Skip to content

Latest commit

 

History

History
160 lines (132 loc) · 4.78 KB

utils.md

File metadata and controls

160 lines (132 loc) · 4.78 KB

Table of Contents

Precision@k
Data Shuffling
Conda Enviroment
Fix Random Seed
LaTeX Tabular to DataFrame
LaTex Tree Diagram with TiKZ

precision@k (Pytorch)

from __future__ import print_function, absolute_import

__all__ = ['accuracy']

def accuracy(output, target, topk=(1,)):
    """Computes the precision@k for the specified values of k in Pytorch"""
    maxk = max(topk)
    batch_size = target.size(0)

    _, pred = output.topk(maxk, 1, True, True)
    pred = pred.t()
    correct = pred.eq(target.view(1, -1).expand_as(pred))

    res = []
    for k in topk:
        correct_k = correct[:k].view(-1).float().sum(0)
        res.append(correct_k.mul_(100.0 / batch_size))
    return res

Shuffle (sample, label) paris

from sklearn.utils import shuffle
train_data, train_labels = shuffle(train_data, train_labels)

Create mini-batch

class Batch(object):
  def __init__(self, X, y, batch_size):
    self.batch_size = batch_size
    self.X = X
    self.y = y
    self.size = X.shape[0]
  def getBatch(self):
    indices = np.random.choice(range(self.size), self.batch_size)
    return self.X[indices], self.y[indices]

x_train = x_train.reshape([-1, 28, 28, 1])

batch_size = 512
batch = Batch(x_train, y_train, batch_size)

Anaconda create/remove environment

Tutorial

Fix random seeds (for reproducibility)

Pytorch Notes

Conversion between LaTeX Tabular and Pandas DataFrame

# -*- coding: utf-8 -*-
import pandas as pd
import numpy as np
import re

def tabular2df(filename):
    with open(filename, 'r') as fp:
        table = fp.read().splitlines()

    df = []
    for line in table:
        if line.startswith('\\'):
            continue
        if line[-1] == '\\':
            line = line[:-2]
        line = line.split('&')
        for i, cell in enumerate(line):
            if i == 0:
                continue
            regex = re.search(r"\d", cell)
            if regex == None:
                continue
            else:
                line[i] = cell[regex.start():]
        df.append(line)

    df = np.array(df)
    columns = df[0, 1:]
    index = df[1:, 0]
    data = df[1:, 1:]
    df = pd.DataFrame(data=data, columns=columns, index=index)
    return df

def calc_average_and_rank(df, column_wise=True):
    if column_wise == False:
        df = df.T

    avg = []
    for col in df.columns:
        data = df[col]
        if isinstance(data[0], str):
            data = list(map(float, data))
        avg.append(np.mean(data))
    avg = np.array(avg)
    tmp = np.argsort(avg)[::-1]
    rnk = np.arange(len(avg))
    for i, v in enumerate(tmp):
        rnk[v] = i + 1

    return avg, rnk

def highlight_max(df):
    # row-wise max
    for i in range(df.shape[0]):
        row = df.iloc[i]
        if isinstance(row[0], str):
            row = list(map(float, row))
        pos = np.argmax(row)
        df.iloc[i, pos] = '\\bf' + df.iloc[i, pos]
    return df

def df2tabular(df):
    return df.to_latex(float_format="%.4f", escape=False)

Tree Diagrame with TiKZ

\begin{figure*}
\centering
\begin{tikzpicture}[every tree node/.style={draw=black,thick,rounded corners},sibling distance=.25cm, grow = right, level distance = 8em, anchor = west, edge from parent fork right,
level 1/.style={sibling distance=3em},level 2/.style={sibling distance=2em} ]

\Tree[.XML 
[.Evaluation [.{Label propensity~\cite{jain2016propensity}} ]
[.{Coverage~\cite{babbar2017dismec}} ]
]
[.{Data scarcity}
[.{Missing label} {LEML~\cite{yu2014leml}} {Propensity scored loss~\cite{jain2016propensity}} ]
[.{Tail label} {Specific losses~\cite{jain2016propensity}} {Tail label pruning~\cite{wei2018ijcai}} {Label-specific feature~\cite{wei2019ijcai}} {Matrix decomposition~\cite{xu2016reml}} {Knowledge transfer~\cite{anonymous2020deepxml}} ]
[.{Feature sparsity} {Dense embedding~\cite{you2018attentionxml,chuanguo2019nips,anonymous2020deepxml,xbert2019}} ]
]
[.Scalability 
[.{Model size} {Sparse regularizer~\cite{yen2016pd,yen2017ppd}} {Weight pruning~\cite{babbar2017dismec}} ] 
[.{Inference time} {Tree-based methods~\cite{Prabhu2014fastxml,jain2016propensity,khandagale2019bonsai,you2018attentionxml}} {Label filter~\cite{niculescu2016labelfilter}} ]
[.{Training time} {Parallelization~\cite{babbar2017dismec}} {Label partitioning~\cite{Prabhu2014fastxml,prabhu2018parabel}} {GPU~\cite{xbert2019,chuanguo2019nips}} ]
]
]
\end{tikzpicture}
\end{figure*}

back to top