FY2020

Physics and Biology Unit 
Professor Jonathan Miller

Abstract

The unit conducts theoretical, computational, and experimental studies of collective phenomena, from epidemiological physics and machine learning to novel model organisms for information processing.

1. Staff

  • Dr. Shotaro Funai, Staff Scientist
  • Dr. Teresa Iglesias, Postdoctoral Scholar (- May 2020)
  • Dr. Zdenek Lajbner, Technician
  • Dr. Reuven Pnini, Technician
  • John Parker, Technician (Feb 2021 - )
  • Mayu Suzuki, Administrative Assistant (Sep 2020-)

2. Collaborations

2.1 Thermodynamics and feature extraction by machine learning

  • Type of collaboration: Joint research
  • Researchers:
    • Dr. Dimitrios Giataganas, National Center for Theoretical Sciences (NCTS), National Tsing-Hua University, Taiwan.
    • Shotaro Shiba Funai, OIST.

2.2 Comparative cytogenetics

  • Type of collaboration: Joint research
  • Researchers:
    • Professor Petr Ráb, Institute of Animal Physiology and Genetics of the Czech Academy of Sciences, Libechov, Czechia.
    • Zdenek Lajbner, OIST.

2.3 Epidemiological physics of the Covid-19 pandemic.

  • Type of collaboration: Joint research
  • Researchers:
    • Greg Huber, Mason Kamb, Lucy M Li, Aaron McGeever; Chan Zuckerberg Biohub, San Francisco, CA 94158, USA.
    • Kyle Kawagoe, Department of Physics and Kadanoff Center for Theoretical Physics, University of Chicago, Chicago, IL 60637, USA.
    • Boris Veytsman, Chan Zuckerberg Initiative, Redwood City, CA 94063, USA; School of Systems Biology, George Mason University, Fairfax, VA 22030, USA. 
    • Dan Zigmond; Jikoji Zen Center, 12100 Skyline Boulevard, Los Gatos, CA 95033, USA.
    • Jonathan Miller, OIST.

3. Activities and Findings

3.1 Thermodynamics and feature extraction by machine learning

Contributors: D. Giataganas, National Tsing-Hua Univ. (Taiwan); S. S. Funai, OIST. Phys. Rev. Research 2020.

Background

In the field of machine learning, image recognition technology has developed rapidly in recent years. The machine can extract features of images and recast them as vector embeddings. It was anticipated that such feature extraction might be analogous to coarse-graining or renormalization in statistical physics. We studied a 2-d Ising model from statistical physics, regarding the spin configurations as black-and-white images. We train the machine so that it can extract features from these images and investigate a potential analogy to renormalization.

Figure 3.2.1: The machine learning of spin configurations in an Ising model. The machine is trained to reconstruct input images optimally, subject to a fixed number of hidden nodes. As a result, the machine can extract features of the images.

Results

In the renormalization group (RG), irrelevant (\(\simeq\) unimportant) parameters flow to smaller values. The RBM, or Restricted Boltzmann Machine, is a variety of neural network popular for feature extraction. Inspired by RG flow, we define the RBM flow where the extracted features are amplified in the process of image reconstruction. Our calculations show that for the 2-d Ising model, the RBM flow (feature extraction) differs from the RG flow (renormalization). We find instead that the fixed point of the RBM flow is coincident with the parameters for maximal specific heat in Ising model.

Figure 3.2.2: RBM flow (red->green->blue) in the parameter space of temperature (T) and external magnetic field (H).

Figure 3.2.3: Fixed points of RBM flow in the space of (T,H). They are coincident with the maximal specific heat.

Conclusions

Our observations for the 2-d Ising model suggest that the feature extraction from images by machine learning is not directly analogous to renormalization (coarse-graining) but nevertheless corresponds to thermodynamical properties. We are searching for an explanation of why feature extraction is related to thermodynamics.

3.2 Oval squid – emerging model for camouflage

Contributors: Zdeněk Lajbner, Ryuta Nakajima, Tamar Gutnick, Teresa L. Iglesias, Keishu Asada, Takahiro Nishibayashi, Michael J. Kuba, Jonathan Miller

To date, studies of cephalopod camouflage have been directed mainly at benthic species. A recent study conducted in our cephalopod breeding facility demonstrated that semi-pelagic teuthid oval squid Sepioteuthis lessoniana Sp.2 (Shiro-ika) display two different forms of substrate-specific camouflage behavior: 1) motion camouflage and 2) situational motionless camouflage. In contrast to cuttlefish, oval squid combine a color-change with semitransparency. After 7 generations of Shiro-ika captive breeding, we have so far observed no sign of inbreeding depression. Together, these characteristics make the oval squid an especially suitable model animal for study of camouflage.

Fig 1. The seventh captive generation of Shiro-ika camouflaging to substrate at OIST Marine Science Station (image courtesy Zdenek Lajbner).

3.3 Present and Future Salmonid Cytogenetics

Contributors: Gaffaroglu Muhammet, Zuzana Majtánová, Radka Symonová, Šárka Pelikánová, Sevgi Unal, Zdeněk Lajbner, Petr Ráb

The complex phylogeny and taxonomy of Salmonids are challenging for traditional approaches in research. Firstly, we contributed to discovery of the hitherto unknown cytogenetic characteristics of the Anatolian endemic flathead trout, Salmo platycephalus, and summarization of the presently known, but highly complicated, situation in the genus Salmo. Secondly, by outlining future directions of salmonid cytogenomics, we have contributed to the prototypical virtual karyotype of Salmo trutta, the closest relative of S. platycephalus. This achievement became possible thanks to the high-quality genome assembled to the chromosome level in S. trutta via soft-masking, including a direct labelling of repetitive sequences along the chromosome sequence. Repetitive sequences were crucial for traditional cytogenetics and hence should play a central role in cytogenomics. As such virtual karyotypes become increasingly available in the very near future, it will be fruitful to take full advantage of existing and future developments in this technology. Finally, we contributed to a demonstration of how presumptively repetitive sequences in salmonids can alter and improve the understanding of the overall relationship between genome size and G+C content. Salmonids are extremely important economically and scientifically; we anticipate rapid future advances in this area.

Fig 1. Male flathead trout (image courtesy Zdenek Lajbner).

Fig 2. Female flathead trout (image courtesy Zdenek Lajbner).

4. Publications

4.1 Journals​​

  1. Eric Edsinger*, Reuven Pnini*I, Natsumi Ono, Ryoko Yanagisawa, Kathryn Dever, Jonathan Miller. “Social tolerance in Octopus laqueus—A maximum entropy model”, Plos One Vol. 15, Issue 6, e0233834. June 10, 2020. DOI https://doi.org/10.1371/journal.pone.0233834 
  2. Shotaro Shiba Funai*, Hirotaka Sugawara, “Current algebra formulation of quantum gravity and its application to cosmology”, Progress of Theoretical and Experimental Physics, Vol. 2020, Issue 9, 093B08, September 2020, Research, https://doi.org/10.1093/ptep/ptaa108
  3. Shotaro Shiba Funai*, Dimitrios Giataganas, “Thermodynamics and feature extraction by machine learning”, Physics Review Research, Issue number 2, Pages 033415, September 2020, Research, https://doi.org/10.1103/PhysRevResearch.2.033415　
  4. Muhammet Gaffaroglu, Zuzana Majtánová, Radka Symonová, Šárka Pelikánová, Sevgi Unal, Zdeněk Lajbner, Petr Ráb, “Present and Future Salmonid Cytogenetics”, Genes, 11 no. 12 (2020), Pages 1462, December 6, 2020, https://doi.org/10.3390/genes11121462
  5. Hiromasa Watanabe, Georg Bergner, Norbert Bodendorfer, Shotaro Shiba Funai*, Masanori Hanada, Enrico Rinaldi, Andreas Schäfer, Pavlos Vranas, “Partial Deconfinement at Strong Coupling on the Lattice”, Journal of High Energy Physics, 2102, 004, Feb 1, 2021, Research, https://doi.org/10.1007/JHEP02(2021)004
  6. Greg Huber, Mason Kamb, Kyle Kawagoe, Lucy M Li, Aaron McGeever, Jonathan Miller, Boris A Veytsman, Dan Zigmond, “A minimal model for household-based testing and tracing in epidemics”, Physical biology, January 12, 2021, Research,  https://doi.org/10.1088/1478-3975/abdacd, https://iopscience.iop.org/article/10.1088/1478-3975/abdacd/pdf
  7. Warren DL, Dornburg A, Zapfe K, Iglesias TL (2021) The effects of climate change on Australia’s only endemic Pokémon: Measuring bias in species distribution models. Methods in Ecology and Evolution 12:985-95 https://doi.org/10.1111/2041-210X.13591
  8. Drerup C, Sykes AV, Cooke GM (2020) Behavioural aspects of the spotty bobtail squid Euprymna parva (Cephalopoda: Sepiolidae). Journal of Experimental Marine Biology and Ecology 530: 151442. https://doi.org/10.1016/j.jembe.2020.151442

4.2 Books and other one-time publications

Nothing to report

4.3 Oral and Poster Presentations

[Oral Presentations]

  1. Shotaro Funai, “How can AI solve social problems?”, ACCJ (The American Chamber of Commerce in Japan) Kansai, Online(Zoom), June 5, 2020 / 船井正太郎, "AIは社会課題をどう解決できるか”, 在日米国商工会議所, 関西支部（Zoomでの講演), 2020年6月5日 https://www.accj.or.jp/kansai-chapter 
  2. Shotaro Funai, “New money creation with GDP gap as a background using AI”, Matsuda Policy Institute, Online(Zoom) June 28, 2020 / 船井正太郎, “AIを活用したGDPギャップをバックとする新しいマネーの創造”, 松田政策研究所,（丹羽経済塾, Zoomでの講演), 2020年6月28日 http://matsuda-pi.com/niwa.html
  3. Shotaro Shiba Funai, “Research with Matsuo-sensei and After”, Yutaka Matsuo Sexagenarian Memorial Workshop, Online,  September 19, 2020 / 船井（柴）正太郎, “松尾先生との研究とその後”, 松尾泰先生還暦記念研究会　(オンラインでの講演),  2020年9月19日 https://sites.google.com/view/ym60/
  4. Ryuta Nakajima, Zdeněk Lajbner, Tamar Gutnick, Teresa L. Iglesias, Keishu Asada, Takahiro Nishibayashi, Michael J. Kuba, Jonathan Miller  (2020) Squid camouflage to substrate. In: CephRes2020 Biology and Life History of Cephalopods: behavior, cognition, evolution, ecology, fisheries, genomics, neuroscience, taxonomy. International Cephalopod Scientific Community Symposium, Naples, Italy (Virtual Event) . 16.09.2020 - 21.09.2020, Napoli, Italy. Book of abstracts p 67.　https://www.cephalopodresearch.org/cephres_wordpress/wp-content/uploads/2020/09/CephRes2020-Event-Day_6.pdf
  5. Zdenek Lajbner, Ryuta Nakajima, Jonathan Miller, “Oval squid – emerging model for camouflage” 第三回 イカ・タコ研究会 on the Web, Zoom, October 10, 2020

[Poster Presentation]

  1. Shotaro Shiba Funai, “Can extracted features by machine learning be quantified in terms of physics?”, International Symposium on Artificial Intelligence and Brain Science, Online, October 10-12, 2020, http://www.brain-ai.jp/symposium2020/posters/

[Seminar]

  1. Shotaro Funai, “New form of money using AI: what is PMC (Personal Money Creation)?”, Matsuda Policy Institute, August 25, 2020 / 船井（柴）正太郎, “AIを活用した新しいお金の形：PMC（個人信用創造）とは何か？”, 松田政策研究所,  2020年8月25日, https://youtu.be/I1uyPLs0gCo

5. Intellectual Property Rights and Other Specific Achievements

Nothing to report

6. Meetings and Events

([NOTE]  You can include the following in "6. Meetings and Events":
(1)    Seminars and workshops by guest speaker(s)
(2)    Seminars and workshops by guest speaker(s) and OIST faculty member(s)/unit member(s)

6.1 Seminar Title in Full

  • Date: June 17, 2010
  • Venue: OIST Campus Lab1
  • Speaker: Dr. First Last (University of Something)

6.2 Something Group for Something on Something (SG2S)

  • Date: June 17, 2010
  • Venue: OIST Campus Lab1
  • Co-organizers: The Institute of All But Cats (IABC)
  • Speakers:
    • Dr. First Last (Affiliation)
    • Dr. First Last (Affiliation)
    • Dr. First Last (Affiliation)

7. Other

Nothing to report.