calvinmccarter

\begin{table}[ht]
\centering
\begin{tabular}{rlllllllll}
  \hline
 & MinMAE & AvgMAE & W\_train & W\_test & MedianMAD & R2\_imp & F1\_imp & PercentBias & CoverageRate \\ 
  \hline
KNN & 0.16 (0.03) & 0.16 (0.03) & 0.42 (0.08) & 1.89 (0.49) & 0 (0) & 0.59 (0.09) & 0.75 (0.04) & 1.27 (0.25) & 0.4 (0.11) \\ 
  ICE & 0.1 (0.01) & 0.21 (0.03) & 0.52 (0.09) & 1.99 (0.49) & 0.69 (0.1) & 0.59 (0.09) & 0.74 (0.04) & 1.05 (0.29) & 0.39 (0.09) \\ 
  MICE-Forest & 0.08 (0.02) & 0.13 (0.03) & 0.34 (0.07) & 1.86 (0.48) & 0.29 (0.08) & 0.61 (0.1) & 0.76 (0.04) & 0.61 (0.2) & 0.75 (0.11) \\ 
  MissForest & 0.1 (0.03) & 0.12 (0.03) & 0.32 (0.07) & 1.85 (0.48) & 0.1 (0.03) & 0.61 (0.1) & 0.76 (0.04) & 0.62 (0.22) & 0.79 (0.08) \\ 

calvinmccarter

calvinmccarter

calvinmccarter

calvinmccarter

import numpy as np

def top_p_sampling(n_bins, probs, rng, top_p):
    """A modified implementation of nucleus sampling.
    For the class straddling the top_p boundary, the probability mass beyond top_p is discarded.
    But this class does not receive zero probability mass, so it differs
    from https://gist.github.com/thomwolf/1a5a29f6962089e871b94cbd09daf317 .
    This is more mathematically elegant, in my humble opinion.

    Parameters

calvinmccarter

calvinmccarter

calvinmccarter

calvinmccarter

Keybase proof

I hereby claim:

• I am calvinmccarter on github.
• I am calvinm (https://keybase.io/calvinm) on keybase.
• I have a public key ASD-RPoYtblA-GAI290ob6aspmx8D7O5Dj0jO9b1tkdmjAo

To claim this, I am signing this object:

calvinmccarter