kagaya

https://www.yomiuri.co.jp/culture/book/column/20230411-OYT8T50001/ に書かれた、太刀川氏による文章をもとにChatGPTと会話した記録

katsushi.kagaya@gmail.com

次の文章に誤りもしくは不自然な点があれば指摘してください: 　まず気候変動の緩和とは、二酸化炭素などの温室効果ガスを削減して気候変動を和らげる戦略だ。これは具体的でわかりやすい。一方で気候変動への適応は、起こった気候の変化に社会を合わせる戦略だ。うむ…わかる気もするが、一体どうすれば適応できるのか。生活の全てが適応の手段なので、指標も測りづらい。適応策はこの複雑さ故に、緩和策ほどには進んでいない。

ChatGPT

文章に誤りや不自然な点は見受けられませんが、一部表現をより明確にする提案があります。

現在の文章：

kagaya

library(tidyverse)
library(reticulate)

read_raw_pkl <- function(file_path="/path/to/raw.pkl"){
  # https://gist.github.com/kagaya/32706c38404c99431669ba34b8bccbac
  j <- import("joblib") # python package joblib is loaded via {reticulate}
  arm_xy <- j$load(file_path)
  
  tibble(t=1:length(arm_xy$path), path=arm_xy$path) %>% 
    group_by(t) %>% 

kagaya

kagaya

functions {
  vector my_student_t_rng(int N, real nu, real mu, real sigma){
    vector[N] Y;
    for (n in 1:N){
      Y[n] = student_t_rng(nu, mu, sigma);
    }
    return(Y);
  }

  vector my_ZIP_rng(int N, real q, real lambda){

kagaya

---
title: "an example of model comparison by WAIC (draft)"
author: "Katsushi Kagaya"
date: "`r Sys.time()`"
output:
  html_document: default
---

# motivation and aims

kagaya

library(tidyverse)
library(grid)
library(gtable)


plot_segment_across_facets <- function(p, from=1, to=2, 
                                       from_point_id=1, 
                                       to_point_id=1, 
                                       plotout = F,
                                       gp=gpar(lty=1, alpha=0.5)){

kagaya

waic <- function(log_likelihood) {
  # from https://gist.github.com/MatsuuraKentaro/3f6ae5863e700f5039c19e36a9bdf646
  training_error <- - mean(log(colMeans(exp(log_likelihood))))
  functional_variance_div_N <- mean(colMeans(log_likelihood^2) - colMeans(log_likelihood)^2)
  waic <- training_error + functional_variance_div_N
  return(waic)
}

kagaya

library(tidyverse)
library(ggrepel)
library(rstan)
expose_stan_functions("my_stan_functions.stan") # https://gist.github.com/kagaya/c726f16d82b80026bb1924b408a72b5c
source("utility.R") # https://gist.github.com/kagaya/60c4190ac306840daee54115b3c315c3

plot_choice_random_intercept <- function(d, fit){
  # mu[n,2] =  bias_l[ID[n]] + cwl*C_width[n] + ll*Leg_lack[n];
  # mu[n,3] =  bias_no0 + cwno*C_width[n] + lno*Leg_lack[n]; 
  ms <- rstan::extract(fit)

kagaya

library(rstan)
library(tidyverse)

rstan_options(auto_write = TRUE)
options(mc.cores = parallel::detectCores())

## preprocessing                     
d <- read_csv(file='Dromiidae.csv') # https://gist.github.com/kagaya/0d309397300b7e8294fe53c44b6fc525
d <- d %>%
  select(c('shell_width', 'choice',

kagaya

functions {
  real f(real mu, real s, real x){
      // separately performing numerical integration to make this computing fast
      return(normal_lpdf(x | mu, s));
      }
     
  real log_lik_Simpson(real mu, real s, real a, real b, int M) {
    // numerical integration using the Simpson's rule
    vector[M+1] lp;
    real h;