markdown stringlengths 0 1.02M | code stringlengths 0 832k | output stringlengths 0 1.02M | license stringlengths 3 36 | path stringlengths 6 265 | repo_name stringlengths 6 127 |
|---|---|---|---|---|---|
Performance Analysis One of the advantages of theano is the posibility to create a full profile of the function. This has to be included in at the time of the creation of the function. At the moment it should be active (the downside is larger compilation time and I think also a bit in the computation so be careful if ... | %%timeit
# Compute the block
GeMpy.compute_block_model(geo_data, [0,1,2], verbose = 0)
geo_data.interpolator._interpolate.profile.summary() | Function profiling
==================
Message: ../GeMpy/DataManagement.py:994
Time in 3 calls to Function.__call__: 8.400567e-01s
Time in Function.fn.__call__: 8.395956e-01s (99.945%)
Time in thunks: 8.275988e-01s (98.517%)
Total compile time: 3.540267e+00s
Number of Apply nodes: 342
Theano Optimizer ... | MIT | Prototype Notebook/Example_1_Sandstone.ipynb | nre-aachen/gempy |
Ungraded Lab: Activation in Custom LayersIn this lab, we extend our knowledge of building custom layers by adding an activation parameter. The implementation is pretty straightforward as you'll see below. Imports | try:
# %tensorflow_version only exists in Colab.
%tensorflow_version 2.x
except Exception:
pass
import tensorflow as tf
from tensorflow.keras.layers import Layer | _____no_output_____ | MIT | C1W3_L3_CustomLayerWithActivation.ipynb | 100rab-S/TensorFlow-Advanced-Techniques |
Adding an activation layerTo use the built-in activations in Keras, we can specify an `activation` parameter in the `__init__()` method of our custom layer class. From there, we can initialize it by using the `tf.keras.activations.get()` method. This takes in a string identifier that corresponds to one of the [availab... | class SimpleDense(Layer):
# add an activation parameter
def __init__(self, units=32, activation=None):
super(SimpleDense, self).__init__()
self.units = units
# define the activation to get from the built-in activation layers in Keras
self.activation = tf.keras.activatio... | _____no_output_____ | MIT | C1W3_L3_CustomLayerWithActivation.ipynb | 100rab-S/TensorFlow-Advanced-Techniques |
We can now pass in an activation parameter to our custom layer. The string identifier is mostly the same as the function name so 'relu' below will get `tf.keras.activations.relu`. | mnist = tf.keras.datasets.mnist
(x_train, y_train),(x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(input_shape=(28, 28)),
SimpleDense(128, activation='relu'),
tf.keras.layers.Dropout(0.2),
tf.keras.laye... | Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz
11493376/11490434 [==============================] - 0s 0us/step
Epoch 1/5
1875/1875 [==============================] - 5s 2ms/step - loss: 0.4861 - accuracy: 0.8560
Epoch 2/5
1875/1875 [==============================] - 4s 2ms/... | MIT | C1W3_L3_CustomLayerWithActivation.ipynb | 100rab-S/TensorFlow-Advanced-Techniques |
Using Google Cloud Functions to support event-based triggering of Cloud AI Platform Pipelines> This post shows how you can run a Cloud AI Platform Pipeline from a Google Cloud Function, providing a way for Pipeline runs to be triggered by events.- toc: true - badges: true- comments: true- categories: [ml, pipelines, m... | %env TRIGGER_BUCKET=REPLACE_WITH_YOUR_GCS_BUCKET_NAME | _____no_output_____ | Apache-2.0 | _notebooks/2020-05-12-hosted_kfp_gcf.ipynb | amygdala/fastpages |
Give Cloud Function's service account the necessary accessFirst, make sure the Cloud Function API [is enabled](https://console.cloud.google.com/apis/library/cloudfunctions.googleapis.com?q=functions).Cloud Functions uses the project's 'appspot' acccount for its service account. It will have the form: `PROJECT_ID@apps... | %%bash
mkdir -p functions | _____no_output_____ | Apache-2.0 | _notebooks/2020-05-12-hosted_kfp_gcf.ipynb | amygdala/fastpages |
We'll first create a `requirements.txt` file, to indicate what packages the GCF code requires to be installed. (We won't actually need `kfp` for this first 'sanity check' version of a GCF function, but we'll need it below for the second function we'll create, that deploys a pipeline). | %%writefile functions/requirements.txt
kfp | _____no_output_____ | Apache-2.0 | _notebooks/2020-05-12-hosted_kfp_gcf.ipynb | amygdala/fastpages |
Next, we'll create a simple GCF function in the `functions/main.py` file: | %%writefile functions/main.py
import logging
def gcs_test(data, context):
"""Background Cloud Function to be triggered by Cloud Storage.
This generic function logs relevant data when a file is changed.
Args:
data (dict): The Cloud Functions event payload.
context (google.cloud.functions.Context):... | _____no_output_____ | Apache-2.0 | _notebooks/2020-05-12-hosted_kfp_gcf.ipynb | amygdala/fastpages |
Deploy the GCF function as follows. (You'll need to **wait a moment or two for output of the deployment to display in the notebook**). You can also run this command from a notebook terminal window in the `functions` subdirectory. | %%bash
cd functions
gcloud functions deploy gcs_test --runtime python37 --trigger-resource ${TRIGGER_BUCKET} --trigger-event google.storage.object.finalize | _____no_output_____ | Apache-2.0 | _notebooks/2020-05-12-hosted_kfp_gcf.ipynb | amygdala/fastpages |
After you've deployed, test your deployment by adding a file to the specified `TRIGGER_BUCKET`. You can do this easily by visiting the **Storage** panel in the Cloud Console, clicking on the bucket in the list, and then clicking on **Upload files** in the bucket details view.Then, check in the logs viewer panel (https:... | %%bash
cd functions
mv main.py main.py.bak | _____no_output_____ | Apache-2.0 | _notebooks/2020-05-12-hosted_kfp_gcf.ipynb | amygdala/fastpages |
Then, **before executing the next cell**, **edit the `HOST` variable** in the code below. You'll replace `` with the correct value for your installation.To find this URL, visit the [Pipelines panel](https://console.cloud.google.com/ai-platform/pipelines/) in the Cloud Console. From here, you can find the URL by click... | %%writefile functions/main.py
import logging
import datetime
import logging
import time
import kfp
import kfp.compiler as compiler
import kfp.dsl as dsl
import requests
# TODO: replace with your Pipelines endpoint URL
HOST = 'https://<your_endpoint>.pipelines.googleusercontent.com'
@dsl.pipeline(
name='Seque... | _____no_output_____ | Apache-2.0 | _notebooks/2020-05-12-hosted_kfp_gcf.ipynb | amygdala/fastpages |
Next, deploy the new GCF function. As before, **it will take a moment or two for the results of the deployment to display in the notebook**. | %%bash
cd functions
gcloud functions deploy hosted_kfp_test --runtime python37 --trigger-resource ${TRIGGER_BUCKET} --trigger-event google.storage.object.finalize | _____no_output_____ | Apache-2.0 | _notebooks/2020-05-12-hosted_kfp_gcf.ipynb | amygdala/fastpages |
Herramientas Estadisticas Contenido:1.Estadistica: - Valor medio. - Mediana. - Desviacion estandar. 2.Histogramas: - Histrogramas con python. - Histogramas con numpy. - Como normalizar un histograma. 3.Distribuciones: - Como obtener una distribucion a partir de un histograma. ... | import matplotlib.pyplot as plt
import numpy as np
# %pylab inline
def mi_mediana(lista):
x = sorted(lista)
d = int(len(x)/2)
if(len(x)%2==0):
return (x[d-1] + x[d])*0.5
else:
return x[d-1]
x_input = [1,3,4,5,5,7,7,6,8,6]
mi_mediana(x_input)
print(mi_mediana(x_input) == np.median... | True
| MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
Problemas de no saber estadísticaEste tipo de conceptos parecen sencillos. Pero no siempre son claros para todo el mundo. | x = np.arange(1, 12)
y = np.random.random(11)*10
plt.figure(figsize=(12, 5))
fig = plt.subplot(1, 2, 1)
plt.scatter(x, y, c='purple', alpha=0.8, s=60)
y_mean = np.mean(y)
y_median = np.median(y)
plt.axhline(y_mean, c='g', lw=3, label=r"$\rm{Mean}$")
plt.axhline(y_median, c='r', lw=3, label=r"$\rm{Median}$")
plt.legend(... | [9.33745032 0.46206052 3.07349261 8.65709198 6.44733954 2.5552359
8.93987727 8.24695437 5.62111292 4.64621772 0.05366015]
| MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
Desviacion estandarEs el promedio de las incertidumbres de las mediciones $x_i$$\sigma = \sqrt{\dfrac{1}{n-1} \sum(x_{i} - \bar{x})^2}$Donde $n$ es el número de la muestraAdicionalmente la ${\bf{varianza}}$ se define como:$\bar{x^2} - \bar{x}^{2}$$\sigma^2 = \dfrac{1}{N} \sum(x_{i} - \bar{x})^2$Y es una medida similar... | x = np.arange(1, 12)
y = np.random.random(11)*10
plt.figure(figsize=(9, 5))
y_mean = np.mean(y)
y_median = np.median(y)
plt.axhline(y_mean, c='g', lw=3, label=r"$\rm{Mean}$")
plt.axhline(y_median, c='r', lw=3, label=r"$\rm{Median}$")
sigma_y = np.std(y)
plt.axhspan(y_mean-sigma_y, y_mean + sigma_y, facecolor='g', alpha... | Variancia = 7.888849132964844
Desviacion estandar = 2.8087095138096507
| MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
Referencias: Para mas funciones estadisticas que se pueden usar en python ver: - NumPy: http://docs.scipy.org/doc/numpy/reference/routines.statistics.html- SciPy: http://docs.scipy.org/doc/scipy/reference/stats.html Histogramas 1. histhist es una funcion de python que genera un histograma a partir de un array ... | x = np.random.random(200)
plt.subplot(2,2,1)
plt.title("A simple hist")
h = plt.hist(x)
plt.subplot(2,2,2)
plt.title("bins")
h = plt.hist(x, bins=20)
plt.subplot(2,2,3)
plt.title("alpha")
h = plt.hist(x, bins=20, alpha=0.6)
plt.subplot(2,2,4)
plt.title("histtype")
h = plt.hist(x, bins=20, alpha=0.6, histtype='stepfille... | _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
2. Numpy-histogram | N, bins = np.histogram(caras, bins=15)
plt.plot(bins[0:-1], N) | _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
Histogramas 2D | x = np.random.random(500)
y = np.random.random(500)
plt.subplot(4, 2, 1)
plt.hexbin(x, y, gridsize=15, cmap="gray")
plt.colorbar()
plt.subplot(4, 2, 2)
data = plt.hist2d(x, y, bins=15, cmap="binary")
plt.colorbar()
plt.subplot(4, 2, 3)
plt.hexbin(x, y, gridsize=15)
plt.colorbar()
plt.subplot(4, 2, 4)
data = plt.hist2d... | _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
Como normalizar un histograma. Normalizar un histograma significa que la integral del histograma sea 1. | x = np.random.random(10)*4
plt.title("Como no normalizar un histograma", fontsize=25)
h = plt.hist(x, normed="True")
print ("El numero tamaño del bin debe de ser de la unidad")
plt.title("Como normalizar un histograma", fontsize=25)
h = hist(x, normed="True", bins=4)
| _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
Cual es la probabilidad de sacar 9 veces cara en 10 lanzamientos? Distribución de Probabilidad:Las distribuciones de probabilidad dan información de cual es la probabilidad de que una variable aleatoria $x$ aprezca en un intervalo dado. ¿Si tenemos un conjunto de datos como podemos conocer la distribucion de probabili... | x = np.random.random(100)*10
plt.subplot(1, 2, 1)
h = plt.hist(x)
plt.subplot(1, 2, 2)
histo, bin_edges = np.histogram(x, density=True)
plt.bar(bin_edges[:-1], histo, width=1)
plt.xlim(min(bin_edges), max(bin_edges)) | _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
Distribución Normal: Descripcion Matemática.$f(x, \mu, \sigma) = \dfrac{1}{\sigma \sqrt(2\pi)} e^{-\dfrac{(x-\mu)^2}{2\sigma^2}} $donde $\sigma$ es la desviacion estandar y $\mu$ la media de los datos $x$Es una función de distribucion de probabilidad que esta totalmente determinada por los parametros $\mu$ y $\sigma$.... | import scipy.stats
x = np.linspace(0, 1, 100)
n_dist = scipy.stats.norm(0.5, 0.1)
plt.plot(x, n_dist.pdf(x)) | _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
Podemos generar numeros aleatorios con una distribucion normal: | x = np.random.normal(0.0, 1.0, 1000)
y = np.random.normal(0.0, 2.0, 1000)
w = np.random.normal(0.0, 3.0, 1000)
z = np.random.normal(0.0, 4.0, 1000)
histo = plt.hist(z, alpha=0.2, histtype="stepfilled", color='r')
histo = plt.hist(w, alpha=0.4, histtype="stepfilled", color='b')
histo = plt.hist(y, alpha=0.6, histtype="s... | _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
**Intervalo de confianza**$\sigma_1$ = 68% de los datos van a estar dentro de 1$\sigma$$\sigma_2$ = 95% de los datos van a estar dentro de 2$\sigma$$\sigma_3$ = 99.7% de los datos van a estar dentro de 3$\sigma$ Ejercicio: Generen distribuciones normales con:- $\mu = 5$ y $\sigma = 2$ - $\mu = -3$ y $\sigma = -2$- $\m... | def coinflip(N):
cara = 0
sello = 0
i=0
while i < N:
x = np.random.randint(0, 10)/5.0
if x >= 1.0:
cara+=1
elif x<1.0:
sello+=1
i+=1
return cara/N, sello/N | _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
Función que hace M veces N lanzamientos. | def realizaciones(M, N):
caras=[]
for i in range(M):
x, y = coinflip(N)
caras.append(x)
return caras
hist(caras, normed=True, bins=20)
caras = realizaciones(100000, 30.) | _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
PDF | N, bins = np.histogram(x, density=True)
plt.plot(bins[0:-1], N) | _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
CDF | h = plt.hist(x, cumulative=True, bins=20) | _____no_output_____ | MIT | Notebooks/python 9 - Herramientas_Estadisticas.ipynb | diegour1/HerramientasComputacionales |
01 - Sentence Classification Model Building Parse & clearn labeled training data | import xml.etree.ElementTree as ET
tree = ET.parse('../data/Restaurants_Train.xml')
root = tree.getroot()
root
# Use this dataframe for multilabel classification
# Must use scikitlearn's multilabel binarizer
labeled_reviews = []
for sentence in root.findall("sentence"):
entry = {}
aterms = []
aspects = []
... | _____no_output_____ | Apache-2.0 | 04-Aspect_Based_Opinion_Mining/code/01-Build_Model.ipynb | ayan1995/DS_projects |
Training the model with Naive Bayes1. replace pronouns with neural coref2. train the model with naive bayes | from neuralcoref import Coref
import en_core_web_lg
spacy = en_core_web_lg.load()
coref = Coref(nlp=spacy)
# Define function for replacing pronouns using neuralcoref
def replace_pronouns(text):
coref.one_shot_coref(text)
return coref.get_resolved_utterances()[0]
# Read annotated reviews df, which is the labele... | _____no_output_____ | Apache-2.0 | 04-Aspect_Based_Opinion_Mining/code/01-Build_Model.ipynb | ayan1995/DS_projects |
At this point, we can move on to 02-Sentiment analysis notebook, which will load the fitted Naive bayes model. | #mlb.inverse_transform(predicted)
pred_df = pd.DataFrame(
{'text_pro': X_test,
'pred_category': mlb.inverse_transform(predicted)
})
pd.set_option('display.max_colwidth', -1)
pred_df.head() | _____no_output_____ | Apache-2.0 | 04-Aspect_Based_Opinion_Mining/code/01-Build_Model.ipynb | ayan1995/DS_projects |
Some scrap code below which wasn't used | # Save annotated reviews
labeled_df.to_pickle("annotated_reviews_df.pkl")
labeled_df.head()
# This code was for parsing out terms & their relations to aspects
# However, the terms were not always hyponyms of the aspects, so they were unusable
aspects = {"food":[],"service":[],"anecdotes/miscellaneous":[], "ambience":[]... | _____no_output_____ | Apache-2.0 | 04-Aspect_Based_Opinion_Mining/code/01-Build_Model.ipynb | ayan1995/DS_projects |
Network inference of categorical variables: non-sequential data | import sys
import numpy as np
from scipy import linalg
from sklearn.preprocessing import OneHotEncoder
import matplotlib.pyplot as plt
%matplotlib inline
import inference
import fem
# setting parameter:
np.random.seed(1)
n = 20 # number of positions
m = 5 # number of values at each position
l = int(((n*m)**2)) # num... | _____no_output_____ | MIT | old_versions/1main-v4-MCMC-symmetry.ipynb | danhtaihoang/categorical-variables |
1. You are provided the titanic dataset. Load the dataset and perform splitting into training and test sets with 70:30 ratio randomly using test train split.2. Use the Logistic regression created from scratch (from the prev question) in this question as well.3. Data cleaning plays a major role in this question. Report ... | import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, f1_score,confusion_matrix,r2_score
sns.set(style="darkgrid")
df = pd.read_csv('titanic.csv')
df.head()
print('Missing Values in th... | _____no_output_____ | MIT | question3.ipynb | kanishk779/SMAI-2 |
Data cleaning1. **Removal** :-- Remove *Name* column as this attribute does not affect the *Survived* status of the passenger. And moreover we can see that each person has a unique name hence there is no point considering this column.- Remove *Ticket* because there are 681 unique values of ticket and moreover if there... | df = df.drop(columns=['Name', 'Ticket', 'Cabin', 'PassengerId'])
s1 = sns.barplot(data = df, y='Survived' , hue='Sex' , x='Sex')
s1.set_title('Male-Female Survival')
plt.show() | _____no_output_____ | MIT | question3.ipynb | kanishk779/SMAI-2 |
Females had a better survival rate than male. | sns.pairplot(df, hue='Survived') | _____no_output_____ | MIT | question3.ipynb | kanishk779/SMAI-2 |
Categorical dataFor categorical variables where no ordinal relationship exists, the integer encoding may not be enough, at best, or misleading to the model at worst.Forcing an ordinal relationship via an ordinal encoding and allowing the model to assume a natural ordering between categories may result in poor performa... | from numpy import mean
s1 = sns.barplot(data = df, y='Survived' , hue='Embarked' , x='Embarked', estimator=mean)
s1.set_title('Survival vs Boarding place')
plt.show()
carrier_count = df['Embarked'].value_counts()
sns.barplot(x=carrier_count.index, y=carrier_count.values, alpha=0.9)
plt.title('Frequency Distribution of... | precision recall f1-score support
0.0 0.82 0.86 0.84 153
1.0 0.80 0.75 0.77 115
accuracy 0.81 268
macro avg 0.81 0.80 0.80 268
weighted avg 0.81 0.81 0.81 ... | MIT | question3.ipynb | kanishk779/SMAI-2 |
起手式，導入 numpy, matplotlib | from PIL import Image
import numpy as np
%matplotlib inline
import matplotlib
import matplotlib.pyplot as plt
matplotlib.style.use('bmh')
matplotlib.rcParams['figure.figsize']=(8,5) | _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
使用之前下載的 mnist 資料，載入訓練資料 `train_set` 和測試資料 `test_set` | import gzip
import pickle
with gzip.open('../Week02/mnist.pkl.gz', 'rb') as f:
train_set, validation_set, test_set = pickle.load(f, encoding='latin1')
train_X, train_y = train_set
validation_X, validation_y = validation_set
test_X, test_y = test_set | _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
之前的看圖片函數 | from IPython.display import display
def showX(X):
int_X = (X*255).clip(0,255).astype('uint8')
# N*784 -> N*28*28 -> 28*N*28 -> 28 * 28N
int_X_reshape = int_X.reshape(-1,28,28).swapaxes(0,1).reshape(28,-1)
display(Image.fromarray(int_X_reshape))
# 訓練資料， X 的前 20 筆
showX(train_X[:20]) | _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
train_set 是用來訓練我們的模型用的我們的模型是很簡單的 logistic regression 模型，用到的參數只有一個 784x10 的矩陣 W 和一個長度 10 的向量 b。我們先用均勻隨機亂數來設定 W 和 b 。 | W = np.random.uniform(low=-1, high=1, size=(28*28,10))
b = np.random.uniform(low=-1, high=1, size=10)
| _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
完整的模型如下將圖片看成是長度 784 的向量 x計算 $Wx+b$， 然後再取 $exp$。 最後得到的十個數值。將這些數值除以他們的總和。我們希望出來的數字會符合這張圖片是這個數字的機率。 $ \Pr(Y=i|x, W, b) = \frac {e^{W_i x + b_i}} {\sum_j e^{W_j x + b_j}}$ 先拿第一筆資料試試看， x 是輸入。 y 是這張圖片對應到的數字(以這個例子來說 y=5)。 | x = train_X[0]
y = train_y[0]
showX(x)
y | _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
先計算 $e^{Wx+b} $ | Pr = np.exp(x @ W + b)
Pr.shape | _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
然後 normalize，讓總和變成 1 （符合機率的意義） | Pr = Pr/Pr.sum()
Pr | _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
由於 $W$ 和 $b$ 都是隨機設定的，所以上面我們算出的機率也是隨機的。正確解是 $y=5$， 運氣好有可能猜中為了要評斷我們的預測的品質，要設計一個評斷誤差的方式，我們用的方法如下（不是常見的方差，而是用熵的方式來算，好處是容易微分，效果好） $ loss = - \log(\Pr(Y=y|x, W,b)) $ 上述的誤差評分方式，常常稱作 error 或者 loss，數學式可能有點費解。實際計算其實很簡單，就是下面的式子 | loss = -np.log(Pr[y])
loss | _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
想辦法改進。 我們用一種被稱作是 gradient descent 的方式來改善我們的誤差。因為我們知道 gradient 是讓函數上升最快的方向。所以我們如果朝 gradient 的反方向走一點點（也就是下降最快的方向），那麼得到的函數值應該會小一點。記得我們的變數是 $W$ 和 $b$ (裡面總共有 28*20+10 個變數)，所以我們要把 $loss$ 對 $W$ 和 $b$ 裡面的每一個參數來偏微分。還好這個偏微分是可以用手算出他的形式，而最後偏微分的式子也不會很複雜。 $loss$ 展開後可以寫成$loss = \log(\sum_j e^{W_j x + b_j}) - W_i x - b_i$ 對 $k \neq ... | gradb = Pr.copy()
gradb[y] -= 1
print(gradb) | [ 1.11201478e-03 2.32129668e-06 3.47186834e-03 3.64416088e-03
9.89922844e-01 -9.99616538e-01 4.67890738e-09 3.02581069e-04
1.11720864e-07 1.16063080e-03]
| MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
對 $W$ 的偏微分也不難 對 $k \neq i$ 時, $loss$ 對 $W_{k,t}$ 的偏微分是 $$ \frac{e^{W_k x + b_k} W_{k,t} x_t}{\sum_j e^{W_j x + b_j}} = \Pr(Y=k | x, W, b) x_t$$對 $k = i$ 時, $loss$ 對 $W_{k,t}$ 的偏微分是 $$ \Pr(Y=k | x, W, b) x_t - x_t$$ | print(Pr.shape, x.shape, W.shape)
gradW = x.reshape(784,1) @ Pr.reshape(1,10)
gradW[:, y] -= x | (10,) (784,) (784, 10)
| MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
算好 gradient 後，讓 W 和 b 分別往 gradient 反方向走一點點，得到新的 W 和 b | W -= 0.1 * gradW
b -= 0.1 * gradb | _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
再一次計算 $\Pr$ 以及 $loss$ | Pr = np.exp(x @ W + b)
Pr = Pr/Pr.sum()
loss = -np.log(Pr[y])
loss | _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
Q* 看看 Pr ， 然後找出機率最大者， predict y 值* 再跑一遍上面程序，看看誤差是否變小？* 拿其他的測試資料來看看，我們的 W, b 學到了什麼？ 我們將同樣的方式輪流對五萬筆訓練資料來做，看看情形會如何 | W = np.random.uniform(low=-1, high=1, size=(28*28,10))
b = np.random.uniform(low=-1, high=1, size=10)
score = 0
N=50000*20
d = 0.001
learning_rate = 1e-2
for i in range(N):
if i%50000==0:
print(i, "%5.3f%%"%(score*100))
x = train_X[i%50000]
y = train_y[i%50000]
Pr = np.exp( x @ W +b)
Pr = Pr... | 0 0.000%
50000 87.490%
100000 89.497%
150000 90.022%
200000 90.377%
250000 90.599%
300000 91.002%
350000 91.298%
400000 91.551%
450000 91.613%
500000 91.678%
550000 91.785%
600000 91.792%
650000 91.889%
700000 91.918%
750000 91.946%
800000 91.885%
850000 91.955%
900000 91.954%
950000 92.044%
| MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
結果發現正確率大約是 92%， 但這是對訓練資料而不是對測試資料而且，一筆一筆的訓練資也有點慢，線性代數的特點就是能夠向量運算。如果把很多筆 $x$ 當成列向量組合成一個矩陣（然後叫做 $X$），由於矩陣乘法的原理，我們還是一樣計算 $WX+b$ ， 就可以同時得到多筆結果。下面的函數，可以一次輸入多筆 $x$， 同時一次計算多筆 $x$ 的結果和準確率。 | def compute_Pr(X):
Pr = np.exp(X @ W + b)
return Pr/Pr.sum(axis=1, keepdims=True)
def compute_accuracy(Pr, y):
return (Pr.argmax(axis=1)==y).mean() | _____no_output_____ | MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
下面是更新過得訓練過程， 當 i%100000 時，順便計算一下 test accuracy 和 valid accuracy。 | %%timeit -r 1 -n 1
def compute_Pr(X):
Pr = np.exp(X @ W + b)
return Pr/Pr.sum(axis=1, keepdims=True)
def compute_accuracy(Pr, y):
return (Pr.argmax(axis=1)==y).mean()
W = np.random.uniform(low=-1, high=1, size=(28*28,10))
b = np.random.uniform(low=-1, high=1, size=10)
score = 0
N=20000
batch_size = 128
lea... | 2000 90.50% 90.47%
4000 91.17% 91.56%
6000 91.72% 92.03%
8000 91.86% 92.25%
10000 92.03% 92.52%
12000 92.14% 92.88%
14000 92.34% 92.81%
16000 92.29% 92.99%
18000 92.18% 93.13%
20000 92.06% 93.12%
1 loop, best of 1: 1min 8s per loop
| MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
最後得到的準確率是 92%-93%不算完美，不過畢竟這只有一個矩陣而已。 光看數據沒感覺，我們來看看前十筆測試資料跑起來的情形可以看到前十筆只有錯一個 | Pr = compute_Pr(test_X[:10])
pred_y =Pr.argmax(axis=1)
for i in range(10):
print(pred_y[i], test_y[i])
showX(test_X[i]) | 7 7
| MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
看看前一百筆資料中，是哪些情況算錯 | Pr = compute_Pr(test_X[:100])
pred_y = Pr.argmax(axis=1)
for i in range(100):
if pred_y[i] != test_y[i]:
print(pred_y[i], test_y[i])
showX(test_X[i]) | 6 5
| MIT | Week05/From NumPy to Logistic Regression.ipynb | HowardNTUST/HackNTU_Data_2017 |
Скородумов Александр БВТ1904 Лабораторная работа №2 Методы поиска №1 | #Импорты
from IPython.display import HTML, display
from tabulate import tabulate
import random
import time
#Рандомная генерация
def random_matrix(m = 50, n = 50, min_limit = -250, max_limit = 1016):
return [[random.randint(min_limit, max_limit) for _ in range(n)] for _ in range(m)]
#Бинарный поиск
class BinarySearc... | _____no_output_____ | MIT | .ipynb_checkpoints/Skorodumov.Lab2-checkpoint.ipynb | SkorodumovAlex/SIAODLabs |
№2 | #Простое рехеширование
class HashMap:
def __init__(self):
self.size = 0
self.data = []
self._resize()
def _hash(self, key, i):
return (hash(key) + i) % len(self.data)
def _find(self, key):
i = 0;
index = self._hash(key, i);
while self.dat... | _____no_output_____ | MIT | .ipynb_checkpoints/Skorodumov.Lab2-checkpoint.ipynb | SkorodumovAlex/SIAODLabs |
Сравнение алгоритмов | алгоритмы = {
'Бинарный поиск': BinarySearchMap,
'Фибоначчиева поиск': FibonacciMap,
'Интерполяционный поиск': InterpolateMap,
'Бинарное дерево': BinaryTreeMap,
'Простое рехэширование': HashMap,
'Рехэширование с помощью псевдослучайных чисел': RandomHashMap,
'Метод цепочек': ChainMap,
'С... | _____no_output_____ | MIT | .ipynb_checkpoints/Skorodumov.Lab2-checkpoint.ipynb | SkorodumovAlex/SIAODLabs |
№3 | #Вывод результата
def tag(x, color='white'):
return f'<td style="width:24px;height:24px;text-align:center;" bgcolor="{color}">{x}</td>'
th = ''.join(map(tag, ' abcdefgh '))
def chessboard(data):
row = lambda i: ''.join([
tag('<span style="font-size:24px">*</span>' * v,
color='white' if (i+j+... | _____no_output_____ | MIT | .ipynb_checkpoints/Skorodumov.Lab2-checkpoint.ipynb | SkorodumovAlex/SIAODLabs |
```Pythonx_train = HDF5Matrix("data.h5", "x_train")x_valid = HDF5Matrix("data.h5", "x_valid")```shapes should be:* (1355578, 432, 560, 1)* (420552, 432, 560, 1) | def gen_data(shape=0, name="input"):
data = np.random.rand(512, 512, 4)
label = data[:,:,-1]
return tf.constant(data.reshape(1,512,512,4).astype(np.float32)), tf.constant(label.reshape(1,512,512,1).astype(np.float32))
## NOTE:
## Tensor 4D -> Batch,X,Y,Z
## Tesnor max. float32!
d, l = gen_data(0,0)
p... | _____no_output_____ | BSD-3-Clause | Examples/simple_U-Net.ipynb | thgnaedi/DeepRain |
Anschließend:```Pythonoutput_length = 1input_length = output_length + 1input_shape=(432, 560, input_length)model_1 = unet(input_shape, output_length)model_1.fit(x_train_1, y_train_1, batch_size = 16, epochs = 25, validation_data=(x_valid_1, y_valid_1))``` | d.summary()
#ToDo: now learn something! | _____no_output_____ | BSD-3-Clause | Examples/simple_U-Net.ipynb | thgnaedi/DeepRain |
Implementing a Neural NetworkIn this exercise we will develop a neural network with fully-connected layers to perform classification, and test it out on the CIFAR-10 dataset. | # A bit of setup
import numpy as np
import matplotlib.pyplot as plt
from cs231n.classifiers.neural_net import TwoLayerNet
%matplotlib inline
plt.rcParams['figure.figsize'] = (10.0, 8.0) # set default size of plots
plt.rcParams['image.interpolation'] = 'nearest'
plt.rcParams['image.cmap'] = 'gray'
# for auto-reloadi... | /Users/ayush/anaconda2/lib/python2.7/site-packages/matplotlib/font_manager.py:273: UserWarning: Matplotlib is building the font cache using fc-list. This may take a moment.
warnings.warn('Matplotlib is building the font cache using fc-list. This may take a moment.')
| MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
We will use the class `TwoLayerNet` in the file `cs231n/classifiers/neural_net.py` to represent instances of our network. The network parameters are stored in the instance variable `self.params` where keys are string parameter names and values are numpy arrays. Below, we initialize toy data and a toy model that we will... | # Create a small net and some toy data to check your implementations.
# Note that we set the random seed for repeatable experiments.
input_size = 4
hidden_size = 10
num_classes = 3
num_inputs = 5
def init_toy_model():
np.random.seed(0)
return TwoLayerNet(input_size, hidden_size, num_classes, std=1e-1)
def init_t... | _____no_output_____ | MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
Forward pass: compute scoresOpen the file `cs231n/classifiers/neural_net.py` and look at the method `TwoLayerNet.loss`. This function is very similar to the loss functions you have written for the SVM and Softmax exercises: It takes the data and weights and computes the class scores, the loss, and the gradients on the... | scores = net.loss(X)
print 'Your scores:'
print scores
print
print 'correct scores:'
correct_scores = np.asarray([
[-0.81233741, -1.27654624, -0.70335995],
[-0.17129677, -1.18803311, -0.47310444],
[-0.51590475, -1.01354314, -0.8504215 ],
[-0.15419291, -0.48629638, -0.52901952],
[-0.00618733, -0.12435261, -0.1... | Your scores:
[[-0.81233741 -1.27654624 -0.70335995]
[-0.17129677 -1.18803311 -0.47310444]
[-0.51590475 -1.01354314 -0.8504215 ]
[-0.15419291 -0.48629638 -0.52901952]
[-0.00618733 -0.12435261 -0.15226949]]
correct scores:
[[-0.81233741 -1.27654624 -0.70335995]
[-0.17129677 -1.18803311 -0.47310444]
[-0.51590475 -1... | MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
Forward pass: compute lossIn the same function, implement the second part that computes the data and regularizaion loss. | loss, _ = net.loss(X, y, reg=0.1)
correct_loss = 1.30378789133
# should be very small, we get < 1e-12
print 'Difference between your loss and correct loss:'
print np.sum(np.abs(loss - correct_loss)) | Difference between your loss and correct loss:
1.79856129989e-13
| MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
Backward passImplement the rest of the function. This will compute the gradient of the loss with respect to the variables `W1`, `b1`, `W2`, and `b2`. Now that you (hopefully!) have a correctly implemented forward pass, you can debug your backward pass using a numeric gradient check: | from cs231n.gradient_check import eval_numerical_gradient
# Use numeric gradient checking to check your implementation of the backward pass.
# If your implementation is correct, the difference between the numeric and
# analytic gradients should be less than 1e-8 for each of W1, W2, b1, and b2.
loss, grads = net.loss(... | (4, 10)
W1 max relative error: 1.000000e+00
| MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
Train the networkTo train the network we will use stochastic gradient descent (SGD), similar to the SVM and Softmax classifiers. Look at the function `TwoLayerNet.train` and fill in the missing sections to implement the training procedure. This should be very similar to the training procedure you used for the SVM and ... | net = init_toy_model()
stats = net.train(X, y, X, y,
learning_rate=1e-1, reg=1e-5,
num_iters=100, verbose=False)
print 'Final training loss: ', stats['loss_history'][-1]
# plot the loss history
plt.plot(stats['loss_history'])
plt.xlabel('iteration')
plt.ylabel('training loss')
plt.title('Train... | _____no_output_____ | MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
Load the dataNow that you have implemented a two-layer network that passes gradient checks and works on toy data, it's time to load up our favorite CIFAR-10 data so we can use it to train a classifier on a real dataset. | from cs231n.data_utils import load_CIFAR10
def get_CIFAR10_data(num_training=49000, num_validation=1000, num_test=1000):
"""
Load the CIFAR-10 dataset from disk and perform preprocessing to prepare
it for the two-layer neural net classifier. These are the same steps as
we used for the SVM, but condense... | _____no_output_____ | MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
Train a networkTo train our network we will use SGD with momentum. In addition, we will adjust the learning rate with an exponential learning rate schedule as optimization proceeds; after each epoch, we will reduce the learning rate by multiplying it by a decay rate. | input_size = 32 * 32 * 3
hidden_size = 50
num_classes = 10
net = TwoLayerNet(input_size, hidden_size, num_classes)
# Train the network
stats = net.train(X_train, y_train, X_val, y_val,
num_iters=1000, batch_size=200,
learning_rate=1e-4, learning_rate_decay=0.95,
reg=0.5, verbose=Tru... | _____no_output_____ | MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
Debug the trainingWith the default parameters we provided above, you should get a validation accuracy of about 0.29 on the validation set. This isn't very good.One strategy for getting insight into what's wrong is to plot the loss function and the accuracies on the training and validation sets during optimization.Anot... | # Plot the loss function and train / validation accuracies
plt.subplot(2, 1, 1)
plt.plot(stats['loss_history'])
plt.title('Loss history')
plt.xlabel('Iteration')
plt.ylabel('Loss')
plt.subplot(2, 1, 2)
plt.plot(stats['train_acc_history'], label='train')
plt.plot(stats['val_acc_history'], label='val')
plt.title('Classi... | _____no_output_____ | MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
Tune your hyperparameters**What's wrong?**. Looking at the visualizations above, we see that the loss is decreasing more or less linearly, which seems to suggest that the learning rate may be too low. Moreover, there is no gap between the training and validation accuracy, suggesting that the model we used has low capa... | best_net = None # store the best model into this
#################################################################################
# TODO: Tune hyperparameters using the validation set. Store your best trained #
# model in best_net. #
# ... | _____no_output_____ | MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
Run on the test setWhen you are done experimenting, you should evaluate your final trained network on the test set; you should get above 48%.**We will give you extra bonus point for every 1% of accuracy above 52%.** | test_acc = (best_net.predict(X_test) == y_test).mean()
print 'Test accuracy: ', test_acc | _____no_output_____ | MIT | assignment1/two_layer_net.ipynb | ayush29feb/cs231n |
Social Media Analysis EDA | df = pd.read_csv('./meme_cleaning.csv')
df_sentiment = pd.read_csv('563_df_sentiments.csv')
df_sentiment = df_sentiment.drop(columns=['Unnamed: 0', 'Unnamed: 0.1', 'Unnamed: 0.1.1'])
df_sentiment.head()
#Extract all words that begin with # and turn the results into a dataframe
temp = df_sentiment['Tweet'].str.lower().... | _____no_output_____ | MIT | Final_Colab.ipynb | jared-garalde/sme_deploy_heroku |
Sentiment | g = sns.catplot(x = "No_of_Likes", y = "Normalized_count", hue = "vaderSentiment", data = like_df, kind = "bar")
g = sns.catplot(x = "No_of_Retweets", y = "Normalized_count", hue = "vaderSentiment", data = retweet_df, kind = "bar")
plt.pie(classify_df['vaderSentiment'], labels=classify_df['index']);
l = []
for i in ra... | _____no_output_____ | MIT | Final_Colab.ipynb | jared-garalde/sme_deploy_heroku |
Word Cloud | wordcloud = WordCloud(width = 800, height = 800,
background_color ='white',
min_font_size = 10).generate(str(l))
plt.figure(figsize = (8, 8), facecolor = None)
plt.imshow(wordcloud, interpolation='bilinear')
plt.axis("off")
plt.tight_layout(pad = 0)
plt.show() | _____no_output_____ | MIT | Final_Colab.ipynb | jared-garalde/sme_deploy_heroku |
Topic Modeling | cv = CountVectorizer(stop_words='english')
data_cv = cv.fit_transform(df.Tweet)
words = cv.get_feature_names()
data_dtm = pd.DataFrame(data_cv.toarray(), columns=cv.get_feature_names())
pickle.dump(cv, open("cv_stop.pkl", "wb"))
data_dtm_transpose = data_dtm.transpose()
sparse_counts = scipy.sparse.csr_matrix(data_... | _____no_output_____ | MIT | Final_Colab.ipynb | jared-garalde/sme_deploy_heroku |
1. We shall use the same dataset used in previous assignment - digits. Make a 80-20 train/test split.[Hint: Explore datasets module from scikit learn] 2. Using scikit learn perform a LDA on the dataset. Find out the number of components in the projected subspace.[Hint: Refer to discriminant analysis module of scikit l... | import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.datasets import load_digits
digits = load_digits()
digits.data
digits.data.shape
digits.target.shape
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(digits.d... | _____no_output_____ | MIT | Cls5-Dimentionality Reduction/DimensionalityReduction-CaseStudy2-solution.ipynb | tuhinssam/MLResources |
Project: Create a neural network class---Based on previous code examples, develop a neural network class that is able to classify any dataset provided. The class should create objects based on the desired network architecture:1. Number of inputs2. Number of hidden layers3. Number of neurons per layer4. Number of outpu... | import numpy as np
import matplotlib.pyplot as plt
# Needed for the mnist data
from keras.datasets import mnist
from keras.utils import to_categorical | _____no_output_____ | MIT | Neural Network Assignment.ipynb | DeepLearningVision-2019/a3-neural-network-class-munozgce |
Define the class | class NeuralNetwork:
def __init__(self, architecture, alpha):
'''
layers: List of integers which represents the architecture of the network.
alpha: Learning rate.
'''
# TODO: Initialize the list of weights matrices, then store
# the network... | _____no_output_____ | MIT | Neural Network Assignment.ipynb | DeepLearningVision-2019/a3-neural-network-class-munozgce |
Test datasets XOR | # input dataset
XOR_inputs = np.array([
[0,0],
[0,1],
[1,0],
[1,1]
])
# labels dataset
XOR_labels = np.array([[0,1,1,0]]).T
#TODO: Test the class with the XOR data
| _____no_output_____ | MIT | Neural Network Assignment.ipynb | DeepLearningVision-2019/a3-neural-network-class-munozgce |
Multiple classes | # Creates the data points for each class
class_1 = np.random.randn(700, 2) + np.array([0, -3])
class_2 = np.random.randn(700, 2) + np.array([3, 3])
class_3 = np.random.randn(700, 2) + np.array([-3, 3])
feature_set = np.vstack([class_1, class_2, class_3])
labels = np.array([0]*700 + [1]*700 + [2]*700)
one_hot_labels... | Error: 0.47806009636665425
Error: 0.007121601211763323
Error: 0.005938405234795827
Error: 0.005920131593441376
Error: 0.00585185558757003
Error: 0.0049490985751735606
Error: 0.004301948969147726
Error: 0.0038221899933325782
Error: 0.003507891190406313
Error: 0.003280260509683804
| MIT | Neural Network Assignment.ipynb | DeepLearningVision-2019/a3-neural-network-class-munozgce |
On the mnist data set---Train the network to classify hand drawn digits.For this data set, if the training step is taking too long, you can try to adjust the architecture of the network to have fewer layers, or you could try to train it with fewer input. The data has already been loaded and preprocesed so that it can ... | # Load the train and test data from the mnist data set
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()
# Plot a sample data point
plt.title("Label: " + str(train_labels[0]))
plt.imshow(train_images[0], cmap="gray")
# Standardize the data
# Flatten the images
train_images = train_images.re... | Error: 0.47806009636665425
Error: 0.007121601211763323
Error: 0.005938405234795827
Error: 0.005920131593441376
Error: 0.00585185558757003
Error: 0.0049490985751735606
Error: 0.004301948969147726
Error: 0.0038221899933325782
Error: 0.003507891190406313
Error: 0.003280260509683804
| MIT | Neural Network Assignment.ipynb | DeepLearningVision-2019/a3-neural-network-class-munozgce |
Titanic 4 > `Pclass, Sex, Age` | import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
plt.style.use('seaborn')
sns.set(font_scale=2.5)
import missingno as msno
import warnings
warnings.filterwarnings('ignore')
%matplotlib inline
df_train=pd.read_csv('C:/Users/ehfus/Downloads/titanic/train.csv')
df_test=pd.rea... | _____no_output_____ | Apache-2.0 | _notebooks/2021-12-26-titanic.ipynb | rhkrehtjd/kaggle |
- scale에도 option이 여러가지 있음, google에서 확인해볼 것 > `Embarked : 탑승한 항구` | f,ax=plt.subplots(1,1,figsize=(7,7))
df_train[['Embarked','Survived']]\
.groupby(['Embarked'], as_index=True).mean()\
.sort_values(by='Survived', ascending=False)\
.plot.bar(ax=ax) | _____no_output_____ | Apache-2.0 | _notebooks/2021-12-26-titanic.ipynb | rhkrehtjd/kaggle |
- `sort_values` 또는 `sort_index`도 사용 가능 | f,ax=plt.subplots(2,2,figsize=(20,15)) #2차원임/ 1,2는 1차원
sns.countplot('Embarked',data=df_train, ax=ax[0,0])
ax[0,0].set_title('(1) No. Of Passengers Boared')
sns.countplot('Embarked',hue='Sex',data=df_train, ax=ax[0,1])
ax[0,1].set_title('(2) Male-Female split for embarked')
sns.countplot('Embarked', hue='Survived', d... | _____no_output_____ | Apache-2.0 | _notebooks/2021-12-26-titanic.ipynb | rhkrehtjd/kaggle |
> `Family - SibSp + ParCh` | df_train['FamilySize']=df_train['SibSp'] + df_train['Parch'] + 1
print('Maximum size of Family : ',df_train['FamilySize'].max())
print('Minimum size of Family : ',df_train['FamilySize'].min()) | Maximum size of Family : 11
Minimum size of Family : 1
| Apache-2.0 | _notebooks/2021-12-26-titanic.ipynb | rhkrehtjd/kaggle |
- Pandas series는 연산이 가능 | f,ax=plt.subplots(1,3,figsize=(40,10))
sns.countplot('FamilySize', data=df_train, ax=ax[0])
ax[0].set_title('(1) No. Of Passenger Boarded', y=1.02)
sns.countplot('FamilySize', hue='Survived',data=df_train, ax=ax[1])
ax[1].set_title('(2) Survived countplot depending on FamilySize', y=1.02)
df_train[['FamilySize','Surv... | _____no_output_____ | Apache-2.0 | _notebooks/2021-12-26-titanic.ipynb | rhkrehtjd/kaggle |
> `Fare : 요금, 연속형 변수` - distplot ?? 시리즈에 히스토그램을 그려줌,Skewness? 왜도임 + 첨도도 있음- 왜도? 첨도?- python에서 나타내는 함수는? | fig,ax=plt.subplots(1,1,figsize=(8,8))
g=sns.distplot(df_train['Fare'], color='b',label='Skewness{:.2f}'.format(df_train['Fare'].skew()),ax=ax)
g=g.legend(loc='best') | _____no_output_____ | Apache-2.0 | _notebooks/2021-12-26-titanic.ipynb | rhkrehtjd/kaggle |
- skewness가 5정도로 꽤 큼 -> 좌로 많이 치우쳐져 있음 -> 그대로 모델에 학습시키면 성능이 낮아질 수 있음 | df_train['Fare']=df_train['Fare'].map(lambda i: np.log(i) if i>0 else 0) | _____no_output_____ | Apache-2.0 | _notebooks/2021-12-26-titanic.ipynb | rhkrehtjd/kaggle |
df_train['Fare']의 값을 적절하게 변형 중 | fig,ax=plt.subplots(1,1,figsize=(8,8))
g=sns.distplot(df_train['Fare'], color='b',label='Skewness{:.2f}'.format(df_train['Fare'].skew()),ax=ax)
g=g.legend(loc='best') | _____no_output_____ | Apache-2.0 | _notebooks/2021-12-26-titanic.ipynb | rhkrehtjd/kaggle |
이런 작업(log로 변환)을 통해 skewness가 0으로 근접하게 해주었음 | df_train['Ticket'].value_counts() | _____no_output_____ | Apache-2.0 | _notebooks/2021-12-26-titanic.ipynb | rhkrehtjd/kaggle |
Section 1.2: Dimension reduction and principal component analysis (PCA)One of the iron laws of data science is know as the "curse of dimensionality": as the number of considered features (dimensions) of a feature space increases, the number of data configurations can grow exponentially and thus the number observations... | import pandas as pd
import numpy as np
from sklearn.decomposition import PCA
from sklearn.preprocessing import StandardScaler
import matplotlib.pyplot as plt
%matplotlib inline | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
The dataset we’ll use here is the same one drawn from the [U.S. Department of Agriculture National Nutrient Database for Standard Reference](https://www.ars.usda.gov/northeast-area/beltsville-md-bhnrc/beltsville-human-nutrition-research-center/nutrient-data-laboratory/docs/usda-national-nutrient-database-for-standard-r... | df = pd.read_csv('Data/USDA-nndb-combined.csv', encoding='latin_1') | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
We can check the number of columns and rows using the `info()` method for the `DataFrame`. | df.info() | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
> **Exercise**>> Can you think of a more concise way to check the number of rows and columns in a `DataFrame`? (***Hint:*** Use one of the [attributes](https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.html) of the `DataFrame`.) Handle `null` valuesBecause this is a real-world dataset, it is ... | df.shape | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
Dropping those rows eliminated 76 percent of our data (8989 entries to 2190). An imperfect state of affairs, but we still have enough for our purposes in this section.> **Key takeaway:** Another solution to removing `null` values is to impute values for them, but this can be tricky. Should we handle missing values as e... | desc_df = df.iloc[:, [0, 1, 2]+[i for i in range(50,54)]]
desc_df.set_index('NDB_No', inplace=True)
desc_df.head() | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
> **Question**>> Why was it necessary to structure the `iloc` method call the way we did in the code cell above? What did it accomplish? Why was it necessary set the `desc_df` index to `NDB_No`? | nutr_df = df.iloc[:, :-5]
nutr_df.head() | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
> **Question**>> What did the `iloc` syntax do in the code cell above? | nutr_df = nutr_df.drop(['FoodGroup', 'Shrt_Desc'], axis=1) | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
> **Exercise**>> Now set the index of `nutr_df` to use `NDB_No`. > **Exercise solution**>> The correct code for students to use here is `nutr_df.set_index('NDB_No', inplace=True)`. Now let’s take a look at `nutr_df`. | nutr_df.head() | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
Check for correlation among featuresOne thing that can skew our classification results is correlation among our features. Recall that the whole reason that PCA works is that it exploits the correlation among data points to project our feature-space into a lower-dimensional space. However, if some of our features are h... | nutr_df.drop(['Folate_DFE_(µg)', 'Vit_A_RAE', 'Vit_D_IU'],
inplace=True, axis=1)
nutr_df.head() | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
Normalize and center the dataOur numeric data comes in a variety of mass units (grams, milligrams, and micrograms) and one energy unit (kilocalories). In order to make an apples-to-apples comparison (pun intended) of the nutritional data, we need to first *normalize* the data and make it more normally distributed (tha... | ax = nutr_df.hist(bins=50, xlabelsize=-1, ylabelsize=-1, figsize=(11,11)) | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
Not a bell curve in sight. Worse, a lot of the data is clumped at or around 0. We will use the Box-Cox Transformation on the data, but it requires strictly positive input, so we will add 1 to every value in each column. | nutr_df = nutr_df + 1 | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
Now for the transformation. The [Box-Cox Transformation](https://www.statisticshowto.datasciencecentral.com/box-cox-transformation/) performs the transformation $y(\lambda) = \dfrac{y^{\lambda}-1}{\lambda}$ for $\lambda \neq 0$ and $y(\lambda) = log y$ for $\lambda = 0$ for all values $y$ in a given column. SciPy has a... | from scipy.stats import boxcox
nutr_df_TF = pd.DataFrame(index=nutr_df.index)
for col in nutr_df.columns.values:
nutr_df_TF['{}_TF'.format(col)] = boxcox(nutr_df.loc[:, col])[0] | _____no_output_____ | MIT | Machine Learning 2_Using Advanced Machine Learning Models/Reference Material/190053-Reactors-DS-Tr2-Sec1-2-PCA.ipynb | raspyweather/Reactors |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.