逐次処理を畳み込み処理として並列化
状態空間モデル
ሶ
𝑥 𝑡 = 𝐀𝑥 𝑡 + 𝐁𝑢 𝑡
𝑦 𝑡 = 𝐂𝑥 𝑡 + 𝐃𝑢 𝑡
入力
内部状態
出力
離散化状態空間モデル
𝑥𝑘 = ഥ
𝐀𝑥𝑘−1 + ഥ
𝐁𝑢𝑘
𝑦𝑘 = ҧ
𝐂𝑥𝑘 + ഥ
𝐃𝑢𝑘
19. S4
20
Gu,A., Goel, K., & Ré, C. (2022). Efficiently Modeling Long Sequences
With Structured State Spaces. International Conference on Learning
Representations (ICLR).
畳み込みカーネルഥ
𝑲の計算を様々な数学的テクニックを用いて簡
略化
20. S4
21
Gu,A., Goel, K., & Ré, C. (2022). Efficiently Modeling Long SequencesWith
Structured State Spaces. International Conference on Learning Representations
(ICLR).
畳み込みカーネル𝑲の計算を様々な数学的テクニックを用いて簡略化
Diagonal Plus Low-Rank
𝑨 = 𝚲 − 𝒑𝒒∗
対角行列と低階級の和
ሶ
𝑥 𝑡 = 𝐀𝑥 𝑡 + 𝐁𝑢 𝑡
𝑦 𝑡 = 𝐂𝑥 𝑡
周波数領域で畳
み込みカーネル
𝑲を生成
周波数領域で
𝒖と𝑲を乗算
(=時間領域で
畳み込み)
入力信号𝒖を
フーリエ変換
出信号𝒚を逆
フーリエ変換
21. S4D
22
Gu,A., Gupta,A., Goel, K., & Ré, C. (2022). On the Parameterization and
Initialization of Diagonal State Space Models. Advances in Neural Information
Processing Systems (NeurIPS)
HiPPO行列のDPLRからlow-rank項を取り除き、対角行列のみの形にして
も、実験的にうまくいくことがわかっている。
この現象を数学的に解析し、S4より簡易な手法を提案
22. S4D
23
Gu,A., Gupta,A., Goel, K., & Ré, C. (2022). On the Parameterization and
Initialization of Diagonal State Space Models. Advances in Neural Information
Processing Systems (NeurIPS)
HiPPO行列のDPLRからlow-rank項を取り除き、対角行列のみの形にして
も、実験的にうまくいくことがわかっている。
この現象を数学的に解析し、より簡易な手法を提案
対角行列
23. S5
24
Smith, J.T. H.,Warrington,A., & Linderman, S.W. (2023). Simplified State Space Layers
for Sequence Modeling. International Conference on Learning Representation (ICLR)
S4は入力𝒖のチャネルを個別に処理するのに対し、S5は全チャネルを一括で処理
S4D同様DPLRの対角成分のみ使用
畳み込み処理の代わりに再帰処理を並列化(Parallel Scan)