Year 2024

Please refer to class webpage with overview, class requirements, deadlines, etc. This page is for announcements and class related news.

Course intro
Requirements for projects - More details next week.
Check out the project gudelines at the projects page ! Please note the deadlines!
Questions, requests, etc. concerning the class: contact: szamszimmsc(at)gmail.com
Book chapters up to 14
Overview lecture: slides 1-26 - Nature of intelligence. The process of scientific discovery. Data and models. Extension of senses and brain. An example of scientific big data project: SDSS, the 3D map of the universe.

The AdEMAMix Optimizer: Better, Faster, Older Better neural net optimization? Code
Better precision atomic (actually nuclear) clock Quantamagazine article ; Original Zhang et al. article in Nature
The world’s fastest supercomputer Nature news
Expansion of the universe and the Hubble-tension Adam Riess Nobel laureate explains
Towards the “island of stability” New elements with fuison reactions; Island of stability Wikipedia

Detailed project requirements - by Orsolya Pipek
1st short description deadline this Friday!
Overview lecture: slides 27-60. Moore’s law. The data revolution in astronomy. Simulations, observations and complex model fitting. Data revolution in genomics.

Japan’s planned new zetaFLOPS supercomputer 1000x faster than current best, need 21 nuclear power plant’s electricity
Newest large language model uses “chain of thoughts” for complex reasoning” o1 or “Strawberry”
The standard model is not dead CERN’s calculation of the W boson’s mass agrees with theory, contradicting a previous anomaly that had raised the possibility of new physics.
“Smoke rings” made of electromagnetic field Wang, R., Bao, P.Y., Hu, Z.Q., Shi, S., Wang, B.Z., Zheludev, N.I. and Shen, Y., 2024. Observation of resilient propagation and free-space skyrmions in toroidal electromagnetic pulses. Applied Physics Reviews, 11(3). link
- in the news: https://physicsworld.com/a/vortex-cannon-generates-toroidal-electromagnetic-pulses/
- Original theoretical idea from 1996: Hellwarth, R.W. and Nouchi, P., 1996. Focused one-cycle electromagnetic pulses. Physical Review E, 54(1), p.889. https://pubmed.ncbi.nlm.nih.gov/9965137/
- https://www.youtube.com/watch?v=oruA427cpTg
- Similar even more complex solutions (theory and simulation only): Cameron, R.P., 2018. Monochromatic knots and other unusual electromagnetic disturbances: light localised in 3D. Journal of Physics Communications, 2(1), p.015024. https://iopscience.iop.org/article/10.1088/2399-6528/aa9761

slides 60-97; data revolution in other sciences; science data as high dimensional vector space, projections, principal components.

Feel Moore’s law: Linux boots in 4.76 days on the Intel 4004 cpu from 1971 link - overclocked to 790 kHz :-)
We Finally Know What Creates Static Electricity: ScienceAlert ; Olson, K.P. and Marks, L.D., 2024. What Puts the “Tribo” in Triboelectricity?. Nano Letters
- nice example of partial equation simulation with finite difference method

slides 98-164. Paradigm shift from traditional computational analysis to machine learning. Supervised and unsupervised larning. Tha latent space, autoencoders. Physics of machine learning. Use of machine learning in sciences from cosmology to medical images. State of the art in machine learning, Moore’s law of AI, future promises and challenges.

Loss landscape of neural nets
- a simple analysis with code
- visually stunning loss landscapes and other interesting articles on neural nets
Rewriting Cosmic Calculations: New AI Unlocks the Universe’s Settings
- in the news
- scientific paper: Hahn, C., Lemos, P., Parker, L., Régaldo-Saint Blancard, B., Eickenberg, M., Ho, S., Hou, J., Massara, E., Modi, C., Moradinezhad Dizgah, A. and Spergel, D., 2024. Cosmological constraints from non-Gaussian and nonlinear galaxy clustering using the SimBIG inference framework. Nature Astronomy, pp.1-11. free arxiv version
A real neural network Complete fly brain wiring
- 130,000 cells and 50 million connections.
- Schlegel, P., Yin, Y., Bates, A.S. et al. Whole-brain annotation and multi-connectome cell typing of Drosophila. Nature 634, 139–152 (2024). https://doi.org/10.1038/s41586-024-07686-5 Data is shared. Interesting question: how to analyze, understand such big data?