Yuanpeng's Blog RSS Feed https://iris2020.net/ This post is part-3 of the 4-parts full story about the Langevin dynamics. All references will be presented in part-4. Part VII – The Overdamped (Smoluchowski) Limit 7.1 The physical regime Two timescales are present in the Langevin system: Timescale Symbol Meaning Momentum relaxation \(\tau_p = 1/\gamma = m/\zeta\) time... Sun, 16 Aug 2026 00:00:00 -0400 https://iris2020.net/2026-08-16-langevin_dynamics_3/ https://iris2020.net/2026-08-16-langevin_dynamics_3/ This post is part-2 of the 4-parts full story about the Langevin dynamics. All references will be presented in part-4. Part IV – The Fluctuation–Dissipation Relation The task of this part is to determine the relationship between the two coefficients \(\gamma\) and \(\sigma\) left undetermined in (2.1). Two logically distinct... Sat, 15 Aug 2026 00:00:00 -0400 https://iris2020.net/2026-08-15-langevin_dynamics_2/ https://iris2020.net/2026-08-15-langevin_dynamics_2/ This post is part-1 of the 4-parts full story about the Langevin dynamics. All references will be presented in part-4. 0. Scope, prerequisites, and how to read this This document develops Langevin dynamics from first principles and follows a single thread all the way from classical statistical mechanics to the... Thu, 13 Aug 2026 00:00:00 -0400 https://iris2020.net/2026-08-13-langevin_dynamics_1/ https://iris2020.net/2026-08-13-langevin_dynamics_1/ In one of my earlier posts [1], several typical sampling algorithms were briefly discussed, such as the direct inverse sampling, rejection sampling and Markov Chain Monte Carlo (MCMC) sampling. The direct inverse sampling requires analytical form of the distribution function and reversability. For suitable 1D distribution function, it can be... Thu, 13 Aug 2026 00:00:00 -0400 https://iris2020.net/2026-08-13-gibbs_sampling/ https://iris2020.net/2026-08-13-gibbs_sampling/ The image was generated with Gemini with the prompt of ‘Generate an image as an abstract showcase for the variational autoencoder. Use the Ghibli style’. 1. General Introduction A Variational Autoencoder (VAE) is a generative model that represents data through a probabilistic latent space, allowing new data to be synthesized... Tue, 11 Aug 2026 00:00:00 -0400 https://iris2020.net/2026-08-11-variational_autoencoder/ https://iris2020.net/2026-08-11-variational_autoencoder/ Image generated with Gemini with the prompt ‘generate an image with a person thinking hardly on how diffusion model really works. follow the ghibli style’. My learning for the diffusion model started with the cource by fast.ai, which is a great learning resource. It tries to stay on the practical... Sun, 09 Aug 2026 00:00:00 -0400 https://iris2020.net/2026-08-09-diffusion_algorithm/ https://iris2020.net/2026-08-09-diffusion_algorithm/ This post notes down the understanding of several typical sampling algorithms, including the inverse transform sampling, the rejectin sampling and Markov Chain Monte Carlo sampling. I am not going to cover the nitty-gritty details but just to note down some of what I think as critical points benefiting the understanding... Sun, 02 Aug 2026 00:00:00 -0400 https://iris2020.net/2026-08-02-sampling_algorithms/ https://iris2020.net/2026-08-02-sampling_algorithms/ This post is written by both AI and myself in a ‘collaborative’ way. I was learning through the sampling technique with Claude AI and was asking quite a lot questions back and forth, trying to understand each of the pieces clearly. After that, I asked Claude AI to summarize the... Fri, 31 Jul 2026 00:00:00 -0400 https://iris2020.net/2026-07-31-markov_chain_monte_carlo/ https://iris2020.net/2026-07-31-markov_chain_monte_carlo/ In Reverse Monte Carlo modeling, when the data are not providing strong enough constraint towards the fitting target, potentially we would be having the overfitting problem. The fitting may converge to a state simply because the state entropy is high, i.e., the landscape of the configuration space is very flat... Thu, 30 Jul 2026 00:00:00 -0400 https://iris2020.net/2026-07-30-rmc_sol_uniqueness/ https://iris2020.net/2026-07-30-rmc_sol_uniqueness/ Image generated with Gemini with the prompt ‘robot with intuition for loss function’ (here I am using exactly the same prompt as presented in a reference post about cross entropy loss. [1]). Cross entropy loss is a commonly used loss function for classification problems with neural network. A nice and... Wed, 29 Jul 2026 00:00:00 -0400 https://iris2020.net/2026-07-29-cross_entropy_loss_gradient/ https://iris2020.net/2026-07-29-cross_entropy_loss_gradient/ If we think of network as a big region in which we have a lot of houses (analogous to devices in the network), subnets are just smaller sub-regions inside the big region. Houses are separated according to those sub-regions, which determines how they can talk to each other. This is... Thu, 23 Jul 2026 00:00:00 -0400 https://iris2020.net/2026-07-23-subnet/ https://iris2020.net/2026-07-23-subnet/ This is the first post in the series of using Strands agent SDK for agent development. The walkthrough can be found in the official quick start documentation with Python [1]. This post serves as the side note for the documentation, noting down some technical details that we may need regarding... Sun, 05 Apr 2026 00:00:00 -0400 https://iris2020.net/2026-04-05-strand_agent/ https://iris2020.net/2026-04-05-strand_agent/ In this post, I will be noting down how to use Dufs [1] for setting up a file server on Windows 11. Originally, I was using gohttpserver which is no longer maintained, and Dufs is one of the recommended options on the gohttpserver GitHub [2]. First, we need to install... Fri, 03 Apr 2026 00:00:00 -0400 https://iris2020.net/2026-04-03-win_fserver/ https://iris2020.net/2026-04-03-win_fserver/ Taking the \(Y_4^3\) function as the example, the full mathematical form of the function is, \[Y_4^3 = \frac{3}{4}\sqrt{\frac{35}{\pi}}sin^3\theta cos\theta cos(3\phi)\] To visualize the function, we first need to make it clear what we are trying to visualize. First, physically, this functions gives the distribution of, e.g., the electron density, as... Mon, 05 Jan 2026 00:00:00 -0500 https://iris2020.net/2026-01-05-spherical_harmonics_viz/ https://iris2020.net/2026-01-05-spherical_harmonics_viz/ In the age of AI, is it really making any real sense to write blogs to put down something that AI can probably generate in a flash? In fact, when I was writing this blog, initially I was using the phrase ‘in a blinking time’ but as a non-native English... Thu, 01 Jan 2026 00:00:00 -0500 https://iris2020.net/2026-01-01-goldstone_mode_symmetry/ https://iris2020.net/2026-01-01-goldstone_mode_symmetry/ In one of my previous posts, I had some notes about the preferred orientation and texture and the two typical ways to represent them [1]. In this post, I am trying to put down some follow-up notes about the topic, hoping to add some more understanding of the topic. Preferred... Thu, 11 Dec 2025 00:00:00 -0500 https://iris2020.net/2025-12-11-preferred_orientation_texture_followup/ https://iris2020.net/2025-12-11-preferred_orientation_texture_followup/ In the low temperature region, the Dulong-Petit describing the specific heat \(C_V\) as \(3R\) fails – \(C_V\) will approach 0 while \(3R\) stays a constant across the whole temperature range. Fundamenally, this is due to the failure of the equipartition theorem in the low temperature region. Einstein gave the problem... Thu, 04 Dec 2025 00:00:00 -0500 https://iris2020.net/2025-12-04-einstein_debye_model/ https://iris2020.net/2025-12-04-einstein_debye_model/ When we have a high-dimension configuration space, e.g., a system containing hundreds of thousands of atoms with each atom carrying 3 degrees of freedom (DOF), it becomes difficult to sample the system. For example, if we want to sample the configuration with low energy (thus considered as energetically stable) or... Sun, 30 Nov 2025 00:00:00 -0500 https://iris2020.net/2025-11-30-metropolis_monte_carlo/ https://iris2020.net/2025-11-30-metropolis_monte_carlo/ Understanding ion diffusion is essential in battery materials research and characterizing diffusion pathways is critical for rational materials design and optimization. While using the diffraction technique for the characterization of battery materials, we could obtain the refined structure configuration via fitting the diffraction data, with the small-box or big-box approach,... Sun, 26 Oct 2025 00:00:00 -0400 https://iris2020.net/2025-10-26-bondstr_bvs/ https://iris2020.net/2025-10-26-bondstr_bvs/ GSAS-II is a powerful software for analyzing the crystallographic structure from diffraction data. It is written in Python, with some underlying routines in Fortran and C++. Conveniently, it provides the scriptable interface so that we can perform the refinement in a programming way. This is pretty handy when some batch... Mon, 25 Aug 2025 00:00:00 -0400 https://iris2020.net/2025-08-25-gsasii_scriptable/ https://iris2020.net/2025-08-25-gsasii_scriptable/