January 6, 2025 at 12:55 AM
DeepWalkにテキスト情報を埋め込む論文。ネットワーク構造の埋め込みとノードのテキスト情報（引用ネットワークな要約のベクトル表現とか）を結合するよりも各タスクの予測精度が高いことを示す論文。ミソはDeepWalkが行列分解で定式化することができ、その行つ分解にテキスト情報(e.g.,埋め込み表現、200次元のテキストのベクトル表現をつかって200×|V|)を入れ込むこと。論文ではこれをTADW(text-associated Deep Walk)と読んでいる。
Network representation learning with rich text information
Network representation learning with rich text information
December 29, 2025 at 10:27 PM
DeepWalkにテキスト情報を埋め込む論文。ネットワーク構造の埋め込みとノードのテキスト情報（引用ネットワークな要約のベクトル表現とか）を結合するよりも各タスクの予測精度が高いことを示す論文。ミソはDeepWalkが行列分解で定式化することができ、その行つ分解にテキスト情報(e.g.,埋め込み表現、200次元のテキストのベクトル表現をつかって200×|V|)を入れ込むこと。論文ではこれをTADW(text-associated Deep Walk)と読んでいる。
Network representation learning with rich text information
Network representation learning with rich text information
node2vec、結局これはnegative samplingで学習を効率化しつつサンプリングの方法をBFS/DFSの度合いをパラメータで設定して調整できるようにして、ランダムウォークじゃないようにしたのか。その点がdeepwalkと違うと。なるほどね。
node2vec: Scalable Feature Learning for Networks
node2vec: Scalable Feature Learning for Networks
December 22, 2025 at 11:32 PM
node2vec、結局これはnegative samplingで学習を効率化しつつサンプリングの方法をBFS/DFSの度合いをパラメータで設定して調整できるようにして、ランダムウォークじゃないようにしたのか。その点がdeepwalkと違うと。なるほどね。
node2vec: Scalable Feature Learning for Networks
node2vec: Scalable Feature Learning for Networks
ノードの埋め込み表現にランダムサーフを利用する手法DNGRを対案している論文。Deepwalkなどはランダムにネットワークの流列を学習するが、この手法はノードがKステップ内の共起をとらえたPCO行列を利用し大局的な情報を得る。PCO行列から「実際に一緒に起きた確率」と「偶然一緒に起きる確率」の比率を計算した、PPMI行列で頻度の高いノードの過大評価を防ぎ(NLPでいうstop wordsのようなもの)、最後にSDAE（オートエンコーダ）で非線形な特徴を学習するのが特徴。
Deep Learning for Learning Graph Representations
Deep Learning for Learning Graph Representations
January 7, 2026 at 10:22 PM
ノードの埋め込み表現にランダムサーフを利用する手法DNGRを対案している論文。Deepwalkなどはランダムにネットワークの流列を学習するが、この手法はノードがKステップ内の共起をとらえたPCO行列を利用し大局的な情報を得る。PCO行列から「実際に一緒に起きた確率」と「偶然一緒に起きる確率」の比率を計算した、PPMI行列で頻度の高いノードの過大評価を防ぎ(NLPでいうstop wordsのようなもの)、最後にSDAE（オートエンコーダ）で非線形な特徴を学習するのが特徴。
Deep Learning for Learning Graph Representations
Deep Learning for Learning Graph Representations
## 알고리즘 기반 바이오 분석: 단백질 상호작용 네트워크 분석을 위한 그래프 임베딩 기반 준지도 학습 프레임워크 개발
**초록** 본 연구는 단백질 상호작용 네트워크(Protein-Protein Interaction Network, PPI network) 분석의 정확도와 효율성을 극대화하기 위해 그래프 임베딩 기법과 준지도 학습(Semi-Supervised Learning, SSL)을 결합한 새로운 프레임워크를 제안한다. 기존의 네트워크 분석 방법들은 고차원 데이터의 복잡성을 효과적으로 처리하지 못하거나, 레이블링된 데이터…
**초록** 본 연구는 단백질 상호작용 네트워크(Protein-Protein Interaction Network, PPI network) 분석의 정확도와 효율성을 극대화하기 위해 그래프 임베딩 기법과 준지도 학습(Semi-Supervised Learning, SSL)을 결합한 새로운 프레임워크를 제안한다. 기존의 네트워크 분석 방법들은 고차원 데이터의 복잡성을 효과적으로 처리하지 못하거나, 레이블링된 데이터…
## 알고리즘 기반 바이오 분석: 단백질 상호작용 네트워크 분석을 위한 그래프 임베딩 기반 준지도 학습 프레임워크 개발
**초록** 본 연구는 단백질 상호작용 네트워크(Protein-Protein Interaction Network, PPI network) 분석의 정확도와 효율성을 극대화하기 위해 그래프 임베딩 기법과 준지도 학습(Semi-Supervised Learning, SSL)을 결합한 새로운 프레임워크를 제안한다. 기존의 네트워크 분석 방법들은 고차원 데이터의 복잡성을 효과적으로 처리하지 못하거나, 레이블링된 데이터 부족 문제에 직면하여 정확도가 제한적이었다. 본 연구에서는 DeepWalk, node2vec과 같은 그래프 임베딩 기법을 활용하여 PPI 네트워크의 위상학적 정보를 저차원 벡터 공간에 효율적으로 표현하고, 소량의 레이블링된 데이터를 활용하여 전체 네트워크의 단백질 기능을 예측하는 준지도 학습 모델을 개발한다.
freederia.com
March 22, 2025 at 5:27 AM
## 알고리즘 기반 바이오 분석: 단백질 상호작용 네트워크 분석을 위한 그래프 임베딩 기반 준지도 학습 프레임워크 개발
**초록** 본 연구는 단백질 상호작용 네트워크(Protein-Protein Interaction Network, PPI network) 분석의 정확도와 효율성을 극대화하기 위해 그래프 임베딩 기법과 준지도 학습(Semi-Supervised Learning, SSL)을 결합한 새로운 프레임워크를 제안한다. 기존의 네트워크 분석 방법들은 고차원 데이터의 복잡성을 효과적으로 처리하지 못하거나, 레이블링된 데이터…
**초록** 본 연구는 단백질 상호작용 네트워크(Protein-Protein Interaction Network, PPI network) 분석의 정확도와 효율성을 극대화하기 위해 그래프 임베딩 기법과 준지도 학습(Semi-Supervised Learning, SSL)을 결합한 새로운 프레임워크를 제안한다. 기존의 네트워크 분석 방법들은 고차원 데이터의 복잡성을 효과적으로 처리하지 못하거나, 레이블링된 데이터…
Houlin Zhou, Yejin Wang, Xufei Tang, Dan Zhuang
Spectral Dynamics of DeepWalk Embeddings for Dynamic Network Change-Point Detection
https://arxiv.org/abs/2609.17893
Spectral Dynamics of DeepWalk Embeddings for Dynamic Network Change-Point Detection
https://arxiv.org/abs/2609.17893
September 18, 2026 at 7:25 AM
Houlin Zhou, Yejin Wang, Xufei Tang, Dan Zhuang
Spectral Dynamics of DeepWalk Embeddings for Dynamic Network Change-Point Detection
https://arxiv.org/abs/2609.17893
Spectral Dynamics of DeepWalk Embeddings for Dynamic Network Change-Point Detection
https://arxiv.org/abs/2609.17893
## 알고리즘 기반 바이오 분석: 단백질 상호작용 네트워크 분석을 위한 그래프 임베딩 기반 준지도 학습 프레임워크 개발
**초록** 본 연구는 단백질 상호작용 네트워크(Protein-Protein Interaction Network, PPI network) 분석의 정확도와 효율성을 극대화하기 위해 그래프 임베딩 기법과 준지도 학습(Semi-Supervised Learning, SSL)을 결합한 새로운 프레임워크를 제안한다. 기존의 네트워크 분석 방법들은 고차원 데이터의 복잡성을 효과적으로 처리하지 못하거나, 레이블링된 데이터…
**초록** 본 연구는 단백질 상호작용 네트워크(Protein-Protein Interaction Network, PPI network) 분석의 정확도와 효율성을 극대화하기 위해 그래프 임베딩 기법과 준지도 학습(Semi-Supervised Learning, SSL)을 결합한 새로운 프레임워크를 제안한다. 기존의 네트워크 분석 방법들은 고차원 데이터의 복잡성을 효과적으로 처리하지 못하거나, 레이블링된 데이터…
## 알고리즘 기반 바이오 분석: 단백질 상호작용 네트워크 분석을 위한 그래프 임베딩 기반 준지도 학습 프레임워크 개발
**초록** 본 연구는 단백질 상호작용 네트워크(Protein-Protein Interaction Network, PPI network) 분석의 정확도와 효율성을 극대화하기 위해 그래프 임베딩 기법과 준지도 학습(Semi-Supervised Learning, SSL)을 결합한 새로운 프레임워크를 제안한다. 기존의 네트워크 분석 방법들은 고차원 데이터의 복잡성을 효과적으로 처리하지 못하거나, 레이블링된 데이터 부족 문제에 직면하여 정확도가 제한적이었다. 본 연구에서는 DeepWalk, node2vec과 같은 그래프 임베딩 기법을 활용하여 PPI 네트워크의 위상학적 정보를 저차원 벡터 공간에 효율적으로 표현하고, 소량의 레이블링된 데이터를 활용하여 전체 네트워크의 단백질 기능을 예측하는 준지도 학습 모델을 개발한다.
freederia.com
March 22, 2025 at 5:27 AM
## 알고리즘 기반 바이오 분석: 단백질 상호작용 네트워크 분석을 위한 그래프 임베딩 기반 준지도 학습 프레임워크 개발
**초록** 본 연구는 단백질 상호작용 네트워크(Protein-Protein Interaction Network, PPI network) 분석의 정확도와 효율성을 극대화하기 위해 그래프 임베딩 기법과 준지도 학습(Semi-Supervised Learning, SSL)을 결합한 새로운 프레임워크를 제안한다. 기존의 네트워크 분석 방법들은 고차원 데이터의 복잡성을 효과적으로 처리하지 못하거나, 레이블링된 데이터…
**초록** 본 연구는 단백질 상호작용 네트워크(Protein-Protein Interaction Network, PPI network) 분석의 정확도와 효율성을 극대화하기 위해 그래프 임베딩 기법과 준지도 학습(Semi-Supervised Learning, SSL)을 결합한 새로운 프레임워크를 제안한다. 기존의 네트워크 분석 방법들은 고차원 데이터의 복잡성을 효과적으로 처리하지 못하거나, 레이블링된 데이터…
Houlin Zhou, Yejin Wang, Xufei Tang, Dan Zhuang: Spectral Dynamics of DeepWalk Embeddings for Dynamic Network Change-Point Detection https://arxiv.org/abs/2609.17893 https://arxiv.org/pdf/2609.17893 https://arxiv.org/html/2609.17893
September 17, 2026 at 7:03 AM
Houlin Zhou, Yejin Wang, Xufei Tang, Dan Zhuang: Spectral Dynamics of DeepWalk Embeddings for Dynamic Network Change-Point Detection https://arxiv.org/abs/2609.17893 https://arxiv.org/pdf/2609.17893 https://arxiv.org/html/2609.17893
🧬 Identifying miRNA–disease associations through biological experiments can be time-consuming & costly. This study introduces a DeepWalk-based approach designed to predict these associations and explores its performance. #miRNA #Bioinformatics
mdpi.com/2227-9059/13...
mdpi.com/2227-9059/13...
September 1, 2026 at 11:26 PM
🧬 Identifying miRNA–disease associations through biological experiments can be time-consuming & costly. This study introduces a DeepWalk-based approach designed to predict these associations and explores its performance. #miRNA #Bioinformatics
mdpi.com/2227-9059/13...
mdpi.com/2227-9059/13...
DeepWalk: 5-Minute @Neo4j Interview with Dr. @StevenSkiena. #DataScience #Analytics #Graph http://bit.ly/2WWxaDg
November 23, 2024 at 6:05 PM
DeepWalk: 5-Minute @Neo4j Interview with Dr. @StevenSkiena. #DataScience #Analytics #Graph http://bit.ly/2WWxaDg
🔥REMOTE JOB(Worldwide): DeepWalk is hiring Staff Software Engineer. Apply → https://tryremote.com/remote-jobs/344526-staff-software-engineer-deepwalk
#remote #remotejob #techjob #hiring #WorkFromHome #JobSearch
#remote #remotejob #techjob #hiring #WorkFromHome #JobSearch
Staff Software Engineer at DeepWalk | Work from home
<p>Remote tech job → Worldwide, DeepWalk is hiring a Staff Software Engineer. About DeepWalkDeepWalk is a fast-growing venture that helps cities keep people safe using computer vision to map and monitor their sidewalks. We have ongoing contracts with cities, universities, and e...</p>
tryremote.com
May 8, 2026 at 9:00 AM
🔥REMOTE JOB(Worldwide): DeepWalk is hiring Staff Software Engineer. Apply → https://tryremote.com/remote-jobs/344526-staff-software-engineer-deepwalk
#remote #remotejob #techjob #hiring #WorkFromHome #JobSearch
#remote #remotejob #techjob #hiring #WorkFromHome #JobSearch
Spokane is investing $500,000 in a groundbreaking sidewalk pilot project that uses cutting-edge lidar technology to assess condition and inform future funding strategies!
Click to read more!
#SpokaneSpokaneCounty #WA #CitizenPortal #PublicSafety #ClimateResilience #TransportationPolicy
Click to read more!
#SpokaneSpokaneCounty #WA #CitizenPortal #PublicSafety #ClimateResilience #TransportationPolicy
City presents sidewalk pilot using lidar data; $500,000 pilot to define needs and options
City staff outlined a $500,000 sidewalk pilot that prioritizes data collection and policy design over immediate construction, using DeepWalk lidar scans and in‑house GIS processing to estimate citywide sidewalk condition and costs; ADA ramp upgrades and funding mechanisms were discussed.
citizenportal.ai
November 27, 2025 at 11:04 AM
Spokane is investing $500,000 in a groundbreaking sidewalk pilot project that uses cutting-edge lidar technology to assess condition and inform future funding strategies!
Click to read more!
#SpokaneSpokaneCounty #WA #CitizenPortal #PublicSafety #ClimateResilience #TransportationPolicy
Click to read more!
#SpokaneSpokaneCounty #WA #CitizenPortal #PublicSafety #ClimateResilience #TransportationPolicy
Two Layer Walk: A Community-Aware Graph Embedding
https://arxiv.org/abs/2412.12933
Community structures are critical for understanding the mesoscopic organization of networks, bridging local and global patterns. While methods such as DeepWalk and node2vec capture local positional information ...📈🤖
https://arxiv.org/abs/2412.12933
Community structures are critical for understanding the mesoscopic organization of networks, bridging local and global patterns. While methods such as DeepWalk and node2vec capture local positional information ...📈🤖
December 18, 2024 at 4:51 PM
Two Layer Walk: A Community-Aware Graph Embedding
https://arxiv.org/abs/2412.12933
Community structures are critical for understanding the mesoscopic organization of networks, bridging local and global patterns. While methods such as DeepWalk and node2vec capture local positional information ...📈🤖
https://arxiv.org/abs/2412.12933
Community structures are critical for understanding the mesoscopic organization of networks, bridging local and global patterns. While methods such as DeepWalk and node2vec capture local positional information ...📈🤖
Analyzing the Evolution of Graphs and Texts
https://arxiv.org/abs/2411.06295
With the recent advance of representation learning algorithms on graphs (e.g., DeepWalk/GraphSage) and natural languages (e.g., Word2Vec/BERT) , the state-of-the art models can even achieve human-level performance ov...📈🤖
https://arxiv.org/abs/2411.06295
With the recent advance of representation learning algorithms on graphs (e.g., DeepWalk/GraphSage) and natural languages (e.g., Word2Vec/BERT) , the state-of-the art models can even achieve human-level performance ov...📈🤖
November 12, 2024 at 4:31 PM
Analyzing the Evolution of Graphs and Texts
https://arxiv.org/abs/2411.06295
With the recent advance of representation learning algorithms on graphs (e.g., DeepWalk/GraphSage) and natural languages (e.g., Word2Vec/BERT) , the state-of-the art models can even achieve human-level performance ov...📈🤖
https://arxiv.org/abs/2411.06295
With the recent advance of representation learning algorithms on graphs (e.g., DeepWalk/GraphSage) and natural languages (e.g., Word2Vec/BERT) , the state-of-the art models can even achieve human-level performance ov...📈🤖
Revolutionizing ADA Compliance: Harnessing GIS for Streamlined Reporting
Free webinar Wed Sep 10 | GIS and AI for Local Government Compliance Reporting and Asset Management WEBINAR SERIES
events.zoom.us/e/view/DUzT8...
#GIS #Bad_Elf #Carahsoft #DeepWalk #ADA #ADACompliance
Free webinar Wed Sep 10 | GIS and AI for Local Government Compliance Reporting and Asset Management WEBINAR SERIES
events.zoom.us/e/view/DUzT8...
#GIS #Bad_Elf #Carahsoft #DeepWalk #ADA #ADACompliance
September 2, 2025 at 8:36 PM
Revolutionizing ADA Compliance: Harnessing GIS for Streamlined Reporting
Free webinar Wed Sep 10 | GIS and AI for Local Government Compliance Reporting and Asset Management WEBINAR SERIES
events.zoom.us/e/view/DUzT8...
#GIS #Bad_Elf #Carahsoft #DeepWalk #ADA #ADACompliance
Free webinar Wed Sep 10 | GIS and AI for Local Government Compliance Reporting and Asset Management WEBINAR SERIES
events.zoom.us/e/view/DUzT8...
#GIS #Bad_Elf #Carahsoft #DeepWalk #ADA #ADACompliance
and classification. While classical methods such as DeepWalk, node2vec, and graph convolutional networks learn node embeddings by capturing structural and relational patterns in graphs, they often require significant computational resources and [2/7 of https://arxiv.org/abs/2503.06332v1]
March 11, 2025 at 6:47 AM
and classification. While classical methods such as DeepWalk, node2vec, and graph convolutional networks learn node embeddings by capturing structural and relational patterns in graphs, they often require significant computational resources and [2/7 of https://arxiv.org/abs/2503.06332v1]
🎤 CAMINATA AL ORIGEN DE LA PACHA
👍🏽 Una caminata para mirar el pasado, actuar en el presente y decidir nuestro futuro.
Facebook
https://f.mtr.cool/rmtlovtkhl
SPOTIFY
https://f.mtr.cool/xmklpomoxx
YouTube
https://f.mtr.cool/wzmnxfejkc
#caminataalorigen #Deepwalk #Ecuador
👍🏽 Una caminata para mirar el pasado, actuar en el presente y decidir nuestro futuro.
https://f.mtr.cool/rmtlovtkhl
SPOTIFY
https://f.mtr.cool/xmklpomoxx
YouTube
https://f.mtr.cool/wzmnxfejkc
#caminataalorigen #Deepwalk #Ecuador
June 26, 2025 at 12:30 PM
🎤 CAMINATA AL ORIGEN DE LA PACHA
👍🏽 Una caminata para mirar el pasado, actuar en el presente y decidir nuestro futuro.
Facebook
https://f.mtr.cool/rmtlovtkhl
SPOTIFY
https://f.mtr.cool/xmklpomoxx
YouTube
https://f.mtr.cool/wzmnxfejkc
#caminataalorigen #Deepwalk #Ecuador
👍🏽 Una caminata para mirar el pasado, actuar en el presente y decidir nuestro futuro.
https://f.mtr.cool/rmtlovtkhl
SPOTIFY
https://f.mtr.cool/xmklpomoxx
YouTube
https://f.mtr.cool/wzmnxfejkc
#caminataalorigen #Deepwalk #Ecuador
ノードn_ijからkステップの遷移確率行列A^kでノードの特徴量を、学習し、グローバルな構造情報を獲得する手法GraRepを提案。
設定したステップ毎に行列分解し、concatしてKステップ毎の情報を明示的に保持。DeepWalkはKステップなど同じ空間に同じ空間に射影するのが問題と指摘し、クラスタリングの精度で提案手法は勝ると主張。比較がフェアになるように、DeepWalkにおいてエッジの重みでサンプルしたE-SGNSを提案し比較。
GraRep: Learning Graph Representations with Global Structural Information
設定したステップ毎に行列分解し、concatしてKステップ毎の情報を明示的に保持。DeepWalkはKステップなど同じ空間に同じ空間に射影するのが問題と指摘し、クラスタリングの精度で提案手法は勝ると主張。比較がフェアになるように、DeepWalkにおいてエッジの重みでサンプルしたE-SGNSを提案し比較。
GraRep: Learning Graph Representations with Global Structural Information
December 30, 2025 at 10:52 PM
ノードn_ijからkステップの遷移確率行列A^kでノードの特徴量を、学習し、グローバルな構造情報を獲得する手法GraRepを提案。
設定したステップ毎に行列分解し、concatしてKステップ毎の情報を明示的に保持。DeepWalkはKステップなど同じ空間に同じ空間に射影するのが問題と指摘し、クラスタリングの精度で提案手法は勝ると主張。比較がフェアになるように、DeepWalkにおいてエッジの重みでサンプルしたE-SGNSを提案し比較。
GraRep: Learning Graph Representations with Global Structural Information
設定したステップ毎に行列分解し、concatしてKステップ毎の情報を明示的に保持。DeepWalkはKステップなど同じ空間に同じ空間に射影するのが問題と指摘し、クラスタリングの精度で提案手法は勝ると主張。比較がフェアになるように、DeepWalkにおいてエッジの重みでサンプルしたE-SGNSを提案し比較。
GraRep: Learning Graph Representations with Global Structural Information
DeepWalkやLINEは有向性を考えていないので、近接性を捉えた表現をえる手法の提案。論文のタイトルにもある非対称な推移性（Asymmetric Transitivity）とは有向だとu→vが近ければv→uも近いとはいえないよねということ。これがその当時(2016)では考えられていない。ポイントは著者らの考える有向グラフを考慮している近接性が行列で定義できたのでその分解した表現を計算していること。ちなみにnode2vecも同じKDDで発表されている。
Asymmetric Transitivity Preserving Graph Embedding
Asymmetric Transitivity Preserving Graph Embedding
December 28, 2025 at 9:25 PM
DeepWalkやLINEは有向性を考えていないので、近接性を捉えた表現をえる手法の提案。論文のタイトルにもある非対称な推移性（Asymmetric Transitivity）とは有向だとu→vが近ければv→uも近いとはいえないよねということ。これがその当時(2016)では考えられていない。ポイントは著者らの考える有向グラフを考慮している近接性が行列で定義できたのでその分解した表現を計算していること。ちなみにnode2vecも同じKDDで発表されている。
Asymmetric Transitivity Preserving Graph Embedding
Asymmetric Transitivity Preserving Graph Embedding
DeepWalkみたいにランダムウォークとかnode2vecのように探査の方法をDFS/BFSで切り替えるのではなく、コミュニティ内で探査するのか、ランダムウォークなのかを切り替えている論文。nodeを学習するsequenceをどう獲得するかで色々亜種ができるんだな。その点DFS/BFSで切り替えるnode2vecは賢いと思った。
Community Aware Random Walk for Network Embedding
Community Aware Random Walk for Network Embedding
December 27, 2025 at 10:37 PM
DeepWalkみたいにランダムウォークとかnode2vecのように探査の方法をDFS/BFSで切り替えるのではなく、コミュニティ内で探査するのか、ランダムウォークなのかを切り替えている論文。nodeを学習するsequenceをどう獲得するかで色々亜種ができるんだな。その点DFS/BFSで切り替えるnode2vecは賢いと思った。
Community Aware Random Walk for Network Embedding
Community Aware Random Walk for Network Embedding
ネットワーク埋め込みは行列分解だよ論文を読んだ。
「G=(V,E,A)をつかってV→R^dとなるVの要素の特徴を表すd 次元のベクトル(ただし d≪∣V∣)を学習することが目的」というのは言われてみればそう。各手法のアルゴリズムを清書しているのもよかった。
Network Embedding as Matrix Factorization: Unifying DeepWalk, LINE, PTE, and node2vec
「G=(V,E,A)をつかってV→R^dとなるVの要素の特徴を表すd 次元のベクトル(ただし d≪∣V∣)を学習することが目的」というのは言われてみればそう。各手法のアルゴリズムを清書しているのもよかった。
Network Embedding as Matrix Factorization: Unifying DeepWalk, LINE, PTE, and node2vec
December 24, 2025 at 9:52 PM
ネットワーク埋め込みは行列分解だよ論文を読んだ。
「G=(V,E,A)をつかってV→R^dとなるVの要素の特徴を表すd 次元のベクトル(ただし d≪∣V∣)を学習することが目的」というのは言われてみればそう。各手法のアルゴリズムを清書しているのもよかった。
Network Embedding as Matrix Factorization: Unifying DeepWalk, LINE, PTE, and node2vec
「G=(V,E,A)をつかってV→R^dとなるVの要素の特徴を表すd 次元のベクトル(ただし d≪∣V∣)を学習することが目的」というのは言われてみればそう。各手法のアルゴリズムを清書しているのもよかった。
Network Embedding as Matrix Factorization: Unifying DeepWalk, LINE, PTE, and node2vec
LINE論文読んだ。流れ的にはdeepwalk-LINE-node2vecなんだけど。LINEはエッジのサンプリングを重みを使って加重してサンプリングをしていて、学習につかうエッジは1と2次近接性のものだけに絞っている。1ホップ（直接のつながり）と2ホップ（共通の友人がいる関係）までの情報を保存すして、DeepWalkやnode2vecのように、高次のつながり（3、4ホップ）を深く探索して学習するわけでないし、系列ではなく、あくまでエッジを入力として学習している。
LINE: Large-scale Information Network Embedding
LINE: Large-scale Information Network Embedding
December 23, 2025 at 9:03 PM
LINE論文読んだ。流れ的にはdeepwalk-LINE-node2vecなんだけど。LINEはエッジのサンプリングを重みを使って加重してサンプリングをしていて、学習につかうエッジは1と2次近接性のものだけに絞っている。1ホップ（直接のつながり）と2ホップ（共通の友人がいる関係）までの情報を保存すして、DeepWalkやnode2vecのように、高次のつながり（3、4ホップ）を深く探索して学習するわけでないし、系列ではなく、あくまでエッジを入力として学習している。
LINE: Large-scale Information Network Embedding
LINE: Large-scale Information Network Embedding
DeepWalk論文を読んだ。これって、ランダムウォークをつかって対象nodeが出現するsequenceをつくってそのノードの特徴量を得る上に、独立だからwalkerを複数走らせて並列化できる手法なのか。頭いいな。
arxiv.org/abs/1403.6652
arxiv.org/abs/1403.6652
December 22, 2025 at 2:31 PM
DeepWalk論文を読んだ。これって、ランダムウォークをつかって対象nodeが出現するsequenceをつくってそのノードの特徴量を得る上に、独立だからwalkerを複数走らせて並列化できる手法なのか。頭いいな。
arxiv.org/abs/1403.6652
arxiv.org/abs/1403.6652
nodes are embedded into a high-dimensional vector space for downstream processing. State-of-the-art algorithms for these two paradigms, force-directed layouts and random-walk-based contrastive learning (such as DeepWalk and node2vec), have little in [2/5 of https://arxiv.org/abs/2503.23822v1]
April 1, 2025 at 6:20 AM
nodes are embedded into a high-dimensional vector space for downstream processing. State-of-the-art algorithms for these two paradigms, force-directed layouts and random-walk-based contrastive learning (such as DeepWalk and node2vec), have little in [2/5 of https://arxiv.org/abs/2503.23822v1]