together with brief user bio and description of her/his academic activity.

[Link to my homepage](https://sites.google.com/view/danisch/home)
## I will read the following papers.
- [PageRank as a Function of the Damping Factor](https://papers-gamma.link/paper/106)
- [Graph Stream Algorithms: A Survey](https://papers-gamma.link/paper/102)
- [Network Sampling: From Static to Streaming Graphs](https://papers-gamma.link/paper/122)
- [The Protein-Folding Problem, 50 Years On](https://papers-gamma.link/paper/78)
- [Computational inference of gene regulatory networks: Approaches, limitations and opportunitie](https://papers-gamma.link/paper/77)
- [Graph complexity analysis identifies an ETV5 tumor-specific network in human and murine low-grade glioma](https://papers-gamma.link/paper/79)
- [Gene Networks in Plant Biology: Approaches in Reconstruction and Analysis](https://papers-gamma.link/paper/76)
- [The non-convex Burer–Monteiro approach works on smooth semidefinite programs](https://papers-gamma.link/paper/80)
- [Solving SDPs for synchronization and MaxCut problems via the Grothendieck inequality](https://papers-gamma.link/paper/81)
- [Influence maximization in complex networks through optimal percolation](https://papers-gamma.link/paper/70)
- [Motifs in Temporal Networks](https://papers-gamma.link/paper/61)
- [Deep Sparse Rectifier Neural Networks](https://papers-gamma.link/paper/69)
- [Sparse Convolutional Neural Networks](https://papers-gamma.link/paper/67)
- [A fast and simple algorithm for training neural probabilistic language models](https://papers-gamma.link/paper/58)
- [Adding One Neuron Can Eliminate All Bad Local Minima](https://papers-gamma.link/paper/71)
## I read the following papers.
### 2018-2019:
- [Hierarchical Taxonomy Aware Network Embedding](https://papers-gamma.link/paper/116)
- [Billion-scale Network Embedding with Iterative Random Projection](https://papers-gamma.link/paper/110)
- [HARP: Hierarchical Representation Learning for Networks](https://papers-gamma.link/paper/109/)
- [Layered Label Propagation: A MultiResolution Coordinate-Free Ordering for Compressing Social Networks](https://papers-gamma.link/paper/105)
### 2017-2018:
- [Link Prediction in Graph Streams](https://papers-gamma.link/paper/101)
- [The Community-search Problem and How to Plan a Successful Cocktail Party](https://papers-gamma.link/paper/74)
- [A Nonlinear Programming Algorithm for Solving Semidefinite Programs via Low-rank Factorization](https://papers-gamma.link/paper/55)
- [Deep Learning](https://papers-gamma.link/paper/68)
- [Reducing the Dimensionality of Data with Neural Networks](https://papers-gamma.link/paper/65)
- [Representation Learning on Graphs: Methods and Applications](https://papers-gamma.link/paper/60)
- [Improved Approximation Algorithms for MAX k-CUT and MAX BISECTION](https://papers-gamma.link/paper/56)
- [Cauchy Graph Embedding](https://papers-gamma.link/paper/53)
- [Phase Transitions in Semidefinite Relaxations](https://papers-gamma.link/paper/57)
- [Graph Embedding Techniques, Applications, and Performance: A Survey](https://papers-gamma.link/paper/52)
- [VERSE: Versatile Graph Embeddings from Similarity Measures](https://papers-gamma.link/paper/48)
- [Hierarchical Clustering Beyond the Worst-Case](https://papers-gamma.link/paper/45)
- [Scalable Motif-aware Graph Clustering](https://papers-gamma.link/paper/18)
- [Practical Algorithms for Linear Boolean-width](https://papers-gamma.link/paper/40)
- [New Perspectives and Methods in Link Prediction](https://papers-gamma.link/paper/28/New%20Perspectives%20and%20Methods%20in%20Link%20Prediction)
- [In-Core Computation of Geometric Centralities with HyperBall: A Hundred Billion Nodes and Beyond](https://papers-gamma.link/paper/37)
- [Diversity is All You Need: Learning Skills without a Reward Function](https://papers-gamma.link/paper/36)
- [When Hashes Met Wedges: A Distributed Algorithm for Finding High Similarity Vectors](https://papers-gamma.link/paper/23)
- [Fast Approximation of Centrality](https://papers-gamma.link/paper/35/Fast%20Approximation%20of%20Centrality)
- [Indexing Public-Private Graphs](https://papers-gamma.link/paper/19/Indexing%20Public-Private%20Graphs)
- [On the uniform generation of random graphs with prescribed degree sequences](https://papers-gamma.link/paper/26/On%20the%20uniform%20generation%20of%20random%20graphs%20with%20prescribed%20d%20egree%20sequences)
- [Linear Additive Markov Processes](https://papers-gamma.link/paper/21/Linear%20Additive%20Markov%20Processes)
- [ESCAPE: Efficiently Counting All 5-Vertex Subgraphs](https://papers-gamma.link/paper/17/ESCAPE:%20Efficiently%20Counting%20All%205-Vertex%20Subgraphs)
- [The k-peak Decomposition: Mapping the Global Structure of Graphs](https://papers-gamma.link/paper/16/The%20k-peak%20Decomposition:%20Mapping%20the%20Global%20Structure%20of%20Graphs)
- [A Fast and Provable Method for Estimating Clique Counts Using Turán’s Theorem](https://papers-gamma.link/paper/24)

☆

2

### Can be extended to k-cliques and k-motifs:
- To list k-cliques: https://github.com/maxdan94/kClist
- To list k-motifs: https://github.com/maxdan94/kmotif
### Typos:
- ", cf.[23],"
- "conductance problem.Notice that"
- "(CNM) [12] , Cfinder"
- "Louvaine" instead of "Louvain" in Table 2

☆

1

## About scalability:
"We apply a standard cleaning procedure (for similarity) and remove high out-degrees. In other words, if some vertex v has more than 10K followers (outdegree > 10K), we remove
all these edges. (We do not remove the vertex, but rather only its out-edges.) Intuitively, the fact that two vertices are followed by v is not a useful signal for similarity. In
flock and friendster, such vertices are typically spammers and should be ignored. For webgraphs, a page linking to more than 10K other pages is probably not useful for similarity
measurement."
- Typo: should be "more than 10K followeEs (outdegree > 10K)" not "followers".
- Such a cleaning procedure makes sense, in addition, it has a valuable side effect: in practice, after applying such a cleaning procedure, many pairs of nodes with very small non-zero similarity (in the original directed graph) have a similarity of zero (in the new graph with the out-edges of the out-hubs removed).
- Given that such a cleaning procedure is applied, a brute force approach (that computes the similarity of each pair of nodes sharing at least one in-neighbor) is scalable to a large extent.
- The following C code and experiments support the previous claim: https://github.com/maxdan94/cosineSparseMatrix
- If scalability is still an issue, in-edges of nodes with very high in-degree could also be removed.
"In our example, this turns out to be more than 100 trillion triples. This is infeasible even for a large industrial-strength cluster.".
Listing 100 trillion triples seems doable with a medium-strength cluster.
"If a matrix A has a 100 billion non-zeroes, it takes upwards of 1TB just to store the entries. This is more than an order of magnitude of the storage of a commodity machine in a cluster. Any approach of partitioning A into submatrices cannot scale."
Indeed, memory seems to be a problem. However, it is unclear why an approach partitioning A into submatrices cannot scale.
## Dataset:
"The dataset friendster is a social network, and is available from the Stanford Large Network Dataset Collection [28]."
It is unclear which friendster dataset is used. In SNAP, http://snap.stanford.edu/data/com-Friendster.html seems to be "undirected" and has 1.8B edges, not "1.6B edges" as written in Table 1 (even after removing edges of nodes with degree larger than 10,000).
## Typos and minors:
- Inside the paper $O(100B)$ looks a bit strange, because according to [the definition](https://en.wikipedia.org/wiki/Big_O_notation#Formal_definition) the set $O(100B)$ is equal to $O(0.00001)$
- should be "more than 10K followeEs (outdegree > 10K)" not "followers".

## Comments: