PyTorch Linear Regression

Source

Table of Contents

Beginning

Imports

# python
from collections import namedtuple
from functools import partial

# pypi
from torch import nn
from torch.utils.data import Dataset, DataLoader

import hvplot.pandas
import numpy
import pandas

# local stuff
from graeae import EmbedHoloviews

Set Up

random_generator = numpy.random.default_rng(seed=2021)

slug = "pytorch-linear-regression"
Embed = partial(EmbedHoloviews, folder_path=f"files/posts/pytorch/{slug}")

Plot = namedtuple("Plot", ["width", "height", "fontscale", "tan", "blue", "red"])
PLOT = Plot(
    width=900,
    height=750,
    fontscale=2,
    tan="#ddb377",
    blue="#4687b7",
    red="#ce7b6d",
 )

def sample(start: float, stop: float, shape: tuple, uniform: bool=True) -> numpy.ndarray:
    """Create a random sample

    Args:
     start: lowest allowed value
     stop: highest allowed value
     shape: shape for the final array (just an int for single values)
     uniform: use the uniform distribution instead of the standard normal
    """
    if uniform:
        return (stop - start) * random_generator.random(shape) + start
    return (stop - start) * random_generator.standard_normal(shape) + start

Middle

SAMPLES = 200
X_RANGE = 5
x_values = sample(-X_RANGE, X_RANGE, SAMPLES)
SLOPE = sample(-5, 5, 1)
INTERCEPT = sample(-5, 5, 1)
noise = sample(-2, 2, SAMPLES, uniform=False)
y_values = SLOPE * x_values + INTERCEPT + noise

data_frame = pandas.DataFrame.from_dict(dict(X=x_values, Y=y_values))
first, last = x_values.min(), x_values.max()
line_frame = pandas.DataFrame.from_dict(
    dict(X=[first, last],
         Y=[SLOPE * first + INTERCEPT,
            SLOPE * last + INTERCEPT]))
line_plot = line_frame.hvplot(x="X", y="Y", color=PLOT.blue)
data_plot = data_frame.hvplot.scatter(x="X", y="Y", title="Sample Data",
                                      color=PLOT.tan)
plot = (data_plot * line_plot).opts(
    height=PLOT.height,
    width=PLOT.width,
    fontscale=PLOT.fontscale
)
output = Embed(plot=plot, file_name="sample_data")()

Figure Missing

 
class XY(Dataset):
    def __init__(self, x, y):
        self.x = x
        self.y = y
        return

    def __len__(self):
        return len(self.x)

    def __getitem__(self, index):
        if torch.is_tensor(index):
            index = index.tolist()

        return {"x": self.x[index], "y": self.y[index]}

dataset = XY(x_values, y_values)
loader = DataLoader(dataset, batch_size=4)

model = nn.Linear(1, 1)