https://mott.in/ Recent content on Davide Mottin Hugo -- gohugo.io en-us Mon, 18 Sep 2023 00:00:00 +0200 https://mott.in/projects/gai/ Tue, 01 Jun 2021 00:00:00 +0000 https://mott.in/projects/gai/ Abstract Publications https://mott.in/projects/graph-exploration/ Mon, 10 Oct 2016 00:00:00 +0000 https://mott.in/projects/graph-exploration/ Abstract The increasing interest in social networks, knowledge graphs, protein-interaction, and many other types of networks has raised the question how users can explore such large and complex graph structures easily. In this regard, graph exploration has emerged as a complementary toolbox for graph management, graph mining, or graph visualization in which the user is a first class citizen. Graph exploration combines and expands database, data mining, and machine learning approaches with the user eye on one side and the system perspective on the other. https://mott.in/projects/exemplars/ Sun, 28 Sep 2014 00:00:00 +0000 https://mott.in/projects/exemplars/ Abstract Search engines are continuously employing advanced techniques that aim to capture user intentions and provide results that go beyond the data that simply satisfy the query conditions. Examples include the personalized results, related searches, similarity search, popular and relaxed queries. In this work we introduce a novel query paradigm that considers a user query as an example of the data in which the user is interested. We call these queries «exemplar queries», and claim that they can play an important role in dealing with the information deluge. https://mott.in/projects/kare/ Sat, 01 Jan 2022 00:00:00 +0000 https://mott.in/projects/kare/ Abstract Publications https://mott.in/projects/aavanguard/ Tue, 01 Jun 2021 00:00:00 +0000 https://mott.in/projects/aavanguard/ Abstract Publications https://mott.in/projects/nebula/ Sat, 01 May 2021 00:00:00 +0000 https://mott.in/projects/nebula/ Abstract Publications https://mott.in/projects/graph-similarities/ Wed, 14 Feb 2018 00:00:00 +0000 https://mott.in/projects/graph-similarities/ Abstract Graphs elegantly model a plethora of phenomena, from social interactions, to biological and chemical processes. Without any prior information comparing two graphs can only be done through notion of similarity. However, designing the right similarity for a certain task is a convolotued process which requires domain expertise and graph expertise. We propose self-learned similarity measures which scale to large graphs and are automatically harvested from the graph itself. https://mott.in/publications/publications/ucode/ Mon, 18 Sep 2023 00:00:00 +0200 https://mott.in/publications/publications/ucode/ UCode captures both overlapping and non-overlapping community switching a single parameter. TL;DR A single unified method for community detection in attributed graph that detects both overlapping and non-overlapping communities via a novel contrastive loss function. In this paper: We introduce a new GNN method, UCoDe, for com- munity detection on graphs. We devise a simple effective single score model which leverages state-of-the-art representations; UCoDe features a novel contrastive loss function that promotes both overlapping and non-overlapping communities, thus being the first approach to achieve competitive results across these tasks with a single model. https://mott.in/publications/publications/marigold/ Sat, 26 Aug 2023 00:00:00 +0200 https://mott.in/publications/publications/marigold/ On high-dimensional data, such as physics experiments, MARIGOLD is the scalable answer to compute k-means. TL;DR Marigold delivers the exact result of Lloyd’s algorithm for k-means in at least one order of magnitude less time on high-dimensional data. In this paper: We propose MARIGOLD, an enhanced Lloyd’s algorithm for high-dimensional data. We enhange Lloyds by introducing triangle inequality pruning and a novel stepwise method that approximates the centroid’s distances through compression. https://mott.in/publications/publications/actup/ Tue, 15 Aug 2023 00:00:00 +0200 https://mott.in/publications/publications/actup/ A single method (GDR) can recreate tSNE and UMAP outputs just by changing the normalization. TL;DR One method (GDR) recreate both tSNE and UMAP embedding, since the difference between them is only the normalization. In this paper: We perform the first comprehensive analysis of the dif- ferences between tSNE and UMAP, showing the effect of each algorithmic choice on the embeddings. We theoretically and experimentally show that changing the normalization is a sufficient condition for switching between the two methods. https://mott.in/publications/bookchapters/pkgs/ Thu, 03 Aug 2023 00:00:00 +0200 https://mott.in/publications/bookchapters/pkgs/ TL;DR Personal Knowledge Graph contain information about a single user. Being able to summarize their content is instrumental for both efficiency and usability. Abstract Summarization for personalized KGs has several key challenges that need to be addressed when designing new computational approaches: Data volume. As mentioned above, KGs tend to be massive as they contain facts about the world. Though summarization aims to reduce the size of the data, summarization approaches need to be carefully designed in order to be scalable and process these graphs efficiently. https://mott.in/news/vldb2023/ Thu, 01 Jun 2023 00:00:00 +0000 https://mott.in/news/vldb2023/ Our paper on speeding up Lloyd’s algorithm for k-means on high-dimensional data have been published at VLDB 2023! See you in Vancouver, looking forward to meet colleagues in VLDB after a long time. https://mott.in/publications/publications/evalign/ Tue, 28 Mar 2023 00:00:00 +0200 https://mott.in/publications/publications/evalign/ TL;DR Comprehensive becnhmark evaluation of graph alignment algorithms on general graphs. Our benchmark includes: First, complete, experimental evaluation of nine undirected, unattributed graph alignment algorithms. Carefully tuning of the algorithms hyperparameters based on network size and using the same assignment algorithm. Thorough evaluation of the algorithms on real and synthetic graphs, with different levels and types of noise. Memory and time scalability experiments on the algorithms. An experimental framework for graph alignment with reproducible experiments and available data and code. https://mott.in/news/graph-align/ Fri, 30 Dec 2022 00:00:00 +0000 https://mott.in/news/graph-align/ A prolific year for my (freshly graduated) PhD student Judith Hermanns. A paper on Evaluating graph alignment algorithms accepted for publication at EDBT 2023 An extension of our GRASP published at the prestigious TKDD journal. Thanks to the great work of the coauthors [Judith Hermanns](https://pure.au.dk/portal/en/persons/judith-franziska-hermanns(99033892-1e04-4a9c-81cf-401ee50fd9da), Anton Tsitsulin, Marina Munkhoeva, Alex Bronstein, and Panagiotis Karras. https://mott.in/publications/publications/grasp-tkdd/ Thu, 29 Dec 2022 00:00:00 +0100 https://mott.in/publications/publications/grasp-tkdd/ GRASP has solid roots into spectral theory and functional mappings. The eigenvector of the two graphs are comparable functions. TL;DR We propose GRASP, short for GRaph Alignment through SPectral Signatures, a principled approach towards detecting a good alignment among graphs, grounded on their spectral characteristics, i.e., eigenvalues and eigenvectors of their Laplacian matrices. In this paper, we extend our previously published GRASP as follows: We extensively discuss the modular graph alignment framework and possible enhancements to GRASP. https://mott.in/publications/publications/grab/ Tue, 01 Nov 2022 00:00:00 +0100 https://mott.in/publications/publications/grab/ GraB introduces new datasets for overlapping community detection on heterogeneous, attributed networks TL;DR We introduce GraB, a benchmark for graph clustering with unique characteristics. Our graphs are at the same time heterogeneous, i.e., include different types of nodes and node attributes, and comprise overlapping clusters, i.e., a node may belong to multiple clusters. We empirically show the arduous characteristics of the datasets. In this paper: We propose GraB, a set of benchmark datasets for overlapping community detection on heterogeneous, attributed netwroks. https://mott.in/news/cidr2022/ Tue, 11 Jan 2022 00:00:00 +0000 https://mott.in/news/cidr2022/ Our system idea of Knowledge Graph Exploration has been published at CIDR 2022 within the paper “Knowledge Graph Exploration Systems: are we lost”. We hope we can inspire the research on the area of Knowledge Graph Exploration as we believe there is a huge potential for helping people explore large KGs in disciplines such as health, journalism, engineering, and many more. Thanks to the work of the coauthors Matteo Lissandrini, Torben Bach Pedersen, and Katja Hose. https://mott.in/publications/publications/kgexp-cidr/ Tue, 11 Jan 2022 00:00:00 +0100 https://mott.in/publications/publications/kgexp-cidr/ KG Exploration system architecture. TL;DR This survey expands our preliminary analysis and survey, by providing a more thorough literature review and drafting an idea for a KG exploration system (see image above). In this paper: We first identify the limitations of existing DBMSes in supporting the needs of KG data management (Section 2) We provide provide a summarization of the area of KG exploration (Section 3). https://mott.in/news/2021/ Fri, 31 Dec 2021 00:00:00 +0000 https://mott.in/news/2021/ I typically not praise successes in my life, cause I believe everything I receive is a gift, and not my own accomplishments. However, this year I have so much to be grateful that I cannot avoid looking back and think to how lucky I am. First, I got married. It took me a lot of energy to organize a wedding with so many people in Covid time. However, we managed to do that and to happily get married and do celebrations both in Italy and in Romania! https://mott.in/publications/publications/boosting/ Mon, 01 Nov 2021 00:00:00 +0100 https://mott.in/publications/publications/boosting/ The performance of graph alignment algorithms vary a lot by changing small parts of the computation TL;DR Study graph alignment algorithms as modular, by: Noticing that some parts of the algorithm (e.g., the matching step) can be easily substituted Performing an experimental evaluation showing that the performance vary a lot with some small changes Introducing enhanced versions of each algorithms Opening to the possibility of modularizing graph alignment in a unified framework In this paper: https://mott.in/news/grasp-21/ Mon, 23 Aug 2021 00:00:00 +0000 https://mott.in/news/grasp-21/ First paper from my PhD student Judith Hermanns “GRASP: Graph Alignment through Spectral Signatures” accepted for publication at APWeb-WAIM. Thanks to the work of the coauthors [Judith Hermanns](https://pure.au.dk/portal/en/persons/judith-franziska-hermanns(99033892-1e04-4a9c-81cf-401ee50fd9da), Anton Tsitsulin, Marina Munkhoeva, Alex Bronstein, and Panagiotis Karras. https://mott.in/publications/publications/grasp/ Sat, 21 Aug 2021 00:00:00 +0200 https://mott.in/publications/publications/grasp/ With a few removed edges, REGAL, an alignment method based on localfeatures, fails to correctly align the distorted Karate club graph to the original; GRASPidentifies most of nodes (correctly aligned nodes in green). TL;DR We propose GRASP, short for GRaph Alignment through SPectral Signatures, a principled approach towards detecting a good alignment among graphs, grounded on their spectral characteristics, i.e., eigenvalues and eigenvectors of their Laplacian matrices. https://mott.in/news/vldb2021/ Mon, 16 Aug 2021 00:00:00 +0000 https://mott.in/news/vldb2021/ After some years, I come back to VLDB with this incredible paper “FREDE: Anytime Graph Embeddings” accepted for publication at VLDB, one of top venues for the Database community. Thanks to the work of the coauthors Anton Tsitsulin, Marina Munkhoeva, Panagiotis Karras, Ivan Oseledets and Emmanuel Müller. https://mott.in/publications/publications/sofos/ Fri, 25 Jun 2021 00:00:00 +0200 https://mott.in/publications/publications/sofos/ TL;DR Perform analytical (e.g., aggregate) queries over large knowledge graph (RDF) data. We present a system, SOFOS that Addresses the problem of providing fast query answering for analytical queries on KGs Provides a generic solution to be deployed on any RDF triple store with SPARQL query processing Highlights possible limitations of six alternative approaches. Abstract Analytical queries over RDF data are becoming prominent as a result of the proliferation of knowledge graphs. https://mott.in/news/www2021/ Mon, 19 Apr 2021 00:00:00 +0000 https://mott.in/news/www2021/ An incredible experience, “COLT: Few-shot Knowledge Validation using Rules” accepted for publication at TheWebConf, main venue for web technologies and graph mining. Thanks to the work of the coauthors Michael Loster, Paolo Papotti, Jan Ehmueller, Benjamin Feldmann and Felix Naumann. https://mott.in/courses/2021-2022/dm-2022/ Mon, 01 Feb 2021 00:00:00 +0100 https://mott.in/courses/2021-2022/dm-2022/ No description for short course - only external URL https://mott.in/publications/publications/colt/ Sat, 16 Jan 2021 00:00:00 +0100 https://mott.in/publications/publications/colt/ How can be sure of a confidence of a rule in a knowledge graph? For more information about COLT and our free dataset, please consult this page TL;DR Validate both confidence and quality a rule has on the facts of a knowledge graph with: A flexible model that use the power of the crowd Only few interactions with a user that validates the knowledge A method that finds the true confidence of a rule The largest dataset of manually annotated rules ever created - Link In this paper: https://mott.in/publications/publications/frede/ Sat, 16 Jan 2021 00:00:00 +0100 https://mott.in/publications/publications/frede/ FREDE scalably produces an embedding at anytime; at the dotted black line, it outperforms all contenders, including the SVD of a PPR-like similarity matrix, after pro-cessing about 20% of matrix rows. TL;DR First anytime optimal algorithm for graph embedding with state-of-the-art performance after visiting a fraction of the matrix. In this paper: We interpret a state-of-the-art graph embedding method, VERSE, as factorizing a transformed PPR similarity matrix We propose FREDE, ananytime graph embedding algorithm that minimizes covariance error on that PPR-like matrix via sketching We attain space complexity linear in the number ofn odes and time linear in the number of processed rows In a thorough experimental evaluation with real graphs we confirm that FREDE is competitive against the state-of-the-art and scales to large networks https://mott.in/publications/publications/fasth/ Thu, 10 Dec 2020 00:00:00 +0100 https://mott.in/publications/publications/fasth/ FastH reduces the time for matrix inversion up to 2.7x compared with the fastest sequential method implemented in PyTorch’. TL;DR We propose a parallel algorithm that exploits GPU power to parametrize orthogonal Householder matrices used for the SVD. Various expensive matrix operations enjoy a substantial speedup using our approach. In this paper: We introduce a novel algorithm, FastH, which increases core utilization, leaving less cores to run idle. https://mott.in/news/neurips-20/ Mon, 28 Sep 2020 00:00:00 +0000 https://mott.in/news/neurips-20/ I am very excited to share our new paper “What if neural networks had SVDs?" has been accepted for spotlight presentation at NeurIPS, one of top machine learning conferences held (sadly) online due Covid. Thanks to the outstanding work of brilliant PhD and MSc students Alexander Mathiasen, Frederik Hvilshøj, Jakob Rødsgaard Jørgensen, and Anshul Nasery. https://mott.in/news/sigweb/ Tue, 28 Jul 2020 00:00:00 +0000 https://mott.in/news/sigweb/ Our survey paper “Knowledge graph exploration: where are we and where are we going?" has just been published on SIGWEB. Thanks to the work of the coauthors Matteo Lissandrini, Torben Bach Pedersen, and Katja Hose. https://mott.in/publications/publications/kgexp/ Mon, 27 Jul 2020 00:00:00 +0200 https://mott.in/publications/publications/kgexp/ Our taxonomy of Knowledge Graph Exploration techniques. TL;DR In this survey we look at what techniques assist users in exploring large knowledge graphs. Knowledge graph exploration entails the process of finding information in knowledge graphs without knowing what we are looking for. In this paper: We propose a taxonomy for knowledge graph exploration techniques (see Figure above). We categorize the available work in three large categories: Summarization & Profiling, Exploratory analytics, and Exploratory Search. https://mott.in/news/iclr2020/ Thu, 30 Apr 2020 00:00:00 +0000 https://mott.in/news/iclr2020/ Our paper “The Shape of Data: Intrinsic Distance for Data Distributions” has been accepted for poster presentation at ICLR 2020, a top machine learning conference held this year at Addis Ababa. Thanks to the work of the coauthors Anton Tsitsulin, Marina Munkhoeva, Panagiotis Karras, Alex Bronstein, Ivan Oseledets and Emmanuel Müller. https://mott.in/news/www2020/ Thu, 30 Apr 2020 00:00:00 +0000 https://mott.in/news/www2020/ Our short paper “Graph-Query Suggestions for Knowledge Graph Exploration” has been accepted for oral presentation at The Web Conference 2020, a top web and data mining conference held this year at Taipei. Thanks to the work of the coauthors Matteo Lissandrini, Themis Palpanas, and Yannis Velegrakis. https://mott.in/publications/publications/imd/ Thu, 30 Apr 2020 00:00:00 +0200 https://mott.in/publications/publications/imd/ TL;DR We propose a metric for comparing data distributions based on their geometry while not relying on any positional information. In this paper: We start out from the observation that models capturing the multi-scale nature of the data manifold by utilizing higher distribution moment matching perform consistently better than their single-scale; We propose IMD, an Intrinsic Multi-scale Distance that is able to compare distributions using only intrinsic information about the data; We provide an efficient approximation thereof that renders computational complexity nearly linear; We empirically demonstrate that IMD effectively quantifies change in model representations; We use IMD to assess the sample quality of GANs and provide reliable insights into the layer-wise output dynamics of neural networks https://mott.in/publications/publications/int-exq/ Mon, 20 Apr 2020 00:00:00 +0200 https://mott.in/publications/publications/int-exq/ Two interactions in the graph query formulation process for ‘Einstein Academic Education’. TL;DR We consider the task of exploratory search through graph queries on knowledge graphs. We propose to assist the user by expanding the query with intuitive suggestions that can retrieve more detailed and relevant answers. In this paper: We formally define the problem of Suggesting Graph-Query Expansions (Section 3) and provide a model based on intuitions from language-modelling and relevance feedback. https://mott.in/news/seadata-19/ Mon, 30 Mar 2020 00:00:00 +0000 https://mott.in/news/seadata-19/ I am proud to announce that this year Matteo Lissandrini (Aaalborg University), Yannis Velegrakis (Utrecht University) and I are organizing the first Search, Exploration, and Analysis in Heterogeneous Datastores Workshop at EDBT/ICDT 2020. Outline The SEA Data workshop will provide a forum for researchers and practitioners to exchange ideas, results, and visions on challenges in data management, information extraction, exploration, and analysis of heterogeneous data and multiple data models at once, such as data lakes, polystores, knowledge bases and knowledge graphs. https://mott.in/news/turin-visit/ Mon, 17 Feb 2020 00:00:00 +0000 https://mott.in/news/turin-visit/ I will visit Francesco Bonchi at the ISI research center February 17-28. I’m looking forward to collaborations and discussion in the amazing research lab there. https://mott.in/courses/2020-2021/dm-2021/ Sat, 01 Feb 2020 00:00:00 +0100 https://mott.in/courses/2020-2021/dm-2021/ No description for short course - only external URL https://mott.in/news/icdm2019/ Fri, 08 Nov 2019 00:00:00 +0000 https://mott.in/news/icdm2019/ Our paper “Personalized Knowledge Graph Summarization: From the Cloud to Your Pocket” has been accepted for oral presentation (9% acceptance rate) at ICDM 2019, a top data mining conference. Thanks to the work of the coauthors Tara Safavi, Caleb Belth, Lukas Faber, Emmanuel Müller, and Danai Koutra. https://mott.in/news/itu-talk/ Wed, 06 Nov 2019 00:00:00 +0000 https://mott.in/news/itu-talk/ I will be given a talk at the IT University of Copenhagen about “The user and the machine: Blending database and machine learning for smarter devices.” Abstract: Can a system discover what a user wants circumventing complex search languages? Can a system answer future enquiries without accessing the data again? In this talk, I will introduce two recent results that build upon machine learning and database techniques to learn user preferences as fast as possible with the fewest number of interactions with the user. https://mott.in/publications/publications/glimpse/ Fri, 16 Aug 2019 00:00:00 +0200 https://mott.in/publications/publications/glimpse/ Is it possible to find a summary of a knowledge graph which reflects the user preferences? TL;DR First summarization method for knowledge graphs workload- and space-aware. In this paper: We summarize a knowledge graph on a personalized basis inferring preferences from user-search or queries. We formulate the problem as maximizing the likelihood of answering user queries given the summary and a space constraint. We reduce the problem to that of a maximum coverage. https://mott.in/publications/publications/mfasc/ Sun, 11 Aug 2019 00:00:00 +0200 https://mott.in/publications/publications/mfasc/ Can the system help the user in the decision? TL;DR Efficient interactive method to learn user preferences in an active way: Fast – Only 10 steps to build a model of the user Multiple information – incorporate shallow preferences from the system to help the user Flexible – recommendations for multi-dimensional and graph data In this paper: A novel active search formulation that fuses continuous user scores with correlated computationally derived scores Experiments on synthetic data, in which MF-ASC outperforms multifidelity methods for function optimization Two real case-studies on tabular consumer ratings and information graphs. https://mott.in/publications/others/frede-dlg/ Tue, 06 Aug 2019 00:00:00 +0200 https://mott.in/publications/others/frede-dlg/ Abstract Low-dimensional representations, or embeddings, of a graph’s nodes facilitate tasks such as community detection, link prediction, and classification. Embeddings learn similarities among nodes, either implicitly, as a side-effect of adapting word embedding methods to graphs, or explicitly, by reconstructing a similarity measure. As the similarity matrix is quadratic, past research has resorted to heuristic or linear-factorization solutions, compromising quality. In this paper we develop FREDE (FRequent Directions Embeddings), an incremental algorithm that embeds a nonlinear transformation of the similarity matrix optimally in terms of an objective relaxed from the optimal low-rank approximation given by SVD. https://mott.in/news/kdd2019/ Sun, 04 Aug 2019 00:00:00 +0000 https://mott.in/news/kdd2019/ Our paper “Figuring out the User in a Few Steps: Bayesian Multifidelity Active Search with Cokriging” has been accepted for oral presentation (9% acceptance rate) at KDD 2019, the top data mining conference. Thanks to the work of the coauthors Nikita Klyuchnikov, Georgia Koutrika, Emmanuel Müller, and Panagiotis Karras. https://mott.in/news/tutorial/ Sun, 30 Jun 2019 00:00:00 +0000 https://mott.in/news/tutorial/ Our tutorial “Exploring the Data Wilderness through Examples” has been accepted for presentation at SIGMOD 2019. A more practical and IR version of the tutorial “Example-driven Search: a New Frontier for Exploratory Search” will be presented at SIGIR 2019, the top-tier conference in Information Retrieval. I will present the tutorial with Matteo Lissandrini, Yannis Velegrakis, and Themis Palpanas. Looking forward to presenting in Amsterdam and Paris! https://mott.in/courses/2018-2019/example-based/ Mon, 20 May 2019 00:00:00 +0200 https://mott.in/courses/2018-2019/example-based/ No description for short course - only external URL https://mott.in/news/eigenestelle/ Wed, 01 May 2019 00:00:00 +0000 https://mott.in/news/eigenestelle/ The Deutsche FoschungsGemainschaft (DFG), the German Federal Funding agency, awarded 330.000€ as a personal grant (eigenestelle) for a project on “Query correction in large knowledge graphs”. Thanks DFG! Sadly, the grant cannot be received as I moved to Denmark, but I hope there will be other future fruitful projects with the German agency. https://mott.in/publications/tutorials/expl-methods-sigir/ Wed, 17 Apr 2019 00:00:00 +0000 https://mott.in/publications/tutorials/expl-methods-sigir/ TL;DR This tutorial explores Methods for exploring large datasets using examples Algorithmic solutions to search without query languages Interactive methods and user-in-the-loop feedback Machine learning for adaptive, online methods Prototypes of exploration systems Abstract Exploration is one of the primordial ways to accrue knowledge about the world and its nature. As we accumulate, mostly automatically, data at unprecedented volumes and speed, our datasets have become complex and hard to understand. https://mott.in/publications/publications/beyond-journal/ Wed, 30 Jan 2019 11:10:23 +0200 https://mott.in/publications/publications/beyond-journal/ Can we find personalized graph patterns? TL;DR Pattern mining in graphs with preferences (weighted) Antimonotonic scoring functions for weighted graphs Mine patterns in multi-weight graphs (e.g., Amazon) Exact and approximate solutions for the problem Extensive evaluation on real graphs [New!] Distributed algorithm for multi-weighted graphs and evaluation Abstract Graph pattern mining aims at identifying structures that appear frequently in large graphs, under the assumption that frequency signifies importance. https://mott.in/news/dapd-19/ Wed, 30 Jan 2019 00:00:00 +0000 https://mott.in/news/dapd-19/ Our paper “Mining Patterns in Graphs with Multiple Weights” has been accepted for publication in the Distributed and Parallel Databases journal. Thanks to my coauthors Giulia Preti, Matteo Lissandrini, and Yannis Velegrakis! https://mott.in/news/amazon/ Sun, 20 Jan 2019 00:00:00 +0000 https://mott.in/news/amazon/ A work in collaboration with Danai Koutra from the University of Michigan has been selected for the Amazon research award! The award is granted to continue our work on novel knowledge graph summaries. https://mott.in/courses/2018-2019/db-2019/ Sun, 20 Jan 2019 00:00:00 +0100 https://mott.in/courses/2018-2019/db-2019/ https://mott.in/courses/2019-2020/ml-2019/ Sun, 20 Jan 2019 00:00:00 +0100 https://mott.in/courses/2019-2020/ml-2019/ https://mott.in/courses/2018-2019/adma-2018/ Sat, 19 Jan 2019 00:00:00 +0100 https://mott.in/courses/2018-2019/adma-2018/ No description for short course - only external URL https://mott.in/courses/2019-2020/adma-2019/ Sat, 19 Jan 2019 00:00:00 +0100 https://mott.in/courses/2019-2020/adma-2019/ No description for short course - only external URL https://mott.in/courses/2019-2020/dm-2020/ Sat, 19 Jan 2019 00:00:00 +0100 https://mott.in/courses/2019-2020/dm-2020/ No description for short course - only external URL https://mott.in/publications/tutorials/expl-methods-sigmod/ Mon, 07 Jan 2019 00:00:00 +0000 https://mott.in/publications/tutorials/expl-methods-sigmod/ TL;DR This tutorial explores Methods for exploring large datasets using examples Algorithmic solutions to search without query languages Interactive methods and user-in-the-loop feedback Machine learning for adaptive, online methods Abstract Exploration is one of the primordial ways to accrue knowledge about the world and its nature. As we accumulate, mostly automatically, data at unprecedented volumes and speed, our datasets have become complex and hard to understand. In this context exploratory search provides a handy tool for progressively gather the necessary knowledge by starting from a tentative query that hopefully leads to answers at least partially relevant and that can provide cues about the next queries to issue. https://mott.in/publications/books/exploration/ Tue, 01 Jan 2019 00:00:00 +0100 https://mott.in/publications/books/exploration/ Example-based approaches and book outline TL;DR A lecture-style book on example-based approaches For exploratory tasks using examples instead of queries to retrieve data Connecting the different works in the area Providing novel insights to the use of machine learning for user understanding Highlighting visionary research directions on the area of example-based tasks and learning Abstract Data usually comes in a plethora of formats and dimensions, rendering the exploration and information extraction processes challenging. https://mott.in/news/port/ Mon, 17 Dec 2018 00:00:00 +0000 https://mott.in/news/port/ Thanks to Pedro Manuel Pinto I had the honor to be a speaker Graph Exploration in the Porto Winter School on Network Science. Over 80 attendees of which 43 Students (28 PhD, 11 MSc, 4 BSc), 12 University Professors, 8 PostDoc/Researchers (Academic), 12 Data Scientists (Private Company), and 5 Engineers/Developers (Private Company) have been exposed to different topics on the broad (and expanding) area of Network Science. The event featured experts in different disciplines for a full-three day immersion in this beautiful topic at the intersection of physics, math, computer science, and statistics. https://mott.in/news/book/ Sat, 01 Dec 2018 00:00:00 +0000 https://mott.in/news/book/ Our book on Data Exploration Using Example-based Methods has been published by Morgan & Claypool! This book was an incredible effort that wouldn’t have been possible without the work of my coauthors Matteo Lissandrini, Yannis Velegrakis, and Themis Palpanas. Our book is out! https://mott.in/news/professor/ Mon, 01 Oct 2018 00:00:00 +0000 https://mott.in/news/professor/ I’m very excited to start as a professor in the department of Computer Science of Aarhus University. Looking forward to the start of the new adventure and the new challenges ahead. https://mott.in/news/marathon/ Sun, 16 Sep 2018 00:00:00 +0000 https://mott.in/news/marathon/ There is little of my academic endeavors in this story, but completing a marathon with its 42.195 km is something that makes me feel proud. This year has been a very successful part of my life and I think that dreams whatever and wherever they are should be pursued. Thanks all who supported me (especially Thijs and Anna)! After months of training and 42km, finally the end That’s happiness https://mott.in/publications/publications/x2q/ Sat, 01 Sep 2018 00:00:00 +0200 https://mott.in/publications/publications/x2q/ TL;DR Interactive exploration of large knowledge graphs Interactive – Starts from a single entity and guides the user towards the results Expressive – explore the results with Exemplar Queries [1] and Multiple Exemplar [2] Fast – takes less than 1s to get the results In this demontration paper: We showcase \(\tt X_2Q\) a search-and-explore interactie system to explore large knowledge graphs We demonstrate the effectiveness of the exemplar query paradigm We remove the need of complex query languages and natural language ambiguity We suggest and guide the user towards the answer easily and fast Search-and-explore with X2Q https://mott.in/publications/others/adgraphsum/ Fri, 24 Aug 2018 11:27:52 +0200 https://mott.in/publications/others/adgraphsum/ Can we compress a graph based on user queries? Abstract Knowledge graphs, which are rich networks of entities and concepts connected via multiple types of relationships, have gained traction as powerful structures for natural language understanding and question answering. Although recent research efforts have started to address efficient querying and storage of knowledge graphs, such methods are neither user-driven nor flexible to changes in the data, both of which are important in the real world. https://mott.in/publications/others/sgr/ Fri, 24 Aug 2018 11:26:37 +0200 https://mott.in/publications/others/sgr/ https://mott.in/publications/tutorials/graph-exploration-kdd/ Fri, 24 Aug 2018 00:00:00 +0000 https://mott.in/publications/tutorials/graph-exploration-kdd/ Abstract The increasing interest in social networks, knowledge graphs, protein-interaction, and many other types of networks has raised the question how users can explore such large and complex graph structures easily. Current tools focus on graph management, graph mining, or graph visualization but lack user-driven methods for graph exploration.In many cases graph methods try to scale to the size and complexity of a real network. However, methods miss user requirements such as exploratory graph query processing, intuitive graph explanation, and interactivity in graph exploration. https://mott.in/publications/publications/netlsd/ Fri, 24 Aug 2018 00:00:00 +0200 https://mott.in/publications/publications/netlsd/ How can we model the structure of this graph on multiple scales? TL;DR Fast graph descriptors that allows to compare graphs: Fast – in \(O(1)\) after fast precomputation On multiple scales – from connectivity to community structure Of different sizes – we can correct for size of graphs In this paper: We take the geometrical perspective for computing the descriptors. We show how to compute them fast for million-scale graphs. https://mott.in/publications/publications/verse/ Wed, 25 Apr 2018 00:00:00 +0200 https://mott.in/publications/publications/verse/ Can a single model capture three highlighted node properties? TL;DR State-of-the-art graph embedding algorithms are neural similarity matrix compression algorithms. In this paper: We take the perspective of node similarities for node embedding. We show how previous works (DeepWalk, node2vec, LINE) implicitly utilize similarities. We develop an algorithm (VERSE) to efficiently preserve similarities. We create two VERSE versions: full that uses all node-to-node information, and sampled one that runs in linear time, yet provably converges to the full variant. https://mott.in/publications/publications/metaexp/ Sun, 15 Apr 2018 18:57:23 +0200 https://mott.in/publications/publications/metaexp/ TL;DR Metaexp is a system that allows Definition of arbitrary similarities on large knowledge graphs Interactive exploration of knowledge graphs through embeddings Explanation of similarities through metapaths Working with complex domains like biological networks What is similar in a knowledge graph? People Freya Behrens Sebastian Bischoff Pius Ladenburger Julius Rückin Laurenz Seidel Fabian Stolp Michael Vaichenker Adrian Ziegler Davide Mottin Emmanuel Müller in collaboration with Martin Preusse (Computational Cell Maps, Institute of Computational Biology, Helmholtz Zentrum München, Germany / Knowing) Nikola Müller (Computational Cell Maps, Institute of Computational Biology, Helmholtz Zentrum München, Germany / Knowing) Michael Hunger (neo4j) https://mott.in/publications/publications/multi-exemplar/ Sun, 01 Apr 2018 18:58:34 +0200 https://mott.in/publications/publications/multi-exemplar/ Can we find results without query languages? TL;DR Expressive queries based on multiple fragments of examples Formalization of multi-exemplar queries Allows combinations of non-homomorhphic examples Efficient exact and top-k methods that scale to large graphs Adapts to generic ranking functions Scale to graphs with millions of nodes and edges Abstract Graph pattern mining aims at identifying structures that appear frequently in large graphs, under the assumption that frequency signifies importance. https://mott.in/publications/publications/notable/ Tue, 20 Mar 2018 18:59:51 +0200 https://mott.in/publications/publications/notable/ How are presidents different? TL;DR Easy comparison of nodes in knowledge graphs Given a set of seed nodes find the notable characteristics Effective computation of metapaths to find context nodes given seed nodes Probabilistic approach to discover notable characteristics Expreimental evaluation through a user study on YAGO and LMDB Abstract Query answering routinely employs knowledge graphs to assist the user in the search process. Given a knowledge graph that represents entities and relationships among them, one aims at complementing the search with intuitive but effective mechanisms. https://mott.in/publications/publications/beyond/ Sat, 17 Mar 2018 11:10:23 +0200 https://mott.in/publications/publications/beyond/ Can we find personalized graph patterns? TL;DR Pattern mining in graphs with preferences (weighted) Antimonotonic scoring functions for weighted graphs Mine patterns in multi-weight graphs (e.g., Amazon) Exact and approximate solutions for the problem Extensive evaluation on real graphs Abstract Graph pattern mining aims at identifying structures that appear frequently in large graphs, under the assumption that frequency signifies importance. Several measures of frequency have been proposed that respect the apriori property, pivotal to an efficient search of the patterns. https://mott.in/publications/bookchapters/bigdata/ Sat, 17 Feb 2018 11:38:39 +0200 https://mott.in/publications/bookchapters/bigdata/ TL;DR The research on graph exploration has revolved around three main pillars: keyword graph queries, exploratory graph analysis, and refinement of query results. Abstract Exploratory methods have been proposed as a mean to extract knowledge from relational data without knowing what to search (Idreos et al. 2015). Graph exploration has been introduced to perform exploratory analyses on graph-shaped data (Mottin and Müller 2017). Graph exploration aims at mitigating the access to the data to the user, even if such user is a novice. https://mott.in/courses/2017-2018/bp-2017/ Wed, 01 Nov 2017 00:00:00 +0100 https://mott.in/courses/2017-2018/bp-2017/ https://mott.in/courses/2017-2018/gm-2017/ Wed, 01 Nov 2017 00:00:00 +0100 https://mott.in/courses/2017-2018/gm-2017/ No description for short course - only external URL https://mott.in/courses/2017-2018/srs-2018/ Wed, 01 Nov 2017 00:00:00 +0100 https://mott.in/courses/2017-2018/srs-2018/ https://mott.in/publications/tutorials/expl-methods/ Fri, 01 Sep 2017 00:00:00 +0000 https://mott.in/publications/tutorials/expl-methods/ TL;DR This tutorial explores Methods for exploring large datasets using examples Algorithmic solutions to search without query languages Interactive methods and user-in-the-loop feedback Machine learning for adaptive, online methods Abstract Data usually comes in a plethora of formats and dimensions, rendering the exploration and information extraction processes cumbersome. Thus, being able to cast exploratory queries in the data with the intent of having an immediate glimpse on some of the data properties is becoming crucial. https://mott.in/publications/tutorials/graph-exploration-sigmod/ Sat, 24 Jun 2017 00:00:00 +0000 https://mott.in/publications/tutorials/graph-exploration-sigmod/ Abstract The increasing interest in social networks, knowledge graphs, protein-interaction, and many other types of networks has raised the question how users can explore such large and complex graph structures easily. Current tools focus on graph management, graph mining, or graph visualization but lack user-driven methods for graph exploration.In many cases graph methods try to scale to the size and complexity of a real network. However, methods miss user requirements such as exploratory graph query processing, intuitive graph explanation, and interactivity in graph exploration. https://mott.in/courses/2016-2017/srs-2017/ Wed, 01 Mar 2017 00:00:00 +0100 https://mott.in/courses/2016-2017/srs-2017/ https://mott.in/courses/2016-2017/gm-2016/ Tue, 01 Nov 2016 00:00:00 +0100 https://mott.in/courses/2016-2017/gm-2016/ No description for short course - only external URL https://mott.in/publications/publications/exemplar-journal/ Sun, 17 Jul 2016 11:12:31 +0200 https://mott.in/publications/publications/exemplar-journal/ Is A1 an answer to Q2? TL;DR Extend exemplar queries with Strong simulation1 for finding answers Extensive experiments on real datasets Related publications Exemplar Queries: Give me an Example of What you Need Ma, S., Cao, Y., Fan, W., Huai, J. and Wo, T., 2014. Strong simulation: Capturing topology in graph pattern matching. ACM Transactions on Database Systems (TODS), 39(1), p. https://mott.in/publications/publications/empty-journal/ Fri, 17 Jun 2016 11:13:46 +0200 https://mott.in/publications/publications/empty-journal/ How can we guide the user towards the answer? TL;DR Extend Empty-answer Framework with Faster and more accurate approximate algorithm Complexity analysis Top-\(k\) query refinements models Experiments, code and algorithms Related publications A Probabilistic Optimization Framework for the Empty-Answer Problem https://mott.in/courses/2015-2016/mp-2016/ Tue, 01 Mar 2016 00:00:00 +0100 https://mott.in/courses/2015-2016/mp-2016/ No description for short course - only external URL https://mott.in/publications/publications/gqref/ Mon, 17 Aug 2015 11:15:23 +0200 https://mott.in/publications/publications/gqref/ How can we find convenient reformulations? TL;DR Refinement of graph queries with \(\frac{1}{2}\)-approximate algorithm Informative reformulations with coverage and diversity Fast exact branch-and-bound heuristic for marginal gain Multiple experiments on real and synthetic datasets Abstract We study a problem of graph-query reformulation enabling explorative query-driven discovery in graph databases. Given a query issued by the user, the system, apart from returning the result patterns, also proposes a number of specializations (i. https://mott.in/home/home/ Sun, 28 Sep 2014 00:00:00 +0000 https://mott.in/home/home/ Introduction https://mott.in/publications/publications/unleashing/ Tue, 16 Sep 2014 18:43:55 +0200 https://mott.in/publications/publications/unleashing/ Abstract Information graphs are generic graphs that model dif- ferent types of information through nodes and edges. Knowledge graphs are the most common type of information graphs in which nodes represent entities and edges represent relationships among them. In this paper, we argue that exploitation of information graphs can lead into novel query answering capabilities that go beyond the existing capabilities of keyword search, and focus on one of them, namely, exemplar queries. https://mott.in/publications/publications/empty/ Mon, 01 Sep 2014 11:17:36 +0200 https://mott.in/publications/publications/empty/ How can we guide the user towards the answer? TL;DR Interactive framework for suggesting changes on user queries, includes Multiple objectives (e.g, user effort, revenue) Exhaustive, fast, and approximate solutions Probabilistic, interactive formulations with black-box preference and knowledge functions User-study and extensive experiments on real data Abstract We propose a principled optimization-based interactive query relaxation framework for queries that return no answers. Given an initial query that returns an empty answer set, our framework dynamically computes and suggests alternative queries with less conditions than those the user has initially requested, in order to help the user arrive at a query with a non-empty answer, or at a query for which no matter how many additional conditions are ignored, the answer will still be empty. https://mott.in/publications/publications/exemplar/ Mon, 01 Sep 2014 00:00:00 +0200 https://mott.in/publications/publications/exemplar/ What if I know only an example of IT-companies acquisition? TL;DR Novel query paradigm that allows Finding answers to unknown query given an example Scales to Freebase1, the largest available knowledge graph Exhaustive and top-k approximate answers Answers in less than 1s Abstract Search engines are continuously employing advanced techniques that aim to capture user intentions and provide results that go beyond the data that simply satisfy the query conditions. https://mott.in/publications/publications/xq/ Tue, 17 Jun 2014 11:21:15 +0200 https://mott.in/publications/publications/xq/ XQ interface, what is similar to Google Youtube? Abstract We demonstrate XQ, a query engine that implements a novel technique for searching relevant information on the web and in various data sources, called Exemplar Queries. While the traditional query model expects the user to provide a set of specifications that the elements of interest need to satisfy, XQ expects the user to provide only an element of interest and we infer the desired answer set based on that element. https://mott.in/publications/publications/iqr/ Tue, 17 Jun 2014 11:20:29 +0200 https://mott.in/publications/publications/iqr/ Abstract We present IQR, a system that demonstrates optimization based interactive relaxations for queries that return an empty answer. Given an empty answer, IQR dynamically suggests one relaxation of the original query conditions at a time to the user, based on certain optimization objectives, and the user responds by either accepting or declining the relaxation, until the user arrives at a non-empty answer, or a non-empty answer is impossible to achieve with any further relaxations. https://mott.in/publications/publications/entrank/ Thu, 17 Oct 2013 11:22:34 +0200 https://mott.in/publications/publications/entrank/ Log information describing the items the users have selected from the set of answers a query engine returns to their queries constitute an excellent form of indirect user feedback that has been extensively used in the web to improve the effectiveness of search engines. In this work we study how the logs can be exploited to improve the ranking of the results returned by an entity search engine. Entity search engines are becoming more and more popular as the web is changing from a web of documents into a “web of things”. https://mott.in/publications/others/kqgq/ Wed, 17 Jul 2013 11:28:59 +0200 https://mott.in/publications/others/kqgq/ https://mott.in/publications/tutorials/graph-exploration/ Mon, 01 Jan 0001 00:00:00 +0000 https://mott.in/publications/tutorials/graph-exploration/ Abstract The increasing interest in social networks, knowledge graphs, protein-interaction, and many other types of networks has raised the question how users can explore such large and complex graph structures easily. Current tools focus on graph management, graph mining, or graph visualization but lack user-driven methods for graph exploration.In many cases graph methods try to scale to the size and complexity of a real network. However, methods miss user requirements such as exploratory graph query processing, intuitive graph explanation, and interactivity in graph exploration.