JLDC

using Plots
using Random
using Statistics

Random.seed!(1)

function eval_pol(pol, n, ε, n_rep)
    ε < 0 && error("ε must be non-negative")
    p_sel = [1, 0, -1]
    R = []

JLDC

def pd_to_latex(df, print_index=True, out_file=None, align='r', col_sep="|",
               digits=2, centering=True):
    # If out_file is specified, outputs a .tex file, otherwise, just output the contents as a string
    if len(align) == 1: # Everything is aligned the same
        align = [align for _ in range((len(df.columns) + print_index))]
    else: # Alignment fully specified, must match df length
        if len(align) != len(df.columns) + print_index:
            raise Exception("Fully specified alignment must match dataframe size")
    align = col_sep + col_sep.join(align) + col_sep
    # Add alignment, top border

JLDC

# -----------------------------------------
# Temporal Convolutional Network in Flux.jl
# Author: Jonathan Chassot, May 17, 2022
# -----------------------------------------
# Reference:
#       Shaojie Bai, J. Zico Kolter, Vladlen Koltun. (2018)
#       An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling
#       https://arxiv.org/abs/1803.01271
# -----------------------------------------
# Note that this gist uses batch normalization instead of weight normalization as in the original paper

JLDC

# All credits go to mcreel, see https://discourse.julialang.org/t/simple-flux-lstm-for-time-series/35494/42
using Flux, Plots, Statistics
using Base.Iterators

# data generating process is autoregressive order 1
# modeling objective is to forecast y conditional
# on lags of y
function AR1(n)
    y = zeros(n)
    for t = 2:n