Friday, April 2, 2021

2021-04-09 / 12:00 ~ 13:00

학과 세미나/콜로퀴엄 - 수리생물학: Highly accurate fluorogenic DNA sequencing with information theory–based error correction

by 홍혁표(KAIST)

We will discuss about “Highly accurate fluorogenic DNA sequencing with information theory–based error correction”, Chen et al., Nature Biotechnology (2017) Eliminating errors in next-generation DNA sequencing has proved challenging. Here we present error-correction code (ECC) sequencing, a method to greatly improve sequencing accuracy by combining fluorogenic sequencing-by-synthesis (SBS) with an information theory–based error-correction algorithm. ECC embeds redundancy in sequencing reads by creating three orthogonal degenerate sequences, generated by alternate dual-base reactions. This is similar to encoding and decoding strategies that have proved effective in detecting and correcting errors in information communication and storage. We show that, when combined with a fluorogenic SBS chemistry with raw accuracy of 98.1%, ECC sequencing provides single-end, error-free sequences up to 200 bp. ECC approaches should enable accurate identification of extremely rare genomic variations in various applications in biology and medicine.

2021-04-07 / 12:00 ~ 13:00

학과 세미나/콜로퀴엄 - 기타: (첫수융합포럼) 곡면 상의 위상적 동역학계와 그 응용

by 백형렬(KAIST 수리과학과)

Thurston classified mappings from a given surface to itself. By iterating the surface mappings, one can view this as a dynamical system. Most of those surface mappings are so-called pseudo-Anosov. We briefly explain how we should understand these pseudo-Anosov maps and their physical meaning.

2021-04-07 / 12:00 ~ 13:00

학과 세미나/콜로퀴엄 - 기타: (첫수융합포럼) 액정물질 기반 위상학적 결함구조의 상전이

by 윤동기(KAIST 화학과)

Topological defect structure is one of the most interesting topics in natural sciences. Especially, the topological defect transition was highlighted by Nobel prize in 2016. But this topic is hard to understand and realize in the practical condition because the size and time-scale are huge in cosmos or so tiny in skyrmion system. So, we proposed to use liquid crystal (LC) materials to directly show this interesting topic, phase transition of topological defect.

2021-04-09 / 16:00 ~ 17:00

학과 세미나/콜로퀴엄 - 대수기하학: Introduction to infinity-categories III

by 조창연(QSMS Seoul National University)

This is the part 3 of the lectures on infinity-categories: This talk will be focused on introducing quasi-categories as our model for infinity-categories. After reviewing some background material needed to define quasi-categories, we'll see how the definition works.

2021-04-09 / 10:00 ~ 12:00

학과 세미나/콜로퀴엄 - SAARC 세미나:

by ()

At Data Science Group, we try to offer computational models for challenging real-world problems. This talk will introduce two such problems that can benefit from collaboration with mathematicians and theorists. One is customs fraud detection, where the goal is to determine a small set of fraudulent transactions that will maximize the tax revenue when caught. We had previously shown a state-of-the-art deep learning model in collaboration with the World Customs Organization [KDD2020]. The next challenge is to consider semi-supervised (i.e., using very few labels) and unsupervised (i.e., no label information) settings that better suit developing countries' conditions. Another research problem is poverty mapping, where the goal is to infer economic index from high-dimensional visual features learned from satellite images. Several innovative algorithms have been proposed for this task [Science2016, AAAI2020, KDD2020]. I will introduce how we approach this problem under extreme conditions with little validation data, as in North Korea.

2021-04-02 / 10:00 ~ 12:00

학과 세미나/콜로퀴엄 - SAARC 세미나:

by 윤철희()

Overparametrized neural networks have infinitely many solutions that achieve zero training loss, but gradient-based optimization methods succeed in finding solutions that generalize well. It is conjectured that the optimization algorithm and the network architecture induce an implicit bias towards favorable solutions, and understanding such a bias has become a popular topic. We study the implicit bias of gradient flow (i.e., gradient descent with infinitesimal step size) applied on linear neural network training. We consider separable classification and underdetermined linear regression problems where there exist many solutions that achieve zero training error, and characterize how the network architecture and initialization affects the final solution found by gradient flow. Our results apply to a general tensor formulation of neural networks that includes linear fully-connected networks and linear convolutional networks as special cases, while removing convergence assumptions required by prior research. We also provide experiments that corroborate our theoretical analysis.

2021-04-02 / 16:00 ~ 17:00

학과 세미나/콜로퀴엄 - 대수기하학: Introduction to infinity-categories II

by 조창연(QSMS Seoul National University)

This is part II of the lecture series in infinity-categories. I'll continue to talk about higher categories and some difficulty in defining them. In the end, a few models for infinity-categories will be introduced very roughly.

2021-04-08 / 16:15 ~ 17:15

학과 세미나/콜로퀴엄 - 콜로퀴엄: Reduction of stochastic systems via resolvent equations

by 서인석(서울대학교)

In this talk, we consider stochastic systems with several stable sets. Typical examples are low-temperature physical systems and stochastic optimization algorithms. The macroscopic description of such systems is usually carried out via a so-called model reduction. We explain a necessary and sufficient condition for model reduction in terms of solutions of certain form of partial differential equations.

Events for the 취소된 행사 포함