Skip to content

Instantly share code, notes, and snippets.

David James xpe

  • Berkeley, CA
Block or report user

Report or block xpe

Hide content and notifications from this user.

Learn more about blocking users

Contact Support about this user’s behavior.

Learn more about reporting abuse

Report abuse
View GitHub Profile
xpe /
Last active Jul 11, 2018
Pros and Cons of Wrapper Types in Rust

Pros and Cons of Wrapper Types in Rust

I've written a mathematically-oriented and computatioally-intensive application in Rust. While the mathematical theory is not necessarily difficult, keeping the various quantities clear required some care.

For this reason, I'm refactoring it to clear it up. I've created some custom structs to wrap various mathematical types.

For example, I created a type to write PDFs (probability density functions)


Various arguments for wiring your home with Cat 6 (gigabit) ethernet or better.

Bottom-line advantages of Cat 6 over 802.11ac WiFi:

  • Better capacity. While 802.11ac, under optimal conditions, can approach 500 Mbps, having a wired connection that guarantees 1000 Mbps can be useful.
  • Better reliability.
  • Better security.
  • Better latency.

All of this said, it isn't either-or -- you can use your wired connections for connecting computers to NAS or when you need maximum speed for uploading.

xpe /
Last active Dec 23, 2018
Policy Gradient Loss function for PyTorch
import torch
import torch.nn as nn
import torch.optim as optim
from torch._jit_internal import weak_module, weak_script_method
class PolicyGradientLoss(nn.Module):
Multiplies an unreduced CrossEntropyLoss by a `q` vector.
xpe /
Last active Oct 7, 2019
Use this when your phrases need numeronyms!
def abbreviate(phrase):
def shorten(w, t=7):
if len(w) <= t:
return w
return "{}{}{}".format(w[0], len(w)-2, w[-1])
return " ".join([shorten(w) for w in phrase.split(" ")])
abbreviate("The Heilmeier Catechism demands a thoughtful approach to risk taking")
# 'The H7r C7m demands a t8l a6h to risk taking'
xpe /
Created Oct 27, 2019
Machine Leaning Theory: Double Descent Curve

Neural networks are capable of interpolating (fitting the training set perfectly) and driving test error lower.

Can the AdaBoost algorithm also do this?

Why or why not?

xpe /
Created Oct 29, 2019
Let's retire Lines of Code and just count characters instead!
You can’t perform that action at this time.