markdown stringlengths 0 37k | code stringlengths 1 33.3k | path stringlengths 8 215 | repo_name stringlengths 6 77 | license stringclasses 15
values |
|---|---|---|---|---|
¿Por qué usar Numpy?
%%capture captura la salida de la ejecución de la celda en la variable dada como parámetro. Después se puede imprimir.
%timeit se utiliza para ejecutar varias veces una instrucción y calcular un promedio de su duración. | %%capture timeit_output
%timeit l1 = range(1,1000)
%timeit l2 = np.arange(1,1000)
print(timeit_output)
x = np.array([[1,2],[3,4]])
print (np.sum(x)) # Compute sum of all elements; prints "10"
print (np.sum(x, axis=0)) # Compute sum of each column; prints "[4 6]"
print (np.sum(x, axis=1)) # Compute sum of each r... | intro/sesion0.ipynb | dsevilla/bdge | mit |
numpy tiene infinidad de funciones, por lo que sería interesante darse una vuelta por su documentación: https://docs.scipy.org/doc/.
Matplotlib
Matplotlib permite generar gráficos de forma sencilla. Lo veremos aquí primero conectado sólo con Numpy y después conectado con Pandas. | x = np.arange(0, 3 * np.pi, 0.1)
y = np.sin(x)
plt.subplot()
# Plot the points using matplotlib
plt.plot(x, y)
plt.show()
plt.subplot(211)
plt.plot(range(12))
plt.subplot(212, facecolor='y')
plt.plot(range(100))
plt.show()
# Compute the x and y coordinates for points on sine and cosine curves
x = np.arange(0, 3 * np.... | intro/sesion0.ipynb | dsevilla/bdge | mit |
Pandas
Tutoriales: 1, 2, 3.
Pandas permite gestionar conjuntos de datos n-dimensionales de diferentes formas, y también conectarlo con matplotlib para hacer gráficas.
Los conceptos principales de Pandas son los Dataframes y las Series. La diferencia entre ambas es que la serie guarda sólo una serie (una columna o una f... | ts = pd.Series(np.random.randn(1000), index=pd.date_range('1/1/2000', periods=1000))
ts
ts.describe()
ts = ts.cumsum()
ts.plot();
df = pd.DataFrame(np.random.randn(1000, 4), index=ts.index, columns=list('ABCD'))
df = df.cumsum()
df.plot(); | intro/sesion0.ipynb | dsevilla/bdge | mit |
Se puede hacer plot también de una columna contra otra. | df3 = pd.DataFrame(np.random.randn(1000, 2), columns=['B', 'C']).cumsum()
df3['A'] = pd.Series(list(range(len(df3))))
df3.plot(x='A', y='B'); | intro/sesion0.ipynb | dsevilla/bdge | mit |
Valores incompletos. Si no se establecen, se pone a NaN (not a number). | d = {'one' : pd.Series([1., 2., 3.], index=['a', 'b', 'c']),
'two' : pd.Series([1., 2., 3., 4.], index=['a', 'b', 'c', 'd'])}
df = pd.DataFrame(d)
df | intro/sesion0.ipynb | dsevilla/bdge | mit |
fillna() permite cambiar el valor de los datos faltantes. | df.fillna(0)
pd.DataFrame(d, index=['d', 'b', 'a'])
pd.DataFrame(d, index=['d', 'b', 'a'], columns=['two', 'three']) | intro/sesion0.ipynb | dsevilla/bdge | mit |
A continuación se muestra un ejemplo de uso de Pandas para leer datos y procesarlos en un Dataframe.
El primer ejemplo completo carga desde el fichero swift-question-dates.txt.gz las fechas de las preguntas en Stackoverflow que contienen el tag "swift".
La función read_csv es capaz de leer cualquier fichero CSV y lo co... | df = pd.read_csv('https://github.com/dsevilla/bdge/raw/master/intro/swift-question-dates.txt.gz',
header=None,
names=['date'],
compression='gzip',
parse_dates=['date'],
index_col='date')
df | intro/sesion0.ipynb | dsevilla/bdge | mit |
De la fecha, extraer sólo la fecha (no la hora, que no nos interesa). | df.index = df.index.date | intro/sesion0.ipynb | dsevilla/bdge | mit |
Añadimos una columna de todo "1" para especificar que cada pregunta cuenta como 1. | df['Count'] = 1
df | intro/sesion0.ipynb | dsevilla/bdge | mit |
A los Dataframe de Pandas también se les puede aplicar operaciones de agregación, como groupby o sum. Finalmente, la funcion plot() permite mostrar los datos en un gráfico. | accum = df.groupby(df.index).sum()
accum
# Los 30 primeros registros que tengan un número de preguntas mayor que 20 por día.
accum = accum[accum.Count > 20][:30]
accum
accum[accum.Count > 30][:30].plot.bar() | intro/sesion0.ipynb | dsevilla/bdge | mit |
A continuación comprobamos con la página de la Wikipedia cuándo apareció el lenguaje Swift: | !pip install lxml
dfwiki = pd.read_html('https://en.wikipedia.org/wiki/Swift_(programming_language)',attrs={'class': 'infobox vevent'})
dfwiki[0]
firstdate = dfwiki[0][1][4]
firstdate
from dateutil.parser import parse
dt = parse(firstdate.split(';')[0])
print (dt.date().isoformat())
print (accum.index[0].isoformat(... | intro/sesion0.ipynb | dsevilla/bdge | mit |
A continuación se muestra cómo ubicar posiciones en un mapa con el paquete folium. Se muestra también cómo acceder a distintas posiciones del Dataframe con iloc, loc, etc. | # cargar municipios y mostrarlos en el mapa
df = pd.read_csv('https://github.com/dsevilla/bdge/raw/master/intro/municipios-2017.csv.gz',header=0,compression='gzip')
df.head()
df.iloc[0]
df.iloc[0].NOMBRE_ACTUAL
df.loc[:,'NOMBRE_ACTUAL']
df.iloc[:,0]
df.PROVINCIA
df[df.PROVINCIA == 'A Coruña']
mula = df[df.NOMBR... | intro/sesion0.ipynb | dsevilla/bdge | mit |
El paquete folium permite generar mapas de posiciones. El siguiente ejemplo centra un mapa en Mula y pone un marcador con su nombre: | !pip install folium
import folium
map = folium.Map(location=[mula_lat, mula_lon],zoom_start=10)
folium.Marker(location = [mula_lat, mula_lon], popup="{} ({} habitantes)".format(mula.NOMBRE_ACTUAL,mula.POBLACION_MUNI)).add_to(map)
map | intro/sesion0.ipynb | dsevilla/bdge | mit |
Test your code to make sure that the class definition worked. | # code to make sure constructors and get methods all work | lesson15/Lesson15_individual.ipynb | nerdcommander/scientific_computing_2017 | mit |
To run this example, we need to first get the data from [1] and process it so we have dichtonomous $y\in{-1,1}$ outputs and the matrix of predictors has been standardised. In addition, we also add a column of 1s corresponding to a constant term in the regression.
If you are connected to the internet, by instantiating w... | logpdf = pints.toy.GermanCreditLogPDF(download=True) | examples/toy/distribution-german-credit.ipynb | martinjrobins/hobo | bsd-3-clause |
Let's look at the data: x is a matrix of predictors and y is a vector of credit recommendations for 1000 individuals.
Specifically, let's look at the PCA scores and plot the first two dimensions against one another. Here, we see that the two groups overlap substantially, but that there neverless some separation along ... | def pca(X):
# Data matrix X, assumes 0-centered
n, m = X.shape
# Compute covariance matrix
C = np.dot(X.T, X) / (n-1)
# Eigen decomposition
eigen_vals, eigen_vecs = np.linalg.eig(C)
# Project X onto PC space
X_pca = np.dot(X, eigen_vecs)
return X_pca
x, y = logpdf.data()
scores = pca(x)
# colour ind... | examples/toy/distribution-german-credit.ipynb | martinjrobins/hobo | bsd-3-clause |
Now we run HMC to fit the parameters of the model. | xs = [
np.random.uniform(0, 1, size=(logpdf.n_parameters())),
np.random.uniform(0, 1, size=(logpdf.n_parameters())),
np.random.uniform(0, 1, size=(logpdf.n_parameters())),
]
mcmc = pints.MCMCController(logpdf, len(xs), xs, method=pints.HamiltonianMCMC)
mcmc.set_max_iterations(200)
# Set up modest logging
... | examples/toy/distribution-german-credit.ipynb | martinjrobins/hobo | bsd-3-clause |
HMC is quite efficient here at sampling from the posterior distribution. | results = pints.MCMCSummary(chains=chains, time=mcmc.time())
print(results) | examples/toy/distribution-german-credit.ipynb | martinjrobins/hobo | bsd-3-clause |
*2. Set all graphics from matplotlib to display inline | import matplotlib.pyplot as plt
%matplotlib inline | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*3. Import pandas with the right name | #for encoding the command would look smth like this:
#df = pd.read_csv("XXXXXXXXXXXXXXXXX.csv", encoding='mac_roman')
df = pd.read_csv("Animal_Data/07-hw-animals.csv") | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*4. Display the names of the columns in the csv | df.columns | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*5. Display the first 3 animals. | df.head(3) | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*6. Sort the animals to see the 3 longest animals. | df.sort_values(by='length', ascending=False).head(3) | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*7. What are the counts of the different values of the "animal" column? a.k.a. how many cats and how many dogs. | df['animal'].value_counts() | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*8. Only select the dogs. | #df['animal'] == 'dog' this just tests, whether row is a dog or not, True or False
#is_dog = df['animal'] == 'dog'
#df[is_dog]
df[df['animal'] == 'dog'] | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*9. Display all of the animals that are greater than 40 cm. | df[df['length'] > 40]
#del df['feet'] | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*10. 'length' is the animal's length in cm. Create a new column called inches that is the length in inches. | df['inches'] = df['length'] * 0.394
df.head() | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*11. Save the cats to a separate variable called "cats." Save the dogs to a separate variable called "dogs." | dogs = df[df['animal'] == 'dog']
cats = df[df['animal'] == 'cat'] | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*12. Display all of the animals that are cats and above 12 inches long. First do it using the "cats" variable, then do it using your normal dataframe. | cats[cats['inches'] > 12]
#df[df[df[df['animal'] == 'cat']'inches'] > 12]
#df[df['animal'] == 'cat']&
#df[df['inches'] > 12]
#pd.read_csv('imdb.txt')
# .sort(columns='year')
# .filter('year >1990')
# .to_csv('filtered.csv')
df[(df['animal'] == 'cat') & (df['inches'] > 12)]
#3 > 2 & 4 > 3
#true & true
#true
#3 ... | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*13. What's the mean length of a cat? | df[df['animal'] == 'cat'].describe() | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*14. What's the mean length of a dog? | df[df['animal'] == 'dog'].describe() | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*15. Use groupby to accomplish both of the above tasks at once. | df.groupby(['animal'])['inches'].describe() | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*16. Make a histogram of the length of dogs. I apologize that it is so boring. | df[df['animal'] == 'dog'].hist() | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*17. Change your graphing style to be something else (anything else!) | import matplotlib.pyplot as plt
plt.style.available
plt.style.use('ggplot')
dogs['inches'].hist() | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*18. Make a horizontal bar graph of the length of the animals, with their name as the label (look at the billionaires notebook I put on Slack!) | df['length'].plot(kind='bar')
#or:
df.plot(kind='barh', x='name', y='length', legend=False) | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
*19. Make a sorted horizontal bar graph of the cats, with the larger cats on top. | cats_sorted = cats.sort_values(by='length', ascending=True).head(3)
cats_sorted.plot(kind='barh', x='name', y='length', legend=False)
#or:
df[df['animal'] == 'cat'].sort_values(by='length', ascending=True).plot(kind='barh', x='name', y='length') | 07/Animal_Panda_Homework_7_Skinner.ipynb | barjacks/foundations-homework | mit |
Advanced Sounding
Plot a sounding using MetPy with more advanced features.
Beyond just plotting data, this uses calculations from metpy.calc to find the lifted
condensation level (LCL) and the profile of a surface-based parcel. The area between the
ambient profile and the parcel profile is colored as well. | from datetime import datetime
import matplotlib.pyplot as plt
import metpy.calc as mpcalc
from metpy.io import get_upper_air_data
from metpy.io.upperair import UseSampleData
from metpy.plots import SkewT
from metpy.units import concatenate
with UseSampleData(): # Only needed to use our local sample data
# Downl... | v0.4/_downloads/Advanced_Sounding.ipynb | metpy/MetPy | bsd-3-clause |
Create a new figure. The dimensions here give a good aspect ratio | fig = plt.figure(figsize=(9, 9))
skew = SkewT(fig, rotation=45)
# Plot the data using normal plotting functions, in this case using
# log scaling in Y, as dictated by the typical meteorological plot
skew.plot(p, T, 'r')
skew.plot(p, Td, 'g')
skew.plot_barbs(p, u, v)
skew.ax.set_ylim(1000, 100)
skew.ax.set_xlim(-40, 60... | v0.4/_downloads/Advanced_Sounding.ipynb | metpy/MetPy | bsd-3-clause |
To print a value to the screen, we use the function print()
e.g. print(1) | print("Hello World") | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
Jupyter notebooks will always print the value of the last line so you don't have to. You can suppress this with a semicolon ';' | "Hello World"
"Hello World"; | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
TRY IT
Predict and then print the type of 'Orca'
Variables
A variable is a name that you give a value. You can then use this name anywhere you would use the value that the name refers to.
It has some rules.
* It must only contain letters, numbers and/or the underscore character.
* However, it cannot start with a numb... | WHALE = 'Orca'
number_of_whales = 10
weight_of_1_whale = 5003.2 | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
Notice that when you ran that, nothing printed out. To print a variable, you use the same statement you would use to print the value. e.g. print(WHALE) | print(number_of_whales) | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
TRY IT
Assign the name of a sea creature to the variable sea_creature. Then print the value.
Reccomendation
Name your variables with descriptive names. Naming a variable 'a' is easy to type but won't help you figure out what it is doing when you come back to your code six months later.
Operators and operands
Operators... | 1 + 2
fish = 15
fish_left = fish - 3
print(fish_left)
print(3 * 2.1)
number_of_whales ** 2
print(5 / 2) | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
Hint: You can use a variable and assign it to the same variable name in the same statement. | number_of_whales = 8
number_of_whales = number_of_whales + 2
print(number_of_whales) | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
TRY IT
Find the result of 6^18.
Order of operations
You can combine many operators in a single python statement. The way python evaluates it is the same way you were taught to in elementary school. PEMDAS for Please Excuse My Dear Aunt Sally. Or 1. Parentheses, 2. Exponents, 3. Multiplication, 4. Division, 5. Addition,... | 2 * 3 + 4 / 2
(2 * (3 + 4)) / 2 | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
Modulus operator
The modulus operator is not one you were taught in school. It returns the remainder of integer division. It is useful in a few specific cases, but you could go months without using it. | 5 % 2 | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
TRY IT
Find if 12342872 is divisible by 3
String operations
The + operator also works on strings. It is the concatenation operator, meaning it joins the two strings together. | print('Hello ' + 'Coding Circle')
print("The " + WHALE + " lives in the sea.") | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
Hint: Be careful with spaces | print("My name is" + "Charlotte") | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
TRY IT
Print out Good morning to the sea creature you stored in variable named sea_creature earlier.
Asking the user for input
To get an input for the user we use the built-in function input() and assign it to a variable.
NOTE: The result is always a string.
WARNING if you leave an input box without ever putting input ... | my_name = input()
print(my_name) | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
You can pass a string to the input() function to prompt the user for what you are looking for.
e.g. input('How are you feeling?')
Hint, add a new line character "\n" to the end of the prompt to make the user enter it on a new line. | favorite_ocean_animal = input("What is your favorite sea creature?\n")
print("The " + favorite_ocean_animal + " is so cool!") | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
If you want the user to enter a number, you will have to convert the string. Here are the conversion commands.
To convert a variable to an integer, use int -- e.g., int(variable_name)
To convert a variable to a float, use float -- e.g., float(variable_name)
To convert a variable to a string, use str -- e.g., str(varia... | number_of_fish = input("How many fish do you want?\n")
number_of_fish_int = int(number_of_fish)
print(number_of_fish_int * 1.05) | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
TRY IT
Prompt the user for their favorite whale and store the value in a variable called favorite_whale.
Comments
Comments let you explain your progam to someone who is reading your code. Do you know who that person is? It is almost always you in six months. Don't screw over future you. Comment your code.
To make a com... | # Calculate the price of fish that a user wants
number_of_fish = input("How many fish do you want?\n") # Ask user for quantity of fish
number_of_fish_int = int(number_of_fish) # raw_input returns string, so convert to integer
print(number_of_fish_int * 1.05) # multiply by price of fish | Lesson01_Variables/lesson01.ipynb | WomensCodingCircle/CodingCirclePython | mit |
Invoking Modules
Let's instantiate a Dense layer.
- Modules are actually objects in this API, so we provide contructor arguments when initializing the Module. In this case, we only have to provide the output features dimension. | model = nn.Dense(features=3) | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
We need to initialize the Module variables, these include the parameters of the Module as well as any other state variables.
We call the init method on the instantiated Module. If the Module __call__ method has args (self, *args, **kwargs) then we call init with (rngs, *args, **kwargs) so in this case, just (rng, inpu... | # Make RNG Keys and a fake input.
key1, key2 = random.split(random.PRNGKey(0), 2)
x = random.uniform(key1, (4,4))
# provide key and fake input to get initialized variables
init_variables = model.init(key2, x)
init_variables | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
We call the apply method on the instantiated Module. If the Module __call__ method has args (self, *args, **kwargs) then we call apply with (variables, *args, rngs=<RNGS>, mutable=<MUTABLEKINDS>, **kwargs) where
- <RNGS> are the optional call time RNGs for things like dropout. For simple Modules th... | y = model.apply(init_variables, x)
y | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
Additional points:
- If you want to init or apply a Module using a method other than call, you need to provide the method= kwarg to init and apply to use it instead of the default __call__, e.g. method='encode', method='decode' to apply the encode/decode methods of an autoencoder.
Defining Basic Modules
Composing subm... | class ExplicitMLP(nn.Module):
features: Sequence[int]
def setup(self):
# we automatically know what to do with lists, dicts of submodules
self.layers = [nn.Dense(feat) for feat in self.features]
# for single submodules, we would just write:
# self.layer1 = nn.Dense(feat1)
def __call__(self, inpu... | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
Here we show the equivalent compact form of the MLP that declares the submodules inline using the @compact decorator. | class SimpleMLP(nn.Module):
features: Sequence[int]
@nn.compact
def __call__(self, inputs):
x = inputs
for i, feat in enumerate(self.features):
x = nn.Dense(feat, name=f'layers_{i}')(x)
if i != len(self.features) - 1:
x = nn.relu(x)
# providing a name is optional though!
#... | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
Declaring and using variables
Flax uses lazy initialization, which allows declared variables to be initialized only at the first site of their use, using whatever shape information is available a the local call site for shape inference. Once a variable has been initialized, a reference to the data is kept for use in s... | class SimpleDense(nn.Module):
features: int
kernel_init: Callable = nn.initializers.lecun_normal()
bias_init: Callable = nn.initializers.zeros
@nn.compact
def __call__(self, inputs):
kernel = self.param('kernel',
self.kernel_init, # RNG passed implicitly.
... | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
We can also declare variables in setup, though in doing so you can't take advantage of shape inference and have to provide explicit shape information at initialization. The syntax is a little repetitive in this case right now, but we do force agreement of the assigned names. | class ExplicitDense(nn.Module):
features_in: int # <-- explicit input shape
features: int
kernel_init: Callable = nn.initializers.lecun_normal()
bias_init: Callable = nn.initializers.zeros
def setup(self):
self.kernel = self.param('kernel',
self.kernel_init,
... | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
General Variables
For declaring generally mutable variables that may be mutated inside the model we use the call:
self.variable(variable_kind, variable_name, variable_init_fn, *init_args)
with arguments:
- variable_kind the "kind" of state this variable is, i.e. the name of the nested-dict collection that this will be... | class Counter(nn.Module):
@nn.compact
def __call__(self):
# easy pattern to detect if we're initializing
is_initialized = self.has_variable('counter', 'count')
counter = self.variable('counter', 'count', lambda: jnp.zeros((), jnp.int32))
if is_initialized:
counter.value += 1
return counter... | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
Another Mutability and RNGs Example
Let's make an artificial, goofy example that mixes differentiable parameters, stochastic layers, and mutable variables: | class Block(nn.Module):
features: int
training: bool
@nn.compact
def __call__(self, inputs):
x = nn.Dense(self.features)(inputs)
x = nn.Dropout(rate=0.5)(x, deterministic=not self.training)
x = nn.BatchNorm(use_running_average=not self.training)(x)
return x
key1, key2, key3, key4 = random.split... | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
JAX transformations inside modules
JIT
It's not immediately clear what use this has, but you can compile specific submodules if there's a reason to.
Known Gotcha: at the moment, the decorator changes the RNG stream slightly, so comparing jitted an unjitted initializations will look different. | class MLP(nn.Module):
features: Sequence[int]
@nn.compact
def __call__(self, inputs):
x = inputs
for i, feat in enumerate(self.features):
# JIT the Module (it's __call__ fn by default.)
x = nn.jit(nn.Dense)(feat, name=f'layers_{i}')(x)
if i != len(self.features) - 1:
x = nn.relu... | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
Remat
For memory-expensive computations, we can remat our method to recompute a Module's output during a backwards pass.
Known Gotcha: at the moment, the decorator changes the RNG stream slightly, so comparing remat'd and undecorated initializations will look different. | class RematMLP(nn.Module):
features: Sequence[int]
# For all transforms, we can annotate a method, or wrap an existing
# Module class. Here we annotate the method.
@nn.remat
@nn.compact
def __call__(self, inputs):
x = inputs
for i, feat in enumerate(self.features):
x = nn.Dense(feat, name=f'l... | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
Vmap
You can now vmap Modules inside. The transform has a lot of arguments, they have the usual jax vmap args:
- in_axes - an integer or None for each input argument
- out_axes - an integer or None for each output argument
- axis_size - the axis size if you need to give it explicitly
In addition, we provide for eac... | class RawDotProductAttention(nn.Module):
attn_dropout_rate: float = 0.1
train: bool = False
@nn.compact
def __call__(self, query, key, value, bias=None, dtype=jnp.float32):
assert key.ndim == query.ndim
assert key.ndim == value.ndim
n = query.ndim
attn_weights = lax.dot_general(
query,... | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
Scan
Scan allows us to apply lax.scan to Modules, including their parameters and mutable variables. To use it we have to specify how we want each "kind" of variable to be transformed. For scanned variables we specify similar to vmap via in variable_in_axes, variable_out_axes:
- nn.broadcast broadcast the variable ki... | class SimpleScan(nn.Module):
@nn.compact
def __call__(self, xs):
dummy_rng = random.PRNGKey(0)
init_carry = nn.LSTMCell.initialize_carry(dummy_rng,
xs.shape[:1],
xs.shape[-1])
LSTM = nn.scan(nn.LSTMCell,
... | docs/notebooks/linen_intro.ipynb | google/flax | apache-2.0 |
One of the simple things we can do is override the default CSS to customize our DataFrame output.
This specific example is from - Brandon Rhodes' talk at pycon
For the purposes of the notebook, I'm defining CSS as a variable but you could easily read in from a file as well. | CSS = """
body {
margin: 0;
font-family: Helvetica;
}
table.dataframe {
border-collapse: collapse;
border: none;
}
table.dataframe tr {
border: none;
}
table.dataframe td, table.dataframe th {
margin: 0;
border: 1px solid white;
padding-left: 0.25em;
padding-right: 0.25em;
}
table.da... | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
Now add this CSS into the current notebook's HTML. | from IPython.core.display import HTML
HTML('<style>{}</style>'.format(CSS))
SALES=pd.read_csv("../data/sample-sales-tax.csv", parse_dates=True)
SALES.head() | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
You can see how the CSS is now applied to the DataFrame and how you could easily modify it to customize it to your liking.
Jupyter notebooks do a good job of automatically displaying information but sometimes you want to force data to display. Fortunately, ipython provides and option. This is especially useful if you w... | from IPython.display import display
display(SALES.head(2))
display(SALES.tail(2))
display(SALES.describe()) | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
Using pandas settings to control output
Pandas has many different options to control how data is displayed.
You can use max_rows to control how many rows are displayed | pd.set_option("display.max_rows",4)
SALES | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
Depending on the data set, you may only want to display a smaller number of columns. | pd.set_option("display.max_columns",6)
SALES | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
You can control how many decimal points of precision to display | pd.set_option('precision',2)
SALES
pd.set_option('precision',7)
SALES | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
You can also format floating point numbers using float_format | pd.set_option('float_format', '{:.2f}'.format)
SALES | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
This does apply to all the data. In our example, applying dollar signs to everything would not be correct for this example. | pd.set_option('float_format', '${:.2f}'.format)
SALES | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
Third Party Plugins
Qtopian has a useful plugin called qgrid - https://github.com/quantopian/qgrid
Import it and install it. | import qgrid
qgrid.nbinstall(overwrite=True) | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
Showing the data is straighforward. | qgrid.show_grid(SALES, remote_js=True) | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
The plugin is very similar to the capability of an Excel autofilter. It can be handy to quickly filter and sort your data.
Improving your plots
I have mentioned before how the default pandas plots don't look so great. Fortunately, there are style sheets in matplotlib which go a long way towards improving the visualizat... | SALES.groupby('name')['quantity'].sum().plot(kind="bar") | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
We can use some of the matplolib styles available to us to make this look better.
http://matplotlib.org/users/style_sheets.html | plt.style.use('ggplot')
SALES.groupby('name')['quantity'].sum().plot(kind="bar") | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
You can see all the styles available | plt.style.available
plt.style.use('bmh')
SALES.groupby('name')['quantity'].sum().plot(kind="bar")
plt.style.use('fivethirtyeight')
SALES.groupby('name')['quantity'].sum().plot(kind="bar") | notebooks/Ipython-pandas-tips-and-tricks.ipynb | chris1610/pbpython | bsd-3-clause |
Basic principles
Automatic levels
If data is set, the levels are chosen according to the min and max of the data: | dl = DataLevels(a, cmap=cm)
dl.visualize(orientation='horizontal', add_values=True) | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
The Default number of levels is 8, but it's up to you to choose something else: | dl = DataLevels(a, cmap=cm, nlevels=256)
dl.visualize(orientation='horizontal', add_values=True) | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
Automatic levels with min and max value
You can also specify the boundaries of the levels to choose: | dl = DataLevels(a, cmap=cm, nlevels=256, vmax=3)
dl.visualize(orientation='horizontal', add_values=True) | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
You see that the colorbar has been extended. This behavior is forced by DataLevels and prevents unexpected clipping and such. If you set another data, it will remember your choice for a vmax: | dl.set_data(np.arange(5) / 4)
dl.visualize(orientation='horizontal', add_values=True) | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
However vmin has changed, of course. The object remembers what has to be chosen automatically and what was previously set.
User levels
The same is true when the user specifies the levels: | dl = DataLevels(a, cmap=cm, levels=[0, 1, 2, 3])
dl.visualize(orientation='horizontal', add_values=True) | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
Note that the colors have been chosen to cover the full palette, which is much better than the default behavior (see https://github.com/matplotlib/matplotlib/issues/4850).
Since the color choice is automated, when changing the data the colorbar also changes: | dl.set_data(np.arange(5) / 4)
dl.visualize(orientation='horizontal', add_values=True) | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
Since the new data remains within the level range, there is no need to extend the colorbar. Maybe you'd like the two plots above to have the same color code. For this you shoudl set the extend keyword: | dl.set_extend('both')
dl.visualize(orientation='horizontal', add_values=True) | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
Cleo made the choice to warn you if you did something wrong that hides information from the data: | dl = DataLevels(a, cmap=cm, levels=[0, 1, 2, 3], extend='neither')
dl.visualize(orientation='horizontal', add_values=True) | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
Using DataLevels for your own visualisations
The examples above made use of the utilitary function visualize(), which is just here to have a look at the data. Here's a "real world" example with a scatterplot: | # Make the data
x = np.random.randn(1000)
y = np.random.randn(1000)
z = x**2 + y**2
# The datalevel step
dl = DataLevels(z, cmap=cm, levels=np.arange(6))
# The actual plot
fig, ax = plt.subplots(1)
ax.scatter(x, y, color=dl.to_rgb(), s=64)
cbar = dl.append_colorbar(ax, "right", size="5%", pad=0.5) # Note that the Dat... | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
This also works for images: | # Make the data
img, xi, yi = np.histogram2d(x, y)
# The datalevel step
dl = DataLevels(img, cmap=cm, levels=np.arange(6))
# The actual plot
fig, ax = plt.subplots(1)
toplot = dl.to_rgb()
ax.imshow(toplot, interpolation='none')
cbar = dl.append_colorbar(ax, "right", size="5%", pad=0.5) # Note that the DataLevel class h... | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
And with missing data: | cm.set_bad('green')
img[1:5, 1:5] = np.NaN
# The datalevel step
dl = DataLevels(img, cmap=cm, levels=np.arange(6))
# The actual plot
fig, ax = plt.subplots(1)
toplot = dl.to_rgb() # note that the shape is preserved
ax.imshow(toplot, interpolation='none')
cbar = dl.append_colorbar(ax, "right", size="5%", pad=0.5) # No... | examples/DataLevels.ipynb | fmaussion/cleo | gpl-3.0 |
1. Export edges from Retweets, Mentions, or Replies
Run one of three blocks of codes below for your purpose. | # 1. Export edges from Retweets
retweets = tweets[tweets['is_retweet'] == 'Yes']
retweets['original_twitter'] = retweets['text'].str.extract('RT @([a-zA-Z0-9]\w{0,}):', expand=True)
edges = retweets[['screen_name', 'original_twitter','created_at']]
edges.columns = ['Source', 'Target', 'Strength']
# 2. Export edges f... | 20170720-building-social-network-graphs-CSV.ipynb | gwu-libraries/notebooks | mit |
2. Leave only the tweets whose strength level >= user specified level (directed) | strengthLevel = 3 # Network connection strength level: the number of times in total each of the tweeters responded to or mentioned the other.
# If you have 1 as the level, then all tweeters who mentioned or replied to another at least once will be displayed. But if you have 5, only those who have me... | 20170720-building-social-network-graphs-CSV.ipynb | gwu-libraries/notebooks | mit |
3. Export nodes | # Export nodes from the edges and add node attributes for both Sources and Targets.
users = tweets[['screen_name','followers_count','friends_count']]
users = users.sort_values(['screen_name','followers_count'], ascending=[True, False])
users = users.drop_duplicates(['screen_name'], keep='first')
ids = edges2['Source... | 20170720-building-social-network-graphs-CSV.ipynb | gwu-libraries/notebooks | mit |
4. Export nodes and edges to csv files | # change column names for Kumu import (Run this when using Kumu)
edges2.columns = ['From','To','Strength']
# Print nodes to check
nodes.head(3)
# Print edges to check
edges2.head(3)
# Export nodes and edges to csv files
nodes.to_csv('nodes.csv', encoding='utf-8', index=False)
edges2.to_csv('edges.csv', encoding='utf... | 20170720-building-social-network-graphs-CSV.ipynb | gwu-libraries/notebooks | mit |
O indexador i precisa ser inteiro. Entretanto o array que será indexado pode ser qualquer tipo.
f = row[i]
shape(f) é igual ao shape(i)
dtype(f) é o dtype(row) | row = np.arange(0.,100,10)
print('row:', row)
i = np.array([[3,5,0,8],[4,2,7,1]])
f = row[i]
print('i:', i)
print('f=row[i]\n',f)
print(id(i),id(row),id(f)) | master/tutorial_ti_2.ipynb | robertoalotufo/ia898 | mit |
Vejamos agora o caso bidimensional, apropriado para imagens e a transformação de intensidade.
Seja uma imagem f de dimensões (2,3) com os valores de pixels variando de 0 a 2: | f = np.array([[0, 1, 2],
[2, 0, 1]])
print('f=\n',f) | master/tutorial_ti_2.ipynb | robertoalotufo/ia898 | mit |
Seja agora a transformação de intensidade T, especificada por um vetor de 3 elementos, onde
T[0] = 5; T[1] = 6 e T[2] = 7: | T = np.array([5, 6, 7])
print('T:', T)
for i in np.arange(T.size):
print('%d:%d'% (i,T[i])) | master/tutorial_ti_2.ipynb | robertoalotufo/ia898 | mit |
A aplicação da transformação de intensidade é feita utilizando-se a imagem f como índice da
transformação T, como se escreve na equação matemática: | g = T[f]
print('g=T[f]= \n', g)
print('g.shape:', g.shape) | master/tutorial_ti_2.ipynb | robertoalotufo/ia898 | mit |
Note que T[f] tem as mesmas dimensões de f, entretanto, seus pixels passaram pelo
mapeamento da tabela T.
Utilização em imagens
Existem muitas funções úteis que podem ser feitas com o mapeamento T: realce de contraste, equalização de
histograma, thresholding, redução de níveis de cinza, negativo da imagem, entre várias... | T1 = np.arange(256).astype('uint8') # função identidade
T2 = ia.normalize(np.log(T1+1.)) # logaritmica - realce partes escuras
T3 = 255 - T1 # negativo
T4 = ia.normalize(T1 > 128) # threshold 128
T5 = ia.normalize(T1//30) # reduz o número de níveis de cinza
plt.plot(T1)
plt.plot(T2)
plt.plot(T3)
plt.plot(T4)
plt.plot(T... | master/tutorial_ti_2.ipynb | robertoalotufo/ia898 | mit |
Veja a aplicação destas tabelas na imagem "cameraman.tif":
T1: Função identidade | nb = ia.nbshow(2)
f = mpimg.imread('../data/cameraman.tif')
f1 = T1[f]
nb.nbshow(f,'original')
plt.plot(T1)
plt.title('T1: identidade')
nb.nbshow(f1,'T1[f]')
nb.nbshow() | master/tutorial_ti_2.ipynb | robertoalotufo/ia898 | mit |
T2: Função logaritmica | f2 = T2[f]
nb.nbshow(f,'original')
plt.plot(T2)
plt.title('T2: logaritmica')
nb.nbshow(f2,'T2[f]')
nb.nbshow() | master/tutorial_ti_2.ipynb | robertoalotufo/ia898 | mit |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.