Date/Time : 18 Feb 2022 10:00
Venue : via Zoom
Speaker : Yang Haizhao, Purdue University

Join Zoom Meeting
https://nus-sg.zoom.us/j/9775606531?pwd=cVZ6eTRYS3ZzVUlFNXRuSlhMVjJSUT09

Meeting ID: 977 560 6531
Passcode: 123321

Abstract: Learning operators between infinitely dimensional spaces is an important learning task arising in wide applications in machine learning, imaging science, mathematical modeling and simulations, etc. This talk introduces the nonparametric estimation of Lipschitz operators using deep neural networks. Non-asymptotic upper bounds are derived for the generalization error of the empirical risk minimizer over a properly chosen network class. Under the assumption that the target operator exhibits a low dimensional structure, our error bounds decay as the training sample size increases, with an attractive fast rate depending on the intrinsic dimension in our estimation. Our assumptions cover most scenarios in real applications and our results give rise to fast rates by exploiting low dimensional structures of data in operator estimation. We also investigate the influence of network structures (e.g., network width, depth, and sparsity) on the generalization error of the neural network estimator and propose a general suggestion on the choice of network structures to maximize the learning efficiency quantitatively.

Add to calendar :

Date/Time : 11 Feb 2022 10:00
Venue : via Zoom
Speaker : Soufiane Hayou, National University of Singapore

Join Zoom Meeting
https://nus-sg.zoom.us/j/9775606531?pwd=cVZ6eTRYS3ZzVUlFNXRuSlhMVjJSUT09

Meeting ID: 977 560 6531
Passcode: 123321

Abstract: The theoretical analysis of infinite-width neural networks has led to many interesting practical results (choice of initialization schemes, choice of Bayesian priors etc.). However, the traditional infinite-width framework focuses on fixed depth networks and omits the large depth behavior of these models. In this talk, I will cover different topics on the infinite-width-then-infinite-depth networks and show how we can leverage some theoretical results to obtain principled algorithms. The following papers will be covered in this talk:

• On the Impact of the activation function on deep neural networks training. ICML 2019. (Joint work with A. Doucet and J. Rousseau)
• Stable ResNet. AISTATS 2021. (Joint work with E. Clerico, B. He, G. Deligiannidis, A. Doucet, and J. Rousseau)

Add to calendar :

Date/Time : 12 Jan 2022 09:00
Venue : via Zoom
Speaker : Cheng Mao, Georgia Tech

Join Zoom Meeting
https://nus-sg.zoom.us/j/85910065038?pwd=ay9nT0hzR2ROeUs4U3d5dTBoMGNBdz09

Meeting ID: 859 1006 5038
Password: 753065

Abstract: Finding latent structures in networks is an important task in many fields of science and engineering.  For a variety of problems, the task can be formulated as inferring a latent permutation of vertices of a network from noisy data. Problems in this class are often computationally infeasible in the worst case due to the combinatorial nature of permutations, so statistical methods are essential for inference in large-scale networks. This talk will provide an overview of two problems: ranking items from comparisons and matching correlated networks. For each problem, I will introduce an average-case model and the associated task of learning a latent permutation. Then I will discuss recent development of efficient algorithms that come with theoretical guarantees.

Add to calendar :

Date/Time : 07 Jan 2022 10:00
Venue : via Zoom
Speaker : Pavitra Krishnaswamy, I2R, A*STAR

Join Zoom Meeting
https://nus-sg.zoom.us/j/89775139658?pwd=b3VxL0pDT0ZNTnIrUUV6MWNZK28xQT09

Meeting ID: 897 7513 9658
Passcode: 238865

Abstract:  Real-world health data from clinical records, disease registries, consumer health devices, and mobile applications offer great potential for a number of health intelligence applications. However, effectively learning from these observational datasets remains challenging as they lack detailed annotations, exhibit significant irregularities, and are often of variable quality. In this talk, I will discuss our research efforts in developing machine learning capabilities to address these challenges for targeted screening, allied health and clinical decision support applications.

Add to calendar :

Date/Time : 17 Dec 2021 10:00
Venue : via Zoom
Speaker : Masashi Sugiyama, RIKEN and University of Tokyo

Join Zoom Meeting
https://nus-sg.zoom.us/j/9775606531?pwd=cVZ6eTRYS3ZzVUlFNXRuSlhMVjJSUT09

Meeting ID: 977 560 6531
Passcode: 123321

Abstract: Supervised machine learning based on labeled big data has been highly successful in a wide range of applications. Nevertheless, there are still many challenging situations where only limited labeled data is available, e.g., due to high labeling costs, rareness of events, and privacy concerns. In this talk, I will give an overview of our recent advances in weakly supervised classification based on empirical risk minimization. Our framework contains various important weakly supervised classification problems such as positive-unlabeled classification, positive-confidence classification, similar-dissimilar classification, and unlabeled-unlabeled classification for binary problems, and complementary-label classification, partial-label classification, and single-class-confidence classification for multi-class problems. This framework allows us to estimate the classification risk only from weakly supervised data, provides theoretical convergence guarantee, and is compatible with any loss function, classifier, and optimizer.

Add to calendar :

Date/Time : 10 Dec 2021 10:00
Venue : via Zoom
Speaker : Xu Min, Carnegie Mellon University

Join Zoom Meeting
https://nus-sg.zoom.us/j/9775606531?pwd=cVZ6eTRYS3ZzVUlFNXRuSlhMVjJSUT09

Meeting ID: 977 560 6531
Passcode: 123321

The cell is the basic structural and functional unit of all living organisms. Understanding how cells function is fundamental to life science. Macromolecules are nano-machines inside cells that govern the cellular processes. To fully understand such processes, it is necessary to know the native structures and spatial organizations of macromolecules inside single cells, and their interactions with other subcellular components. Such information has been extremely difficult to obtain due to a lack of suitable data acquisition techniques. The recent revolution of Cryo-electron tomography (cryo-ET) 3D imaging technology has made collecting such information possible. Cryo-ET captures a 3D image of a single cell’s subcellular structures at sub-molecular resolution and in a near-native state. It provides unprecedented opportunities for systematically studying the native spatial organization of subcellular structures, especially macromolecules. However, cryo-ET has a high degree of structural complexity and imaging limits,such as high structural diversity and crowding, low signal-to-noise ratio, and missing values. These have made the automated systematic analysis of such images extremely difficult. Since 2008, we have been developing image analysis methods to address this challenge. In particular, we focus on systematic recognition and recovery of the structures of large numbers (millions) of macromolecules captured by cryo-ET, without relying on external structural knowledge. To do so, we have developed different cryo-ET image registration, classification, segmentation restoration techniques. Our effort is a key step for systematic analysis of macromolecules’ structures and spatial organizations inside single cells captured by cryo-ET.

Add to calendar :

Date/Time : 12 Nov 2021 16:00
Venue : via Zoom
Speaker : Alexandros Beskos, University College of London

Join Zoom Meeting
https://nus-sg.zoom.us/j/9775606531?pwd=cVZ6eTRYS3ZzVUlFNXRuSlhMVjJSUT09

Meeting ID: 977 560 6531
Passcode: 123321

Abstract: Bayesian inference for nonlinear diffusions, observed at discrete times,is a challenging task that has prompted the development of a number of algorithms, mainly within the computational statistics community. We propose a new direction, and accompanying methodology – borrowing ideas from statistical physics and computational chemistry – for inferring the posterior distribution of latent diffusion paths and model parameters, given observations of the process. Joint configurations of the underlying process noise and of parameters, mapping onto diffusion paths consistent with observations, form an implicitly defined manifold. Then, by making use of a constrained Hamiltonian Monte Carlo algorithm on the embedded manifold, we are able to perform computationally efficient inference for a class of discretely observed diffusion models. Critically, in contrast with other approaches proposed in the literature, our methodology is highly automated, requiring minimal user intervention and applying alike in a range of settings, including: elliptic or hypo-elliptic systems; observations with or without noise; linear or non-linear observation operators. Exploiting Markovianity, we propose a variant of the method with complexity that scales linearly in the resolution of path discretisation and the number of observation times. Example Python code is given at git.io/m-mcmc.

Add to calendar :

Date/Time : 12 Nov 2021 10:00
Venue : via Zoom
Speaker : Clement Canonne, University of Sydney

Join Zoom Meeting
https://nus-sg.zoom.us/j/9775606531?pwd=cVZ6eTRYS3ZzVUlFNXRuSlhMVjJSUT09

Meeting ID: 977 560 6531
Passcode: 123321

Abstract: In this talk, I will introduce the primitive of “domain compression” over the probability simplex, which enables us to reduce the domain size from k to L while contracting (with high probability) the pairwise total variation distances by a factor at most √(L/k). This can equivalently be seen as some type of dimensionality reduction for probability distributions, preserving (in a one-sided way) the L1 distances.

Starting with the corresponding statement for L=2, I will show how to perform this domain compression in a randomness-efficient fashion, and discuss its optimality. As an application, I will describe how this yields sample-optimal algorithms for goodness-of-fit under various constraints, such as communication (bandwidth) constraints, local privacy, and shuffle privacy.

Based on joint works with Jayadev Acharya, Cody Freitag, Yanjun Han, Ziteng Sun, and Himanshu Tyagi (in particular, (https://arxiv.org/abs/1907.08743; COLT 2020)); and some insights from a paper by Amin, Joseph, and Mao (https://arxiv.org/abs/1911.01452; COLT 2020).

Add to calendar :

Date/Time : 29 Oct 2021 10:00
Venue : via Zoom
Speaker : Jian Tang, Mila-Quebec Al Institute

Join Zoom Meeting
https://nus-sg.zoom.us/j/9775606531?pwd=cVZ6eTRYS3ZzVUlFNXRuSlhMVjJSUT09

Meeting ID: 977 560 6531
Passcode: 123321

Abstract:  Drug discovery is a very long and expensive process, taking on average more than 10 years and costing $2.5B to develop a new drug. Artificial intelligence has the potential to significantly accelerate the process of drug discovery by extracting evidence from a huge amount of biomedical data and hence revolutionizes the entire pharmaceutical industry. In particular, graph representation learning and geometric deep learning–a fast growing topic in the machine learning and data mining community focusing on deep learning for graph-structured and 3D data—has seen great opportunities for drug discovery as many data in the domain are represented as graphs or 3D structures (e.g. molecules, proteins, biomedical knowledge graphs). In this talk, I will introduce our recent progress on geometric deep learning for drug discovery and also a newly released open-source machine learning platform for drug discovery, called TorchDrug.

Add to calendar :