Machine Learning for Human Learning

Human learning, both in the standard classroom setting and in other settings, e.g., online learning, is crucial to advancing human intelligence. Personalized learning, i.e., recommending personalized remediation or enrichment activities to each learner based on their individual background, interests, and learning progress, has the potential to significantly improve human learning.In this talk, I will present a series of machine learning (ML) methods towards the goal of delivering personalized learning experiences at large scale.

High-Performance Sparse Tensor Decomposition for Data Analytics

Many social and scientific domains give rise to data with multi-way relationships that can naturally be represented by tensors, or multi-dimensional arrays. Decomposing – or factoring – tensors can reveal latent properties that are otherwise difficult to see. However, due to the relatively recent rise in popularity of tensor decomposition in HPC, its challenges in performance optimization is poorly understood.

Analyzing and Mitigating Congestion on High Performance Networks

High-performance networks are a critical component of clusters and supercomputers that enable fast communication between compute nodes. On many platforms, the performance of parallel codes is increasingly communication-bound due to a disproportionate increase in the compute capacity per node but only modest increases in network bandwidths. Hence, it is extremely important to optimize communication on the network. On most architectures, communication performance may be degraded due to network congestion arising from message flows of one or multiple jobs sharing the same network resources.

Toward Transdisciplinary Machine Learning: Scalable Text Mining and Social Influence Modeling

Machine Learning has shown remarkable progress in understanding massive data and making data-driven decisions. In this talk, I will first clarify functions of Machine Learning as exploration, prediction, and explanation, demystifying its specialty against other closely related disciplines. Then I will present the state-of-the-art spectral topic modeling for transparent and scalable exploration of multiple modalities such as large text corpora and various user preferences.

Geometric Representations of Graphs

Geometric representations of graphs keep attracting the attention of researchers and practitioners both as a means of information visualization and for interesting theoretical properties. For instance, many optimization problems that are computationally hard for general graphs are solvable in polynomial time for various classes of geometric intersection graphs. We will review some classical characterization theorems, hardness results, current development in the area and long-standing open problems.

Assurance Techniques for Code Generators

Automated code generation is an enabling technology for model-based software development and promises many advantages but the reliability of the generated code is still often considered as a weak point, particularly in safety-critical domains. Traditionally, correctness-by-construction techniques have been seen as the "right" way to assure reliability, but these techniques remain difficult to implement and to scale up, and have not seen widespread use. Currently, generators are validated primarily by testing, which cannot guarantee reliability and quickly becomes excessive.

Control and Logic

We present control operators in the context of the Curry-Howard isomorphism. We start by analyzing the so called abortive control operators, e.g., callcc. The differing expressive power of various abortive control operators is captured in terms of their logical foundation. A new notion of logic, called minimal classical logic, is introduced. Minimal classical logic does not inforce the Ex Falso Quodlibet axiom. Computationally, this translates into operators which do not allow one to abort a program execution.


Subscribe to RSS - Colloquium