All Posts
Posts here include English/Japanese/Chinese posts.
My TLDR for Tsubame 4 @ISCT
A TLDR-like manual of Tsubame (Altair) tailored only to myself being used to Slurm.MDP & Timestep RL for Discrete Diffusion LLMs
A primitive formulation of dLLM RL with denoising timesteps as actions.Training-free Verifiable Process Reward for LLM Reinforcement Finetuning
A training-free process reward system based on formalized theorem proving for LLM reinforcement finetuning.
INFOTH Note 23: Parallel Gaussian Channels, WSS, Distributed Source Compression
Parallel Gaussian Channel Model, Water-Filling Theorem, Szegő Theorem for Colored Noise, Distributed Source Compression and Slepian-Wolf TheoremINFOTH Note 22: AWGN Channel and Shannon Limit
Waveform Channel Model, Additive White Gaussian Noise with its Shannon Theorem, and Spectral EfficiencyINFOTH Note 21: Polar Codes and Related Theories (Bhattacharyya, Martingale)
Polar Codes, Bhattacharyya Parameter, Martingale Theory, and Channel PolarizationINFOTH Note 20: Channel Coding Schemes
Maximum Likelihood Decoding, Block Coding (Binary Linear, Hamming, etc.) & Polar CodesINFOTH Note 19: Channel Coding Theorem for DMC 2
Proof of Shannon's Channel Coding Theorem, Converse TheoremINFOTH Note 18: Channel Coding Theorem for DMC 1
Discrete Memoryless Channel, Shannon's Channel Coding Theorem (2nd)Paper Reading: SEER (Structured Reasoning and Explanation via RL)
SEER: Facilitating Structured Reasoning and Explanation via Reinforcement Learning.
INFOTH Note 9: Differential Entropy
Differential Entropy (for Real-valued/Continuous R.V.s)INFOTH Note 8: BW Transform
Burrows-Wheeler TransformINFOTH Note 7: Huffman & Arithmetic Coding
Huffman Coding and Arithmetic CodingINFOTH Note 6: Kraft's Inequality, Shannon Code
Codes, Kraft's Inequality, Shannon CodeINFOTH Note 5: Fano Ineq., Entropy Rate, AEP
Fano's Inequality, Markov Chain, Entropy Rate, Asymptotic Equipartition Property and Lossless CompressionINFOTH Note 4: Cond. MI & Cond. KL, Data Processing Ineq.
Conditional Mutual Information, Conditional Relative Entropy, Chain Rule for these things, Properties, and Data Processing InequalityINFOTH Note 3: Relative Ent. (KL), Mutual Info.
Log-sum Ineq., Relative Intropy (aka. KL Divergence) and Mutual InformationINFOTH Note 2: Joint & Conditional Entropy, Chain Rule
Definition of Many Entropies and the Chain Rule. We have a specific def for joint entropy.INFOTH Note 1: Axioms & Def of Entropy
Shannon Entropy, Axioms, Entropy and ConcavityReconstruction and Re-evaluation of SFCNN for PLA Scoring
A PyTorch reconstruction and benchmark for protein-ligand binding affinity prediction with SFCNN.
Paper Reading: Pushdown Layers
The paper reading note for Pushdown Layers.
Facial Keypoint Detection by Regression & Pixelwise Classification
CS180 Final Project 1/3
From IMM Face Database (regression) and iBUG dataset (Gaussian heatmap & Pixelwise classification). (CS280 FA22 Project 5 & CS180 Final Project 1/3)
[Flow] Rectified Flow Explained
Reading Flow Matching Across Domains Vol. 1
A simple explanation for the Rectified Flow framework.
2024-11-20
Updated: 2026-10-10
5 min read
DDPM Sampling, Applications & Training from Scratch
CS180 Project 5
A project of sampling and playing with a pretrained diffusion model; and training DDPMs from scratch on MNIST.
CS180: Project Webpage Catalog
The Catalog of Project Showcase Webpages for 2024 Fall UC Berkeley COMPSCI 180 Computer VisionLLM-powered Lecture Generation
An LLM-agent system for generating lectures from PDF materials and supporting long-context question answering.
[CUDA] Transpose Optimization in CUDA
TODO
2024-07-19
1 min read
[2024 Summer GeekPie HPC Lecture Notes] Week 3: Optimization
TODO
2024-07-19
1 min read
[2024 Summer GeekPie HPC Lecture Notes] Week 3: Profiling
TODO
2024-07-19
1 min read