no code implementations • 13 Mar 2024 • Gautham Govind Anil, Pascal Esser, Debarghya Ghoshdastidar
We provide the first convergence results of NTK for contrastive losses, and present a nuanced picture: NTK of wide networks remains almost constant for cosine similarity based contrastive losses, but not for losses based on dot product similarity.
no code implementations • 20 Feb 2024 • Alexandru Crăciun, Debarghya Ghoshdastidar
There currently is a significant interest in understanding the Edge of Stability (EoS) phenomenon, which has been observed in neural networks training, characterized by a non-monotonic decrease of the loss function over epochs, while the sharpness of the loss (spectral norm of the Hessian) progressively approaches and stabilizes around 2/(learning rate).
no code implementations • 15 Feb 2024 • Maximilian Fleissner, Leena Chennuru Vankadara, Debarghya Ghoshdastidar
Despite the growing popularity of explainable and interpretable machine learning, there is still surprisingly limited work on inherently interpretable clustering methods.
no code implementations • 21 Jul 2023 • Anurag Singh, Mahalakshmi Sabanayagam, Krikamol Muandet, Debarghya Ghoshdastidar
Adaptive test-time defenses are used to improve the robustness of deep neural networks to adversarial examples.
no code implementations • 24 Feb 2023 • Satyaki Mukherjee, Soumendu Sundar Mukherjee, Debarghya Ghoshdastidar
We consider the general dimensionality reduction problem of locating in a high-dimensional data cloud, a $k$-dimensional non-Gaussian subspace of interesting features.
no code implementations • 2 Dec 2022 • Pascal Mattia Esser, Satyaki Mukherjee, Mahalakshmi Sabanayagam, Debarghya Ghoshdastidar
The central question in representation learning is what constitutes a good or meaningful representation.
1 code implementation • 29 Nov 2022 • Aishik Mandal, Michaël Perrot, Debarghya Ghoshdastidar
Comparison-based learning addresses the problem of learning when, instead of explicit features or pairwise similarities, one only has access to comparisons of the form: \emph{Object $A$ is more similar to $B$ than to $C$.}
no code implementations • 3 Nov 2022 • Luca Rendsburg, Leena Chennuru Vankadara, Debarghya Ghoshdastidar, Ulrike Von Luxburg
Regression on observational data can fail to capture a causal relationship in the presence of unobserved confounding.
1 code implementation • 18 Oct 2022 • Mahalakshmi Sabanayagam, Pascal Esser, Debarghya Ghoshdastidar
The fundamental principle of Graph Neural Networks (GNNs) is to exploit the structural information of the data by aggregating the neighboring nodes using a `graph convolution' in conjunction with a suitable choice for the network architecture, such as depth and activation functions.
no code implementations • 18 Feb 2022 • Leena Chennuru Vankadara, Luca Rendsburg, Ulrike Von Luxburg, Debarghya Ghoshdastidar
If the confounding strength is negative, causal learning requires weaker regularization than statistical learning, interpolators can be optimal, and the optimal regularization can even be negative.
no code implementations • NeurIPS 2021 • Pascal Mattia Esser, Leena Chennuru Vankadara, Debarghya Ghoshdastidar
While VC Dimension does result in trivial generalisation error bounds in this setting as well, we show that transductive Rademacher complexity can explain the generalisation properties of graph convolutional networks for stochastic block models.
1 code implementation • 18 Nov 2021 • Leena Chennuru Vankadara, Philipp Michael Faller, Michaela Hardt, Lenon Minorics, Debarghya Ghoshdastidar, Dominik Janzing
Under causal sufficiency, the problem of causal generalization amounts to learning under covariate shifts, albeit with additional structure (restriction to interventional distributions under the VAR model).
no code implementations • 18 Oct 2021 • Leena Chennuru Vankadara, Sebastian Bordt, Ulrike Von Luxburg, Debarghya Ghoshdastidar
Despite the ubiquity of kernel-based clustering, surprisingly few statistical guarantees exist beyond settings that consider strong structural assumptions on the data generation process.
no code implementations • 8 Oct 2021 • Mahalakshmi Sabanayagam, Pascal Esser, Debarghya Ghoshdastidar
This paper focuses on semi-supervised learning on graphs, and explains the above observations through the lens of Neural Tangent Kernels (NTKs).
1 code implementation • ICLR 2022 • Mahalakshmi Sabanayagam, Leena Chennuru Vankadara, Debarghya Ghoshdastidar
Using the proposed graph distance, we present two clustering algorithms and show that they achieve state-of-the-art results.
no code implementations • NeurIPS 2020 • Michaël Perrot, Pascal Mattia Esser, Debarghya Ghoshdastidar
The goal of clustering is to group similar objects into meaningful partitions.
no code implementations • 1 Dec 2019 • Leena Chennuru Vankadara, Debarghya Ghoshdastidar
This is the first work that provides such optimality guarantees for the kernel k-means as well as its convex relaxation.
1 code implementation • NeurIPS 2018 • Debarghya Ghoshdastidar, Ulrike Von Luxburg
Hypothesis testing for graphs has been an important tool in applied research fields for more than two decades, and still remains a challenging problem as one often needs to draw inference from few replicates of large graphs.
1 code implementation • NeurIPS 2019 • Debarghya Ghoshdastidar, Michaël Perrot, Ulrike Von Luxburg
We address the classical problem of hierarchical clustering, but in a framework where one does not have access to a representation of the objects or their pairwise similarities.
no code implementations • 4 Jul 2017 • Debarghya Ghoshdastidar, Maurilio Gutzeit, Alexandra Carpentier, Ulrike Von Luxburg
Given a population of $m$ graphs from each model, we derive minimax separation rates for the problem of testing $P=Q$ against $d(P, Q)>\rho$.
no code implementations • 17 May 2017 • Debarghya Ghoshdastidar, Maurilio Gutzeit, Alexandra Carpentier, Ulrike Von Luxburg
We consider a two-sample hypothesis testing problem, where the distributions are defined on the space of undirected graphs, and one has access to only one observation from each model.
no code implementations • 5 Apr 2017 • Siavash Haghiri, Debarghya Ghoshdastidar, Ulrike Von Luxburg
We consider machine learning in a comparison-based setting where we are given a set of points in a metric space, but we have no access to the actual distances between the points.
no code implementations • 21 Feb 2016 • Debarghya Ghoshdastidar, Ambedkar Dukkipati
This work is motivated by two issues that arise when a hypergraph partitioning approach is used to tackle computer vision problems: (i) The uniform hypergraphs constructed for higher-order learning contain all edges, but most have negligible weights.
no code implementations • 7 May 2015 • Debarghya Ghoshdastidar, Ambedkar Dukkipati
Hypergraph partitioning lies at the heart of a number of problems in machine learning and network sciences.
no code implementations • NeurIPS 2014 • Debarghya Ghoshdastidar, Ambedkar Dukkipati
Spectral graph partitioning methods have received significant attention from both practitioners and theorists in computer science.
no code implementations • CVPR 2014 • Debarghya Ghoshdastidar, Ambedkar Dukkipati, Ajay P. Adsul, Aparna S. Vijayan
Motivated by multi-distribution divergences, which originate in information theory, we propose a notion of `multi-point' kernels, and study their applications.
no code implementations • 21 Jun 2012 • Debarghya Ghoshdastidar, Ambedkar Dukkipati, Shalabh Bhatnagar
This motivates us to study SF schemes for gradient estimation using the q-Gaussian distribution.
no code implementations • 3 May 2012 • Ambedkar Dukkipati, Gaurav Pandey, Debarghya Ghoshdastidar, Paramita Koley, D. M. V. Satya Sriram
In this paper, we introduce a maximum entropy classification method with feature selection for large dimensional data such as text datasets that is generative in nature.
no code implementations • 9 Apr 2012 • Debarghya Ghoshdastidar, Ambedkar Dukkipati
Motivated by the importance of power-law distributions in statistical modeling, in this paper, we propose the notion of power-law kernels to investigate power-laws in learning problem.