Coffee Meeting Log


An intriguing property of neural networks (up to date)

Akinori Tanaka (Senior Research Scientist, iTHEMS / Senior Research Scientist, RIKEN Center for Advanced Intelligence Project (AIP))

Neural networks have played a central role in the recent development of machine learning technology, but their properties remain mysterious, and it would be interesting if these could be mathematically modeled. For example, in word embedding in a language model, it is known that "king vector" representing the word "king" appears in the learning process, and that the vector obtained by subtracting "queen vector" from "king vector" becomes a vector representing the change of words from female to male. In this talk, I would like to briefly explain that this kind of "concept arithmetic," is also possible among the weight parameters of trained neural networks, which is called "task arithmetic".


An Introduction of Room Acoustics: Theory of Reverberation Time

Shingo Gibo (Postdoctoral Researcher, iTHEMS)

It is important to predict and optimize the acoustics of a room before building the room. One of the most important indexes in the room acoustics is the reverberation time, which is defined as the time it takes for the sound energy to decay by a factor of 10^{-6}. If the reverberation time is too long, understanding speech in the room becomes difficult. Conversely, if the reverberation time is too short, we may not enjoy music in the room. In this talk, I will briefly explain the theory of the reverberation time.


Unique Characterizations of Thermodynamic Entropy

Yuki Yokokura (Senior Research Scientist, iTHEMS)

Entropy has special properties related to heat, microscopic degrees of freedoms, and macroscopic irreversibility. I will explain that these are connected each other through dynamics, and add a new characterization: symmetry of entropy.


Weather forecast and deep learning

Shigenori Otsuka (Research Scientist, iTHEMS / Research Scientist, Data Assimilation Research Team, RIKEN Center for Computational Science (R-CCS))

In this talk, I will introduce application of deep learning to weather forecasting. Recent years, tech companies, such as NVIDIA, Huawei, and Google, reported their deep learning-based global weather prediction models. These models were trained on so-called atmospheric reanalyses to emulate computationally demanding numerical weather prediction models. Although we still need physically-based models for various purposes, deep learning may change the future of weather predictions.

YouTube: Forming black holes from starsPublic


Forming black holes from stars

Lucy McNeill (Postdoctoral Researcher, iTHEMS)

Black holes which are the remnants of stars are being detected at a rate of a few per month using various optical telescopes and gravitational wave interferometers. They provide fruitful opportunity to test and challenge stellar evolution theory, which depends sensitively on our understanding of general relativity, quantum mechanics, particle physics and nuclear physics. In this coffee talk, I will present the physical concepts and mathematical scalings behind the formation of a black hole, after a star’s iron core collapses. I will quantify the key length, time and energy scales involved, and the (possibly surprising) importance of neutrino transport.


Exotica in Mathematics

Taketo Sano (Special Postdoctoral Researcher, iTHEMS)

Since the discovery of an exotic 7-dimensional sphere by J. Milnor in 1956, the study of exotic phenomena has become one of the central topics in topology. Here, an exotic sphere is a smooth manifold that is homeomorphic, but not diffeomorphic, to the standard sphere. In this talk, I will briefly explain the history of the discoveries of exotic phenomena and discuss some of the recent achievements related to knot theory.


A gentle introduction to fluid turbulence

Camilia Demidem (Research Scientist, iTHEMS)

Turbulence is everywhere around us, manifesting itself in seemingly trivial aspects of daily life, such as the act of pouring milk into coffee, while also shaping critical processes in fusion reactors, atmospheric dynamics and astrophysical phenomena. In this talk, I will try to review some fundamental aspects of turbulence and explain why it is so challenging to model it.

YouTube: Singularity TheoremsPublic


Singularity Theorems

Shigehiro Nagataki (Deputy Program Director, iTHEMS / Chief Scientist, Astrophysical Big Bang Laboratory, RIKEN Cluster for Pioneering Research (CPR))

I am happy to introduce the Singularity Theorem, which was proved by Roger Penrose in 1965. He won the Nobel Prize in Physics in 2020 by the proof. I hope you will feel the outline of the proof and understand that the Singularity corresponds to a point outside of spacetime. If there is 1 minute left in my presentation, I would like to mention that Roger Penrose and Stephen Hawking proved the existence of (a) singularity(ies) in natural conditions at the beginning of our universe. Einstein's equation for general relativity is a kind of God's equation, but singularity theorems strongly suggest the limitation of general relativity. I want to thank Prof. Fujikawa, who requested that I give a presentation on the Singularity Theorems. His request motivated me to prepare for my presentation (originally, I was planning to give a short talk on stellar physics using a part of my notebook that I used in my lecture course at OIST).


Quantum channel characterization

Shunji Matsuura (Visiting Scientist, iTHEMS / Senior Researcher, Hardware Inovation Lab, 1QBit, Canada)

The greatest challenge in building a quantum computer is noise. Suppressing noise in quantum systems is extremely difficult, which has led to a long-standing skepticism about the feasibility of quantum computers. So, what exactly is noise in the context of quantum computers? How is it characterized, and how is it measured? In this talk, we will discuss the nature of noise and, as specific examples of methods for characterizing it, we will talk about randomized benchmarking and tomography.


ORCID to auto-report your research contributions & manage your online visibility

Catherine Beauchemin (Deputy Program Director, iTHEMS / Professor, Department of Physics, Toronto Metropolitan University, Canada)

We all have to report our papers (and grants) activities at least once per year to the institutions we work for. Different employers have different systems (e.g. RIKEN has RARS) and it is tiresome to fill these forms again and again. We also want to share or make this information visible to the wider scientific community, as part of looking for a new jobs or for new collaborators. Again a lot of different tools and databases exist (Scopus, Web of Science, Dimensions, Google Scholar, ResearchGate, Pubmed, etc.). In this talk I want to tell you about ORCID: what it is and what it can do for you. Especially, how it can help you solve the problems of reporting and widely disseminating your research accomplishments to the community across the different platforms, while managing it in just one place: your ORCID record. I'll demonstrate some of nice applications.


What is a Mathematical Model to Replicate Filtration Phenomena?

Ken Furukawa (Postdoctoral Researcher, iTHEMS / Postdoctoral Researcher, Prediction Science Laboratory, RIKEN Cluster for Pioneering Research (CPR))

I will discuss a mathematical model concerning filtration of water and air. I will demonstrate how filtration phenomena can be mathematically replicated using special (less known) boundary conditions. I will also explain why these boundary conditions are necessary and discuss future possibilities.


Kin selection and social evolution

Thomas Hitchcock (Special Postdoctoral Researcher, iTHEMS)

All of life is social, yet the evolution of social traits posed a problem to classical Darwinian thinking for over a century. I will introduce the concepts of kin selection, relatedness, and inclusive fitness and talk about how these resolved the puzzle of altruistic behaviours and revolutionised behavioural ecology in the process. I will then discuss how the scope of social evolution has expanded over the years to tackle a wider set of questions, including the origins of individuality itself.


Molecular evolution and the neutral theory

Jeffrey Fawcett (Senior Research Scientist, iTHEMS)

Understanding how biological diversity is created is one of the most fundamental goals in biology. While the work of Charles Darwin formed the basic framework of Evolution, especially by highlighting the role of natural selection, it was the work of Motoo Kimura, who proposed the Neutral Theory and highlighted the role of stochastic processes, that formed the theoretical framework of molecular (i.e., DNA/RNA/protein) evolution. Here, I will explain the basic ideas of the Neutral Theory and the processes involved in creating the diversity at the molecular level.


Welcome to the new world, "Particle Zoo 2.0"

Takumi Doi (Senior Research Scientist, iTHEMS)

In 1960s, many new "fundamental" particles were found and called "Particle Zoo". Their systematic classification lead to the discovery of elementary particles, quarks. Since the beginning of 21c, however, a new kind of mysterious (exotic) particles are unexpectedly being discovered. In this talk, I will introduce this new world of "Particle Zoo 2.0" and its impact.


The undervalued and misunderstood importance of taxonomy in century 21th

José Said Gutiérrez-Ortega (Special Postdoctoral Researcher, iTHEMS)

Taxonomy, the branch of biology that classifies the living beings and give them scientific names, is not a hypothesis-driven science but a descriptive discipline that suffers of a great undervaluation in century 21th. Taxonomic papers will never be published in high impact journals, they won’t get many citations (if any), and getting funding for pure taxonomic research is basically impossible. Furthermore, taxonomy is often criticized for its apparent arbitrariness on how researchers decide to give name to a species, and attaching a scientific name to a group of organisms often seems trivial in a moment when in biology the definition of “species" is heavily discussed. Certainly, this is not a good moment for a biologist to specialize in taxonomy. (Un)fortunately, my research line somehow directed me to become a “part-time taxonomist”, which compels me to try to understand and overcome the challenges of this discipline. I will tell a few of my experiences as a taxonomist and will tell you about, in my opinion, the hottest topic in the modern history of taxonomy: should taxonomy be reinvented by updating its conventional rules?


Do Mathematicians Dream of Quantum Field Theory?

Yuto Moriwaki (Special Postdoctoral Researcher, iTHEMS)

Leibniz discovered differentiation independently of "classical mechanics" and Riemann discovered Riemannian geometry independently of "theory of gravity". Let's ask: "Will Quantum field theory (a fundamental theory of physics) be reconstructed from pure mathematics?". Here I will talk about what I feel quantum field theory looks like as future mathematics based on my current understanding and expectations. One mathematician said that "quantum field theory is the mathematics of the 22nd century." So this is an extremely incomplete and informal talk, so please listen with an easy mind and a cup of coffee :)


What is a path integral?

Christy Koji Kelly (Special Postdoctoral Researcher, iTHEMS)

Path integrals are an important technical tool in physics for describing quantum systems. We will try to provide an intuitive account of why they appear in quantum theory and show why they are mathematically difficult to understand.


How were elements synthesized?

Tomoya Naito (Special Postdoctoral Researcher, iTHEMS)

I will explain how elements were synthesized in the Universe, which is a important topic in nuclear physics.


Two reasons why hydrogen exists

Tetsuo Hatsuda (Program Director, iTHEMS)


Brane Tiling: a bridge between geometry and gauge theory

Dongwook Ghim (Postdoctoral Researcher, iTHEMS)

I will quickly overview the story of brane tiling, also known as dimer model, involving its relation with supersymmetric gauge theory and toric Calabi-Yau geometries.