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 |
|---|---|---|---|---|---|
Exercise 1. Get a better fit to the data (create a better model & fit it) - try using different optimizers in scipy.optimize | arr = np.arange(100)
arr[10:50]
arr[slice(10,50)] | _____no_output_____ | BSD-3-Clause | notebooks/MarchApril2016_TutorialSession/Notebook - March 18 - Part 1.ipynb | ESO-python/ESOPythonTutorials |
Cloning the pyprobml repo | !git clone https://github.com/probml/pyprobml
%cd pyprobml/scripts | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Installing required software (This may take few minutes) | !apt-get install octave -qq > /dev/null
!apt-get install liboctave-dev -qq > /dev/null
%%capture
%load_ext autoreload
%autoreload 2
DISCLAIMER = 'WARNING : Editing in VM - changes lost after reboot!!'
from google.colab import files
def interactive_script(script, i=True):
if i:
s = open(script).read()
if no... | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Figure 16.1: (a) Illustration of a $K$-nearest neighbors classifier in 2d for $K=5$. The nearest neighbors of test point $\mathbf x $ have labels $\ 1, 1, 1, 0, 0\ $, so we predict $p(y=1|\mathbf x , \mathcal D ) = 3/5$. (b) Illustration of the Voronoi tesselation induced by 1-NN. Adapted from Figure 4.13 of [... | interactive_script("knn_voronoi_plot.py") | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Figure 16.2: Decision boundaries induced by a KNN classifier. (a) $K=1$. (b) $K=2$. (c) $K=5$. (d) Train and test error vs $K$. Figure(s) generated by [knn_classify_demo.py](https://github.com/probml/pyprobml/blob/master/scripts/knn_classify_demo.py) | interactive_script("knn_classify_demo.py") | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Figure 16.3: Illustration of the curse of dimensionality. (a) We embed a small cube of side $s$ inside a larger unit cube. (b) We plot the edge length of a cube needed to cover a given volume of the unit cube as a function of the number of dimensions. Adapted from Figure 2.6 from [HTF09] . Figure(s) generated by ... | interactive_script("curse_dimensionality.py") | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Figure 16.4: Illustration of latent coincidence analysis (LCA) as a directed graphical model. The inputs $\mathbf x , \mathbf x ' \in \mathbb R ^D$ are mapped into Gaussian latent variables $\mathbf z , \mathbf z ' \in \mathbb R ^L$ via a linear mapping $\mathbf W $. If the two latent points coincide (withi... | show_image("/content/pyprobml/notebooks/figures/images/LCA-PGM.png") | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Figure 16.5: Networks for deep metric learning. (a) Siamese network. (b) Triplet network. From Figure 5 of [MH19] . Used with kind permission of Mahmut Kaya. . | show_image("/content/pyprobml/notebooks/figures/images/siameseNet.png")
show_image("/content/pyprobml/notebooks/figures/images/tripletNet.png") | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Figure 16.6: Speeding up triplet loss minimization. (a) Illustration of hard vs easy negatives. Here $a$ is the anchor point, $p$ is a positive point, and $n_i$ are negative points. Adapted from Figure 4 of [MH19] . (b) Standard triplet loss would take $8 \times 3 \times 4 = 96$ calculations, whereas using a proxy... | show_image("/content/pyprobml/notebooks/figures/images/hard-negative-mining.png")
show_image("/content/pyprobml/notebooks/figures/images/tripletBound.png") | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Figure 16.7: Adding spherical embedding constraint to a deep metric learning method. Used with kind permission of Dingyi Zhang. | show_image("/content/pyprobml/notebooks/figures/images/SEC.png") | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Figure 16.8: A comparison of some popular normalized kernels. Figure(s) generated by [smoothingKernelPlot.m](https://github.com/probml/pmtk3/blob/master/demos/smoothingKernelPlot.m) | !octave -W smoothingKernelPlot.m >> _ | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Figure 16.9: A nonparametric (Parzen) density estimator in 1d estimated from 6 data points, denoted by x. Top row: uniform kernel. Bottom row: Gaussian kernel. Left column: bandwidth parameter $h=1$. Right column: bandwidth parameter $h=2$. Adapted from http://en.wikipedia.org/wiki/Kernel_density_estimation . F... | interactive_script("parzen_window_demo2.py") | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Figure 16.10: An example of kernel regression in 1d using a Gaussian kernel. Figure(s) generated by [kernelRegressionDemo.m](https://github.com/probml/pmtk3/blob/master/demos/kernelRegressionDemo.m) | !octave -W kernelRegressionDemo.m >> _ | _____no_output_____ | MIT | notebooks/figures/chapter16_figures.ipynb | kzymgch/pyprobml |
Подключаем необходимые библиотеки | import os
import datetime
import numpy as np
import pandas as pd
from sqlalchemy import create_engine | _____no_output_____ | MIT | project1_sql/scripts/requests.ipynb | mikhailbadin/ds_couse_homeworks |
Настраиваем подключение к БД | engine = create_engine("postgresql://postgres:@{}:{}".format(os.environ["POSGRES_HOST"], os.environ['POSGRES_PORT'])) | _____no_output_____ | MIT | project1_sql/scripts/requests.ipynb | mikhailbadin/ds_couse_homeworks |
Запрос 2: Подсчитать общее количест получателей | emailreceivers = pd.read_sql("select * from emailreceivers", engine)
req2_result = emailreceivers["personid"].drop_duplicates().count()
print(req2_result)
pd.DataFrame(
[req2_result],
columns=['count']).to_csv(os.environ['PANDAS_EXPORT_FOLDER'] + 'req2.csv') | 418
| MIT | project1_sql/scripts/requests.ipynb | mikhailbadin/ds_couse_homeworks |
Запрос 3: Подсчитать количество отправленных писем за 2012 | emails = pd.read_sql("select * from emails", engine)
req3_result = emails[
(emails["metadatadatesent"] >= np.datetime64('2012-01-01')) &
(emails["metadatadatesent"] < np.datetime64('2013-01-01'))
]["id"].count()
print(req3_result)
pd.DataFrame(
[req3_result],
columns=['emails']).to_csv(os.environ['PAND... | 1500
| MIT | project1_sql/scripts/requests.ipynb | mikhailbadin/ds_couse_homeworks |
Запрос 5:Вывести список писем в следующем формате:Отправитель письма, получатель письма, тема письма,и отсортированы по теме письма. | emails = pd.read_sql("select * from emails", engine)
req5_df = emails[
["metadatafrom", "metadatato", "metadatasubject"]
].sort_values(by=["metadatasubject"])
print(req5_df.head())
req5_df.head().to_csv(os.environ['PANDAS_EXPORT_FOLDER'] + 'req5.csv') | metadatafrom metadatato metadatasubject
35711 Mills, Cheryl D H - MORE WHEN WE SPEAK
3951 Mills, Cheryl D H - MORE WHEN WE SPEAK
1528 hrod17@clintonemail.com millscd@state.gov - MORE WHEN WE SPEAK
17504 hrod17@clintonemail.c... | MIT | project1_sql/scripts/requests.ipynb | mikhailbadin/ds_couse_homeworks |
Запрос 6:Вывести среднюю длинну сообщения | emails = pd.read_sql("select * from emails", engine)
req6_result = emails["extractedbodytext"].str.len().mean()
print(req6_result)
pd.DataFrame(
[req6_result],
columns=['avg']).to_csv(os.environ['PANDAS_EXPORT_FOLDER'] + 'req6.csv')
| 533.1624147137348
| MIT | project1_sql/scripts/requests.ipynb | mikhailbadin/ds_couse_homeworks |
Запрос 7:Вывести количество писем, в которых содержится подстрока UNCLASSIFIED | emails = pd.read_sql("select * from emails", engine)
req7_result = emails[
emails["extractedbodytext"].str.contains("UNCLASSIFIED") == True
]["extractedbodytext"].count()
print(req7_result)
pd.DataFrame(
[req7_result],
columns=['text']).to_csv(os.environ['PANDAS_EXPORT_FOLDER'] + 'req7.csv') | 78
| MIT | project1_sql/scripts/requests.ipynb | mikhailbadin/ds_couse_homeworks |
20 Jan | import pandas as pd
import numpy as np
tips = pd.read_csv("/Users/amitkumarsahu/Desktop/pandas/tips.csv")
tips
tips.pivot_table(index=["day", "smoker"])
tips["tip_pct"]=tips["tip"]*100/tips["total_bill"]
tips[:6]
tips.pivot_table(["tip_pct", "size"], index=["time", "day"], columns="smoker") | _____no_output_____ | CC-BY-3.0 | Pivot_table and crosstab.ipynb | priyankakushi/machine-learning |
21 Jan | tips.pivot_table(["tip_pct", "size"], index=["time", "day"],
columns="smoker", margins=True)
tips.pivot_table("tip", index=["time", "smoker"], columns="day", aggfunc=len, margins=True)
tips.pivot_table("tip_pct", index=["time", "size", "smoker"], columns="day", aggfunc="mean", fill_value=0)
from io impo... | _____no_output_____ | CC-BY-3.0 | Pivot_table and crosstab.ipynb | priyankakushi/machine-learning |
OverviewAnother data type that we haven't discussed yet is the list. Lists store values for us, and we can access these values using indices, or by knowing where an element is located to access it. Questions- How can I store multiple values? Objectives- Explain why programs need collections of values.- Write programs... | pressures = [0.273, 0.275, 0.277, 0.276]
print(pressures[0])
print("pressures", pressures)
print("length:", len(pressures)) | pressures [0.273, 0.275, 0.277, 0.276]
length: 4
| MIT | lessonNotebooks/1.11 Lists.ipynb | crowegian/SummerFellowsDBMIWorkshopMaterials |
Use an item’s index to fetch it from a list. | print("Zeroth item:", pressures[0])
print("fourth item:", pressures[3]) | Zeroth item: 0.273
fourth item: 0.276
| MIT | lessonNotebooks/1.11 Lists.ipynb | crowegian/SummerFellowsDBMIWorkshopMaterials |
Lists’ values can be replaced by assigning to them. | print(pressures)
pressures[0] = 0.265
print("current pressures", pressures) | current pressures [0.265, 0.275, 0.277, 0.276]
| MIT | lessonNotebooks/1.11 Lists.ipynb | crowegian/SummerFellowsDBMIWorkshopMaterials |
Appending items to a list lengthens it. | primes = [2, 3, 5]
print("current primes", primes)
primes.append(7)
primes.append(9)
print("primes has become", primes)
primes = primes.append(13)
print(primes)
teen_primes = [11, 13, 17, 19]
middle_aged_primes = [37, 41, 43, 47]
print("primes is currently", primes)
primes.extend(teen_primes)
print("primes is now", pri... | primes is currently [2, 3, 5, 7, 9]
primes is now [2, 3, 5, 7, 9, 11, 13, 17, 19]
| MIT | lessonNotebooks/1.11 Lists.ipynb | crowegian/SummerFellowsDBMIWorkshopMaterials |
Use del to remove items from a list entirely. | primes = [2, 3, 5, 7, 9]
print("current primes", primes)
del primes[2]
print("primes now", primes)
| current primes [2, 3, 5, 7, 9]
primes now [2, 3, 7, 9]
| MIT | lessonNotebooks/1.11 Lists.ipynb | crowegian/SummerFellowsDBMIWorkshopMaterials |
Lists may contain values of different types. | goals = [1, "create lists.", 2, "extract items from lists", 3, "modify lists"]
goals | _____no_output_____ | MIT | lessonNotebooks/1.11 Lists.ipynb | crowegian/SummerFellowsDBMIWorkshopMaterials |
Character strings can be indexed like lists. | element = "carbon"
print(element[0]) | c
| MIT | lessonNotebooks/1.11 Lists.ipynb | crowegian/SummerFellowsDBMIWorkshopMaterials |
Character strings are immutable. | print(element)
element[0] = "C"
print(element) | carbon
| MIT | lessonNotebooks/1.11 Lists.ipynb | crowegian/SummerFellowsDBMIWorkshopMaterials |
Indexing beyond the end of the collection is an error. | print("99th element of element is:", element[99])
primes = [2, 3, 7, 9]
primes.remove(7)
primes
del primes[0]
primes | _____no_output_____ | MIT | lessonNotebooks/1.11 Lists.ipynb | crowegian/SummerFellowsDBMIWorkshopMaterials |
Inference and ValidationNow that you have a trained network, you can use it for making predictions. This is typically called **inference**, a term borrowed from statistics. However, neural networks have a tendency to perform *too well* on the training data and aren't able to generalize to data that hasn't been seen be... | import torch
from torchvision import datasets, transforms
# Define a transform to normalize the data
transform = transforms.Compose([transforms.ToTensor(),
transforms.Normalize((0.5,), (0.5,))])
# Download and load the training data
trainset = datasets.FashionMNIST('~/.pytorch/F_MNIST_d... | _____no_output_____ | Apache-2.0 | 01_introduction/5_Inference_and_Validation.ipynb | thearonn/dtu_mlops |
Here I'll create a model like normal, using the same one from my solution for part 4. | from torch import nn, optim
import torch.nn.functional as F
class Classifier(nn.Module):
def __init__(self):
super().__init__()
self.fc1 = nn.Linear(784, 256)
self.fc2 = nn.Linear(256, 128)
self.fc3 = nn.Linear(128, 64)
self.fc4 = nn.Linear(64, 10)
def forward(s... | _____no_output_____ | Apache-2.0 | 01_introduction/5_Inference_and_Validation.ipynb | thearonn/dtu_mlops |
The goal of validation is to measure the model's performance on data that isn't part of the training set. Performance here is up to the developer to define though. Typically this is just accuracy, the percentage of classes the network predicted correctly. Other options are [precision and recall](https://en.wikipedia.or... | model = Classifier()
images, labels = next(iter(testloader))
# Get the class probabilities
ps = torch.exp(model(images))
# Make sure the shape is appropriate, we should get 10 class probabilities for 64 examples
print(ps.shape) | torch.Size([64, 10])
| Apache-2.0 | 01_introduction/5_Inference_and_Validation.ipynb | thearonn/dtu_mlops |
With the probabilities, we can get the most likely class using the `ps.topk` method. This returns the $k$ highest values. Since we just want the most likely class, we can use `ps.topk(1)`. This returns a tuple of the top-$k$ values and the top-$k$ indices. If the highest value is the fifth element, we'll get back 4 as ... | top_p, top_class = ps.topk(1, dim=1)
# Look at the most likely classes for the first 10 examples
print(top_class[:10,:]) | tensor([[1],
[1],
[1],
[1],
[1],
[1],
[1],
[1],
[1],
[1]])
| Apache-2.0 | 01_introduction/5_Inference_and_Validation.ipynb | thearonn/dtu_mlops |
Now we can check if the predicted classes match the labels. This is simple to do by equating `top_class` and `labels`, but we have to be careful of the shapes. Here `top_class` is a 2D tensor with shape `(64, 1)` while `labels` is 1D with shape `(64)`. To get the equality to work out the way we want, `top_class` and `l... | equals = top_class == labels.view(*top_class.shape)
#print(equals) | _____no_output_____ | Apache-2.0 | 01_introduction/5_Inference_and_Validation.ipynb | thearonn/dtu_mlops |
Now we need to calculate the percentage of correct predictions. `equals` has binary values, either 0 or 1. This means that if we just sum up all the values and divide by the number of values, we get the percentage of correct predictions. This is the same operation as taking the mean, so we can get the accuracy with a c... | accuracy = torch.mean(equals.type(torch.FloatTensor))
print(f'Accuracy: {accuracy.item()*100}%') | Accuracy: 10.9375%
| Apache-2.0 | 01_introduction/5_Inference_and_Validation.ipynb | thearonn/dtu_mlops |
The network is untrained so it's making random guesses and we should see an accuracy around 10%. Now let's train our network and include our validation pass so we can measure how well the network is performing on the test set. Since we're not updating our parameters in the validation pass, we can speed up our code by t... | model = Classifier()
criterion = nn.NLLLoss()
optimizer = optim.Adam(model.parameters(), lr=0.003)
epochs = 30
steps = 0
train_losses, test_losses = [], []
for e in range(epochs):
running_loss = 0
for images, labels in trainloader:
optimizer.zero_grad()
log_ps = model(images)... | racy: 82.8125%
Accuracy: 85.9375%
Accuracy: 75.0%
Accuracy: 90.625%
Accuracy: 89.0625%
Accuracy: 93.75%
Accuracy: 82.8125%
Accuracy: 79.6875%
Accuracy: 79.6875%
Accuracy: 82.8125%
Accuracy: 89.0625%
Accuracy: 87.5%
Accuracy: 90.625%
Accuracy: 85.9375%
Accuracy: 89.0625%
Accuracy: 84.375%
Accuracy: 92.1875%
Accuracy: 82... | Apache-2.0 | 01_introduction/5_Inference_and_Validation.ipynb | thearonn/dtu_mlops |
OverfittingIf we look at the training and validation losses as we train the network, we can see a phenomenon known as overfitting.The network learns the training set better and better, resulting in lower training losses. However, it starts having problems generalizing to data outside the training set leading to the va... | from torch import nn, optim
import torch.nn.functional as F
class Classifier2(nn.Module):
def __init__(self):
super().__init__()
self.fc1 = nn.Linear(784, 256)
self.fc2 = nn.Linear(256, 128)
self.fc3 = nn.Linear(128, 64)
self.fc4 = nn.Linear(64, 10)
self.dropout = n... | _____no_output_____ | Apache-2.0 | 01_introduction/5_Inference_and_Validation.ipynb | thearonn/dtu_mlops |
InferenceNow that the model is trained, we can use it for inference. We've done this before, but now we need to remember to set the model in inference mode with `model.eval()`. You'll also want to turn off autograd with the `torch.no_grad()` context. | # Import helper module (should be in the repo)
import helper
# Test out your network!
model.eval()
dataiter = iter(testloader)
images, labels = dataiter.next()
img = images[0]
# Convert 2D image to 1D vector
img = img.view(1, 784)
# Calculate the class probabilities (softmax) for img
with torch.no_grad():
outpu... | _____no_output_____ | Apache-2.0 | 01_introduction/5_Inference_and_Validation.ipynb | thearonn/dtu_mlops |
TimeEval shared parameter optimization result analysis | # Automatically reload packages:
%load_ext autoreload
%autoreload 2
# imports
import json
import warnings
import pandas as pd
import numpy as np
import scipy as sp
import plotly.offline as py
import plotly.graph_objects as go
import plotly.figure_factory as ff
import plotly.express as px
from plotly.subplots import mak... | _____no_output_____ | MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
ConfigurationTarget parameters that were optimized in this run (per algorithm): | algo_param_mapping = {
"HBOS": ["n_bins"],
"MultiHMM": ["n_bins"],
"MTAD-GAT": ["context_window_size", "mag_window_size", "score_window_size"],
"PST": ["n_bins"]
} | _____no_output_____ | MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
Define data and results folder: | # constants and configuration
data_path = Path("../../data") / "test-cases"
result_root_path = Path("../timeeval_experiments/results")
experiment_result_folder = "2021-10-04_shared-optim2"
# build paths
result_paths = [d for d in result_root_path.iterdir() if d.is_dir()]
print("Available result directories:")
display(... | Available result directories:
| MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
Load results and dataset metadata: | def extract_hyper_params(param_names):
def extract(value):
params = json.loads(value)
result = None
for name in param_names:
try:
value = params[name]
result = pd.Series([name, value], index=["optim_param_name", "optim_param_value"])
... | Reading results from /home/sebastian/Documents/Projects/akita/timeeval/timeeval_experiments/results/2021-10-04_shared-optim2
| MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
Define plotting functions: | def load_scores_df(algorithm_name, dataset_id, optim_params, repetition=1):
params_id = df.loc[(df["algorithm"] == algorithm_name) & (df["collection"] == dataset_id[0]) & (df["dataset"] == dataset_id[1]) & (df["optim_param_name"] == optim_params[0]) & (df["optim_param_value"] == optim_params[1]), "hyper_params_id"]... | _____no_output_____ | MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
Analyze TimeEval results | df[["algorithm", "dataset_name", "status", "AVERAGE_PRECISION", "PR_AUC", "RANGE_PR_AUC", "ROC_AUC", "execute_main_time", "optim_param_name", "optim_param_value"]] | _____no_output_____ | MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
--- Errors | df_error_counts = df.pivot_table(index=["algo_training_type", "algorithm"], columns=["status"], values="repetition", aggfunc="count")
df_error_counts = df_error_counts.fillna(value=0).astype(np.int64) | _____no_output_____ | MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
Aggregation of errors per algorithm grouped by algorithm training type | for tpe in ["SEMI_SUPERVISED", "SUPERVISED", "UNSUPERVISED"]:
if tpe in df_error_counts.index:
print(tpe)
display(df_error_counts.loc[tpe]) | SEMI_SUPERVISED
| MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
Slow algorithmsAlgorithms, for which more than 50% of all executions ran into the timeout. | df_error_counts[df_error_counts["Status.TIMEOUT"] > (df_error_counts["Status.ERROR"] + df_error_counts["Status.OK"])] | _____no_output_____ | MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
Broken algorithmsAlgorithms, which failed for at least 50% of the executions. | error_threshold = 0.5
df_error_counts[df_error_counts["Status.ERROR"] > error_threshold*(
df_error_counts["Status.TIMEOUT"] + df_error_counts["Status.ERROR"] + df_error_counts["Status.OK"]
)] | _____no_output_____ | MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
Detail errors | algo_list = ["MTAD-GAT", "MultiHMM"]
error_list = ["OOM", "Segfault", "ZeroDivisionError", "IncompatibleParameterConfig", "WrongDBNState", "SyntaxError", "other"]
errors = pd.DataFrame(0, index=error_list, columns=algo_list, dtype=np.int_)
for algo in algo_list:
df_tmp = df[(df["algorithm"] == algo) & (df["status"... | _____no_output_____ | MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
--- Parameter assessment | sort_by = ("ROC_AUC", "mean")
metric_agg_type = ["mean", "median"]
time_agg_type = "mean"
aggs = {
"AVERAGE_PRECISION": metric_agg_type,
"RANGE_PR_AUC": metric_agg_type,
"PR_AUC": metric_agg_type,
"ROC_AUC": metric_agg_type,
"train_main_time": time_agg_type,
"execute_main_time": time_agg_type,
... | _____no_output_____ | MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
Selected parameters- HBOS: `n_bins=20` (more is better)- MultiHMM: `n_bins=5` (8 is slightly better, but takes way longer. The scores are very bad anyway!)- MTAD-GAT: `context_window_size=30,mag_window_size=40,score_window_size=52` (very slow)- PST: `n_bins=5` (less is better)> **Note**>> MTAD-GAT is very slow! Exclud... | plot_scores([("MultiHMM", "n_bins", 5), ("MultiHMM", "n_bins", 8)], "sinus-type-mean")
plot_scores([("MTAD-GAT", "context_window_size", 30), ("MTAD-GAT", "context_window_size", 40)], "sinus-type-mean") | _____no_output_____ | MIT | notebooks/TimeEval shared param optimization analysis 2.ipynb | HPI-Information-Systems/TimeEval |
Tacotron 2 inference code Edit the variables **checkpoint_path** and **text** to match yours and run the entire code to generate plots of mel outputs, alignments and audio synthesis from the generated mel-spectrogram using Griffin-Lim. Import libraries and setup matplotlib | import matplotlib
%matplotlib inline
import matplotlib.pylab as plt
import IPython.display as ipd
import sys
sys.path.append('waveglow/')
import numpy as np
import torch
from hparams import create_hparams
from model import Tacotron2
from layers import TacotronSTFT, STFT
from audio_processing import griffin_lim
from... | _____no_output_____ | BSD-3-Clause | inference.ipynb | ncantrell/tacotron2 |
Setup hparams | hparams = create_hparams()
hparams.sampling_rate = 22050 |
WARNING: The TensorFlow contrib module will not be included in TensorFlow 2.0.
For more information, please see:
* https://github.com/tensorflow/community/blob/master/rfcs/20180907-contrib-sunset.md
* https://github.com/tensorflow/addons
If you depend on functionality not listed there, please file an issue.
| BSD-3-Clause | inference.ipynb | ncantrell/tacotron2 |
Load model from checkpoint | checkpoint_path = "tacotron2_statedict.pt"
model = load_model(hparams)
model.load_state_dict(torch.load(checkpoint_path)['state_dict'])
_ = model.cuda().eval().half() | _____no_output_____ | BSD-3-Clause | inference.ipynb | ncantrell/tacotron2 |
Load WaveGlow for mel2audio synthesis and denoiser | waveglow_path = 'waveglow_256channels.pt'
waveglow = torch.load(waveglow_path)['model']
waveglow.cuda().eval().half()
for m in waveglow.modules():
if 'Conv' in str(type(m)):
setattr(m, 'padding_mode', 'zeros')
for k in waveglow.convinv:
k.float()
denoiser = Denoiser(waveglow) | /home/kabakov/VOICE/venv/lib/python3.6/site-packages/torch/serialization.py:454: SourceChangeWarning: source code of class 'torch.nn.modules.conv.ConvTranspose1d' has changed. you can retrieve the original source code by accessing the object's source attribute or set `torch.nn.Module.dump_patches = True` and use the pa... | BSD-3-Clause | inference.ipynb | ncantrell/tacotron2 |
Prepare text input | #%%timeit 77.9 µs ± 237 ns
text = "Waveglow is really awesome!"
sequence = np.array(text_to_sequence(text, ['english_cleaners']))[None, :]
sequence = torch.autograd.Variable(
torch.from_numpy(sequence)).cuda().long() | _____no_output_____ | BSD-3-Clause | inference.ipynb | ncantrell/tacotron2 |
Decode text input and plot results | #%%timeit 240 ms ± 9.72 ms
mel_outputs, mel_outputs_postnet, _, alignments = model.inference(sequence)
plot_data((mel_outputs.float().data.cpu().numpy()[0],
mel_outputs_postnet.float().data.cpu().numpy()[0],
alignments.float().data.cpu().numpy()[0].T)) | _____no_output_____ | BSD-3-Clause | inference.ipynb | ncantrell/tacotron2 |
Synthesize audio from spectrogram using WaveGlow | #%%timeit 193 ms ± 4.87 ms
with torch.no_grad():
audio = waveglow.infer(mel_outputs_postnet, sigma=0.666)
ipd.Audio(audio[0].data.cpu().numpy(), rate=hparams.sampling_rate) | waveglow/glow.py:162: RuntimeWarning: nn.functional.tanh is deprecated. Use torch.tanh instead.
torch.IntTensor([self.n_channels]))
waveglow/glow.py:162: RuntimeWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.
torch.IntTensor([self.n_channels]))
| BSD-3-Clause | inference.ipynb | ncantrell/tacotron2 |
(Optional) Remove WaveGlow bias | audio_denoised = denoiser(audio, strength=0.01)[:, 0]
ipd.Audio(audio_denoised.cpu().numpy(), rate=hparams.sampling_rate) | _____no_output_____ | BSD-3-Clause | inference.ipynb | ncantrell/tacotron2 |
Save result as wav | import librosa
# save
librosa.output.write_wav('./out.wav', audio[0].data.cpu().numpy().astype(np.float32), 22050)
# check
y, sr = librosa.load('out.wav')
ipd.Audio(y, rate=sr) | _____no_output_____ | BSD-3-Clause | inference.ipynb | ncantrell/tacotron2 |
Table of Contents1 label identity hairstyle2 Prepare hairstyle images3 prepare hairstyle manifest | from query.models import Video, FaceIdentity, Identity
from esper.widget import *
from esper.prelude import collect, esper_widget
import pickle
import os
import random
get_ipython().magic('matplotlib inline')
get_ipython().magic('reload_ext autoreload')
get_ipython().magic('autoreload 2') | _____no_output_____ | Apache-2.0 | app/notebooks/hairstyle.ipynb | scanner-research/esper-tv |
label identity hairstyle | identity_hair_dict = {}
identities = Identity.objects.all()
identity_list = [(i.id, i.name) for i in identities]
identity_list.sort()
# 154
hair_color_3 = {0: 'black', 1: 'white', 2: 'blond'}
hair_color_5 = {0: 'black', 1: 'white', 2: 'blond', 3: 'brown', 4: 'gray'}
hair_length = {0: 'long', 1: 'medium', 2: 'short', 3:... | _____no_output_____ | Apache-2.0 | app/notebooks/hairstyle.ipynb | scanner-research/esper-tv |
Prepare hairstyle images | faceIdentities = FaceIdentity.objects \
.filter(identity__name='melania trump') \
.filter(probability__gt=0.9) \
.select_related('face__frame__video')
faceIdentities_sampled = random.sample(list(faceIdentities), 1000)
print("Load %d face identities" % len(faceIdentities_sampled))
identity_grouped = collect(... | _____no_output_____ | Apache-2.0 | app/notebooks/hairstyle.ipynb | scanner-research/esper-tv |
prepare hairstyle manifest | img_list = os.listdir('/app/result/clothing/images/')
len(img_list)
group_by_identity = {}
for name in img_list:
iid = int(name.split('_')[0])
if iid not in group_by_identity:
group_by_identity[iid] = []
else:
group_by_identity[iid].append(name)
identity_label = [id for id, img_list in group... | _____no_output_____ | Apache-2.0 | app/notebooks/hairstyle.ipynb | scanner-research/esper-tv |
loading dataset | data = pd.read_csv("student-data.csv")
data.head()
data.shape
type(data) | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
Exploratory data analysis | import matplotlib.pyplot as plt
import seaborn as sns
a = data.plot()
data.info()
data.isnull().sum()
a = sns.heatmap(data.isnull(),cmap='Blues')
a = sns.heatmap(data.isnull(),cmap='Blues',yticklabels=False) | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
this indicates that we have no any null values in the dataset | a = sns.heatmap(data.isna(),yticklabels=False) | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
this heatmap indicates that we have no any 'NA' values in the dataset | sns.set(style='darkgrid')
sns.countplot(data=data,x='reason') | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
This indicates the count for choosing school of various reasons.A count plot can be thought of as a histogram across a categorical, instead of quantitative, variable. | data.head(7) | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
calculating total passed students | passed = data.loc[data.passed == 'yes']
passed.shape
tot_passed=passed.shape[0]
print('total passed students is: {} '.format(tot_passed)) | total passed students is: 265
| MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
calculating total failed students | failed = data.loc[data.passed == 'no']
print('total failed students is: {}'.format(failed.shape[0])) | total failed students is: 130
| MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
Feature Engineering | data.head() | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
To identity feature and target variable lets first do some feature engineering stuff! | data.columns
data.columns[-1] | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
Here 'passed' is our target variable. Since in this system we need to develop the model that will predict the likelihood that a given student will pass, quantifying whether an intervention is necessary. | target = data.columns[-1]
data.columns[:-1]
#initially taking all columns as our feature variables
feature = list(data.columns[:-1])
data[target].head()
data[feature].head() | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
Now taking feature and target data in seperate dataframe | featuredata = data[feature]
targetdata = data[target] | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
Now we need to convert several non-numeric columns like 'internet' into numerical form for the model to process | def preprocess_features(X):
output = pd.DataFrame(index = X.index)
for col, col_data in X.iteritems():
if col_data.dtype == object:
col_data = col_data.replace(['yes', 'no'], [1, 0])
if col_data.dtype == object:
col_data = pd.get_dummies(col_data, prefix = col)
... | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
MODEL IMPLEMENTATION Decision tree | from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import train_test_split
model=DecisionTreeClassifier()
X_train, X_test, y_train, y_test = train_test_split(featuredata, targetdata, test_size=0.33, random_state=6)
model.fit(X_train,y_train)
from sklearn.metrics import accuracy_score
predictio... | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
K-Nearest Neighbours | from sklearn.neighbors import KNeighborsClassifier
new_classifier = KNeighborsClassifier(n_neighbors=7)
new_classifier.fit(X_train,y_train)
predictions2 = new_classifier.predict(X_test)
accuracy_score(y_test,predictions2)*100 | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
SVM | from sklearn import svm
clf = svm.SVC(random_state=6)
clf.fit(featuredata,targetdata)
clf.score(featuredata,targetdata)
predictions3= clf.predict(X_test)
accuracy_score(y_test,predictions3)*100 | _____no_output_____ | MIT | 020_NIRMAL.ipynb | NirmalSilwal/Machine-Learning |
k-means clustering | import warnings
warnings.filterwarnings('ignore')
%matplotlib inline
import scipy as sc
import scipy.stats as stats
from scipy.spatial.distance import euclidean
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors
plt.style.use('fivethirtyeight')
plt.rcParams['fo... | _____no_output_____ | MIT | codici/kmeans.ipynb | tvml/ml2122 |
1. Install DependenciesFirst install the libraries needed to execute recipes, this only needs to be done once, then click play. | !pip install git+https://github.com/google/starthinker
| _____no_output_____ | Apache-2.0 | colabs/smartsheet_to_bigquery.ipynb | fivestones-apac/starthinker |
2. Get Cloud Project IDTo run this recipe [requires a Google Cloud Project](https://github.com/google/starthinker/blob/master/tutorials/cloud_project.md), this only needs to be done once, then click play. | CLOUD_PROJECT = 'PASTE PROJECT ID HERE'
print("Cloud Project Set To: %s" % CLOUD_PROJECT)
| _____no_output_____ | Apache-2.0 | colabs/smartsheet_to_bigquery.ipynb | fivestones-apac/starthinker |
3. Get Client CredentialsTo read and write to various endpoints requires [downloading client credentials](https://github.com/google/starthinker/blob/master/tutorials/cloud_client_installed.md), this only needs to be done once, then click play. | CLIENT_CREDENTIALS = 'PASTE CREDENTIALS HERE'
print("Client Credentials Set To: %s" % CLIENT_CREDENTIALS)
| _____no_output_____ | Apache-2.0 | colabs/smartsheet_to_bigquery.ipynb | fivestones-apac/starthinker |
4. Enter SmartSheet Sheet To BigQuery ParametersMove sheet data into a BigQuery table. 1. Specify SmartSheet token. 1. Locate the ID of a sheet by viewing its properties. 1. Provide a BigQuery dataset ( must exist ) and table to write the data into. 1. StarThinker will automatically map the correct schema.Modify the va... | FIELDS = {
'auth_read': 'user', # Credentials used for reading data.
'auth_write': 'service', # Credentials used for writing data.
'token': '', # Retrieve from SmartSheet account settings.
'sheet': '', # Retrieve from sheet properties.
'dataset': '', # Existing BigQuery dataset.
'table': '', # Table t... | _____no_output_____ | Apache-2.0 | colabs/smartsheet_to_bigquery.ipynb | fivestones-apac/starthinker |
5. Execute SmartSheet Sheet To BigQueryThis does NOT need to be modified unless you are changing the recipe, click play. | from starthinker.util.project import project
from starthinker.script.parse import json_set_fields
USER_CREDENTIALS = '/content/user.json'
TASKS = [
{
'smartsheet': {
'auth': 'user',
'token': {'field': {'name': 'token','kind': 'string','order': 2,'default': '','description': 'Retrieve from SmartSheet... | _____no_output_____ | Apache-2.0 | colabs/smartsheet_to_bigquery.ipynb | fivestones-apac/starthinker |
TD: prédiction du vote 2016 aux Etats-Unis par arbres de décisions et méthodes ensemblistesLa séance d'aujourd'hui porte sur la prévision du vote en 2016 aux États-Unis. Précisément, les données d'un recensement sont fournies avec diverses informations par comté à travers les États-Unis. L'objectif est de construire ... | %matplotlib inline
from pylab import *
import numpy as np
import os
import random
import matplotlib.pyplot as plt | _____no_output_____ | MIT | MI203-td2_tree_and_forest.ipynb | NataliaDiaz/colab |
Accès aux données* Elles sont disponibles: https://github.com/stepherbin/teaching/tree/master/ENSTA/TD2* Charger le fichier the combined_data.csv sur votre drive puis monter le depuis colab | USE_COLAB = True
UPLOAD_OUTPUTS = False
if USE_COLAB:
# mount the google drive
from google.colab import drive
drive.mount('/content/drive', force_remount=True)
# download data on GoogleDrive
data_dir = "/content/drive/My Drive/teaching/ENSTA/TD_tree/"
else:
data_dir = "data/"
import pandas as pd... | _____no_output_____ | MIT | MI203-td2_tree_and_forest.ipynb | NataliaDiaz/colab |
Analyse préliminaire des donnéesLes données sont organisées en champs:* fips = code du comté à 5 chiffres, le premier ou les deux premiers chiffres indiquent l'état.* votes = nombre de votants* etc..Regarder leur structure, quantité, nature.Où se trouvent les informations pour former les ensembles d'apprentissage et d... | # Exemples de moyens d'accéder aux caractéristiques des données
print(census_data.shape )
print(census_data.columns.values)
print(census_data['fips'])
print(census_data.head(3))
iattr = 10
attrname = census_data.columns[iattr]
print("Mean of {} is {:.1f}".format(attrname,np.array(census_data[attrname]).mean()))
####... | _____no_output_____ | MIT | MI203-td2_tree_and_forest.ipynb | NataliaDiaz/colab |
La classe à prédire ('Democrat') n'est décrite que par un seul attribut binaire.Calculer la répartition des couleurs politiques (quel est a priori la probabilité qu'un comté soit démocrate vs. républicain) | #########################
## METTRE VOTRE CODE ICI
#########################
print("La probabilité qu'un comté soit démocrate est de {:.2f}%%".format(100*proba_dem))
| La probabilité qu'un comté soit démocrate est de 15.45%%
| MIT | MI203-td2_tree_and_forest.ipynb | NataliaDiaz/colab |
Préparation du chantier d'apprentissageOn va préparer les ensembles d'apprentissage et de test. Pour éviter des problèmes de format de données, on choisit une liste d'attributs utiles dans la liste "feature_cols" ci dessous.L'ensemble de test sera constitué des comtés d'un seul état.Info: https://scikit-learn.org/stab... | ## Sous ensembles d'attributs informatifs pour la suite
feature_cols = ['BLACK_FEMALE_rate',
'BLACK_MALE_rate',
'Percent of adults with a bachelor\'s degree or higher, 2011-2015',
'ASIAN_MALE_rate',
'ASIAN_FEMALE_rate',
'25-29_rate',
... | BLACK_FEMALE_rate BLACK_MALE_rate ... Protestant Denomination Zoroastrian
0 0.067586 0.062079 ... 0 0
1 0.067586 0.062079 ... 0 0
[2 rows x 49 columns]
598 0
599 0
Name: Democrat, dtype: int64
| MIT | MI203-td2_tree_and_forest.ipynb | NataliaDiaz/colab |
Apprentissage d'un arbre de décisionOn utilisera la bibliothèque scikit learn * Construire l'arbre sur les données d'entrainement* Prédire le vote sur les comtés de test* Calculer l'erreur et la matrice de confusionFaire varier certains paramètres (profondeur max, pureté, critère...) et visualisez leur influence.Info:... |
from sklearn import tree
#########################
## METTRE VOTRE CODE ICI
#########################
| _____no_output_____ | MIT | MI203-td2_tree_and_forest.ipynb | NataliaDiaz/colab |
Les instructions suivantes permettent de visualiser l'arbre.Interpréter le contenu de la représentation. | import graphviz
dot_data = tree.export_graphviz(clf, out_file=None)
graph = graphviz.Source(dot_data)
dot_data = tree.export_graphviz(clf, out_file=None,
feature_names=X_train.columns.values,
class_names=["R","D"],
filled=True, rounded=True,
... | Predictions per county in state #34 are [0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 1 0 1 0 0 0 0 1 0 0 0
0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 ... | MIT | MI203-td2_tree_and_forest.ipynb | NataliaDiaz/colab |
--- BaggingL'objectif de cette partie est de construire **à la main** une approche de bagging.Le principe de l'approche est de:* Apprendre et collecter plusieurs arbres sur des échantillonnages aléatoires des données d'apprentissage* Agréger les prédictions par vote * Evaluer: Les prédictions agrégées* Comparer avec le... | from sklearn.model_selection import train_test_split
# Données d'apprentissage: X_train, y_train, idx_train
# Données de test: X_test, y_test, idx_test
# Les étapes de conception du prédicteur (apprentissage) sont les suivantes:
# - Construction des sous-ensembles de données
# - Apprentissage d'un arbre
# - Agré... | Taux de bonne prédiction = 76.32%
0.6837740384615385
| MIT | MI203-td2_tree_and_forest.ipynb | NataliaDiaz/colab |
Osnabrück University - Machine Learning (Summer Term 2018) - Prof. Dr.-Ing. G. Heidemann, Ulf Krumnack Exercise Sheet 08 IntroductionThis week's sheet should be solved and handed in before the end of **Sunday, June 3, 2018**. If you need help (and Google and other resources were not enough), feel free to contact your... | import numpy as np
# Activation function σ.
# We use scipy's builtin because it fixes some NaN problems for us.
# sigmoid = lambda x: 1 / (1 + np.exp(-x))
from scipy.special import expit as sigmoid
class Perceptron:
"""Single neuron handling its own weights and bias."""
def __init__(self, dim_in, act_func=s... | _____no_output_____ | MIT | sheet_08/sheet_08_machine-learning_solution.ipynb | ArielMant0/ml2018 |
PerceptronLayerA `PerceptronLayer` is a combination of multiple `Perceptron` instances. It mainly is concerened with passing input and delta values to its individual neurons. There is no math to be done here! InitializationWhen initializing a `PerceptronLayer` (like this: `layer = PerceptronLayer(5, 3)`), the `__init_... | class PerceptronLayer:
"""Layer of multiple neurons.
Attributes:
perceptrons (list): List of perceptron instances in the layer.
"""
def __init__(self, dim_in, dim_out, act_func=sigmoid):
"""Initialize the layer as a list of individual neurons.
A layer contains as many neuro... | _____no_output_____ | MIT | sheet_08/sheet_08_machine-learning_solution.ipynb | ArielMant0/ml2018 |
MultilayerPerceptron Forward-Propagation / ActivationPropagate the input value $x$ through each layer of the network, employing the output of the previous layer as input to the next layer. Back-Propagation / AdaptationThis is the most complex step of the whole task. Split into three separate parts:1. ***Forward propag... | class MultilayerPerceptron:
"""Network of perceptrons, also a set of multiple perceptron layers.
Attributes:
layers (list): List of perceptron layers in the network.
"""
def __init__(self, *layers):
"""Initialize a new network, madeup of individual PerceptronLayers.
Args:
... | _____no_output_____ | MIT | sheet_08/sheet_08_machine-learning_solution.ipynb | ArielMant0/ml2018 |
Classification Problem DefinitionBefore we start, we need a problem to solve. In the following cell we first generate some three dimensional data (= $\text{input_dim}$) between 0 and 1 and label all data according to a binary classification: If the data is close to the center (radius < 2.5), it belongs to one class, ... | def uniform(a, b, n=1):
"""Returns n floats uniformly distributed between a and b."""
return (b - a) * np.random.random_sample(n) + a
n = 1000
radius = 5
r = np.append(uniform(0, radius * .5, n // 2),
uniform(radius * .7, radius, n // 2))
angle = uniform(0, 2 * np.pi, n)
x = r * np.sin(angle) + ... | _____no_output_____ | MIT | sheet_08/sheet_08_machine-learning_solution.ipynb | ArielMant0/ml2018 |
Model DesignThe following cell already contains a simple model with a single layer. Play around with some different configurations! | MLP = MultilayerPerceptron(
PerceptronLayer(2, 1),
)
# Adapt this MLP
### BEGIN SOLUTION
MLP = MultilayerPerceptron(
PerceptronLayer(2, 4),
PerceptronLayer(4, 2),
PerceptronLayer(2, 1),
)
### END SOLUTION | _____no_output_____ | MIT | sheet_08/sheet_08_machine-learning_solution.ipynb | ArielMant0/ml2018 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.