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 |
|---|---|---|---|---|---|
Define the parameter for the photometer reading | Y_e = ufloat(2673.3,1.)
Y_e | _____no_output_____ | CC0-1.0 | empir19nrm02/Jupyter/IBudgetMETAS.ipynb | AndersThorseth/empir19nrm02 |
Define the parameter for the distance measurement | d=ufloat(25.0000, 0.0025)
d | _____no_output_____ | CC0-1.0 | empir19nrm02/Jupyter/IBudgetMETAS.ipynb | AndersThorseth/empir19nrm02 |
The Model | I=k_e*Y_e*d**2
I
[h, result_vecotr] = uncLib_PlotHist(I, xLabel='Luminous intensity / cd')
print('Mean: {}, I0: {}, I1: {}'.format(result_vecotr[0], result_vecotr[1], result_vecotr[2]))
h=uncLib_PlotHist(k_e, xLabel='calibration factor / lx/LSB') | _____no_output_____ | CC0-1.0 | empir19nrm02/Jupyter/IBudgetMETAS.ipynb | AndersThorseth/empir19nrm02 |
Prediction of 2017 World Series winner given data 1905 dataset. | import pandas as pd
tot1905 = pd.read_csv("../clean_data/1905ML.csv")
tot1905 = tot1905.drop({"Unnamed: 0", "H", "HR", "BB", "SB", "HA", "HRA", "BBA", "SOA", "E"}, axis=1)
tot1905
tot2017 = pd.read_csv("../clean_data/2017ML.csv")
tot2017 = tot2017.drop({"Unnamed: 0", "WSWIN"}, axis=1)
tot2017
# Create a function to con... |
WS Probability = 0.000567750444291
R 818
ERA 3.3
WP 0.63
Name: 7, dtype: object
WS Probability = 0.000386572498575
R 770
ERA 3.38
WP 0.642
Name: 13, dtype: object
WS Probability = 0.000337225795312
R 858
ERA 3.72
WP 0.562
Name: 18, dtype: object
WS Probability = 0.0002947... | MIT | world_series_prediction-master/ML/Logistic--2017.ipynb | kchhajed1/baseball_predictions |
Calculating the distance between the Customer's city and the Seller's city | from pyspark.sql import SparkSession, functions as F
import math
spark = SparkSession.builder.getOrCreate()
orders_items_df = spark.read \
.option('escape', '\"') \
.option('quote', '\"') \
.csv('./dataset/olist_order_items_dataset.csv', header=True, multiLine=True, in... | _____no_output_____ | MIT | hypothesis_25.ipynb | IuriSly/DnA-POC-olist |
Grouping data | data_df = orders_df.filter(F.col('order_status') == 'delivered').join(customers_df, 'customer_id')
data_df = orders_items_df.join(data_df, 'order_id') \
.join(sellers_df, 'seller_id') \
.select('customer_state', 'customer_city', 'customer_zip_code_prefix', 'seller_zip_... | +-------------------+-------------------+-------------------+-------------------+-------------+
| customer_lat| customer_lng| seller_lat| seller_lng|freight_value|
+-------------------+-------------------+-------------------+-------------------+-------------+
| -23.50648246805157|-47.4220680... | MIT | hypothesis_25.ipynb | IuriSly/DnA-POC-olist |
Calculating distance | def d(c_lat, c_lng, s_lat, s_lng):
radius = 6371 # km
dlat = math.radians(s_lat-c_lat)
dlon = math.radians(s_lng-c_lng)
a = math.sin(dlat/2) * math.sin(dlat/2) + math.cos(math.radians(c_lat)) \
* math.cos(math.radians(s_lat)) * math.sin(dlon/2) * math.sin(dlon/2)
c = 2 * math.atan2(math.sqr... | _____no_output_____ | MIT | hypothesis_25.ipynb | IuriSly/DnA-POC-olist |
SCOPUS journal Data analysis of Indian Research About: The main aim of this data analysis is to identify the ongoing research in Indian Universities and Indian Industry. It gives a basic answer about research source and trend with top authors and publication. It also shows the participation of Industry and Universitie... | import sqlite3
import matplotlib.pyplot as plt
import operator
sqlite_database = '/home/neel/scopus_data/scopus_data.sqlite'
conn = sqlite3.connect(sqlite_database)
c = conn.cursor()
c.execute("SELECT `Title`,`cited_rank` FROM `AI_scopus` ORDER BY `cited_rank` DESC LIMIT 0, 20;")
data = c.fetchall()
conn.close()
top_p... | _____no_output_____ | MIT | SCOPUS_data_analysis_Python3x_v1.ipynb | NeelShah18/scopus-analysis-for-indian-researcher |
[](https://github.com/lab-ml/python_autocomplete)[](https://colab.research.google.com/github/lab-ml/python_autocomplete/blob/master/notebooks/evaluate.ipynb) Ev... | %%capture
!pip install labml labml_python_autocomplete | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Imports | import string
import torch
from torch import nn
from labml import experiment, logger, lab
from labml_helpers.module import Module
from labml.logger import Text, Style
from labml.utils.pytorch import get_modules
from labml.utils.cache import cache
from labml_helpers.datasets.text import TextDataset
from python_autoco... | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
We load the model from a training run. For this demo I'm loading from a run I trained at home.[](https://web.lab-ml.com/run?uuid=39b03a1e454011ebbaff2b26e3148b3d)*If you want to try this on Colab you need to run this on the same space where you run t... | TRAINING_RUN_UUID = '39b03a1e454011ebbaff2b26e3148b3d' | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
We initialize `Configs` object defined in [`train.py`](https://github.com/lab-ml/python_autocomplete/blob/master/python_autocomplete/train.py). | conf = Configs() | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Create a new experiment in evaluation mode. In evaluation mode a new training run is not created. | experiment.evaluate() | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Load custom configurations/hyper-parameters used in the training run. | custom_conf = experiment.load_configs(TRAINING_RUN_UUID)
custom_conf | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Set the custom configurations | experiment.configs(conf, custom_conf) | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Set models for saving and loading. This will load `conf.model` from the specified run. | experiment.add_pytorch_models({'model': conf.model}) | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Specify which run to load from | experiment.load(TRAINING_RUN_UUID) | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Start the experiment | experiment.start() | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Initialize the `Predictor` defined in [`evaluate.py`](https://github.com/lab-ml/python_autocomplete/blob/master/python_autocomplete/evaluate.py).We load `stoi` and `itos` from cache, so that we don't have to read the dataset to generate them. `stoi` is the map for character to an integer index and `itos` is the map of ... | p = Predictor(conf.model, cache('stoi', lambda: conf.text.stoi), cache('itos', lambda: conf.text.itos)) | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Set model to evaluation mode | _ = conf.model.eval() | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
A python prompt to test completion. | PROMPT = """from torch import nn
from labml_helpers.module import Module
from labml_nn.lstm import LSTM
class LSTM(Module):
def __init__(self, *,
n_tokens: int,
embedding_size: int,
hidden_size int,
n_layers int):""" | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Get a token. `get_token` predicts character by character greedily (no beam search) until it find and end of token character (non alpha-numeric character). | %%time
res = p.get_token(PROMPT)
print('"' + res + '"') | "
super"
CPU times: user 950 ms, sys: 34.7 ms, total: 984 ms
Wall time: 254 ms
| MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Try another token | res = p.get_token(PROMPT + res)
print('"' + res + '"') | "(LSTM"
| MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Load a sample python file to test our model | with open(str(lab.get_data_path() / 'sample.py'), 'r') as f:
sample = f.read()
print(sample[-50:]) | ckpoint()
if __name__ == '__main__':
main()
| MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Test the model on a sample python file`evaluate` function defined in[`evaluate.py`](https://github.com/lab-ml/python_autocomplete/blob/master/python_autocomplete/evaluate.py)will predict token by token using the `Predictor`, and simulates an editor autocompletion.Colors:* yellow: the token predicted is wrong and the u... | %%time
evaluate(p, sample) | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
`accuracy` is the fraction of charactors predicted correctly. `key_strokes` is the number of key strokes required to write the code with help of the model and `length` is the number of characters in the code, that is the number of key strokes required without the model.*Note that this sample is a classic MNIST example,... | anomalies(p, sample) | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
Here we try to autocomplete 100 characters | sample = """import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import torch.utils.data
from torchvision import datasets, transforms
from labml import lab
class Model(nn.Module):
"""
complete(p, sample, 100) | _____no_output_____ | MIT | notebooks/evaluate.ipynb | kalufinnle/python_autocomplete |
MLP 208* Operate on 16000 GenCode 34 seqs.* 5-way cross validation. Save best model per CV.* Report mean accuracy from final re-validation with best 5.* Use Adam with a learn rate decay schdule. | NC_FILENAME='ncRNA.gc34.processed.fasta'
PC_FILENAME='pcRNA.gc34.processed.fasta'
DATAPATH=""
try:
from google.colab import drive
IN_COLAB = True
PATH='/content/drive/'
drive.mount(PATH)
DATAPATH=PATH+'My Drive/data/' # must end in "/"
NC_FILENAME = DATAPATH+NC_FILENAME
PC_FILENAME = DATAPA... | _____no_output_____ | MIT | Workshop/MLP_208.ipynb | ShepherdCode/ShepherdML |
Build model | def compile_model(model):
adam_default_learn_rate = 0.001
schedule = tf.keras.optimizers.schedules.ExponentialDecay(
initial_learning_rate = adam_default_learn_rate*10,
#decay_steps=100000, decay_rate=0.96, staircase=True)
decay_steps=10000, decay_rate=0.99, staircase=True)
# learn r... | _____no_output_____ | MIT | Workshop/MLP_208.ipynb | ShepherdCode/ShepherdML |
Load and partition sequences | # Assume file was preprocessed to contain one line per seq.
# Prefer Pandas dataframe but df does not support append.
# For conversion to tensor, must avoid python lists.
def load_fasta(filename,label):
DEFLINE='>'
labels=[]
seqs=[]
lens=[]
nums=[]
num=0
with open (filename,'r') as infile:
... | _____no_output_____ | MIT | Workshop/MLP_208.ipynb | ShepherdCode/ShepherdML |
Make K-mers | def make_kmer_table(K):
npad='N'*K
shorter_kmers=['']
for i in range(K):
longer_kmers=[]
for mer in shorter_kmers:
longer_kmers.append(mer+'A')
longer_kmers.append(mer+'C')
longer_kmers.append(mer+'G')
longer_kmers.append(mer+'T')
short... | _____no_output_____ | MIT | Workshop/MLP_208.ipynb | ShepherdCode/ShepherdML |
Cross validation | def do_cross_validation(X,y,given_model):
cv_scores = []
fold=0
splitter = ShuffleSplit(n_splits=SPLITS, test_size=0.1, random_state=37863)
for train_index,valid_index in splitter.split(X):
fold += 1
X_train=X[train_index] # use iloc[] for dataframe
y_train=y[train_index]
... | _____no_output_____ | MIT | Workshop/MLP_208.ipynb | ShepherdCode/ShepherdML |
Train on RNA lengths 200-1Kb | MINLEN=200
MAXLEN=1000
print("Load data from files.")
nc_seq=load_fasta(NC_FILENAME,0)
pc_seq=load_fasta(PC_FILENAME,1)
train_set=pd.concat((nc_seq,pc_seq),axis=0)
nc_seq=None
pc_seq=None
print("Ready: train_set")
#train_set
print ("Compile the model")
model=build_model(MAXLEN)
print ("Summarize the model")
print(model... | _____no_output_____ | MIT | Workshop/MLP_208.ipynb | ShepherdCode/ShepherdML |
JAXの乱数生成について調べてみたけどよくわからない> JAXにおける乱数生成について調べたけどよくわからない- toc: true - badges: true- comments: true- categories: [Python, JAX, DeepLearning]- image: images/jax-samune.png JAX流行ってますね。JAXについての詳しい説明は、[たくさんの記事](https://www.google.com/search?q=jax%E3%81%A8%E3%81%AF)や[https://github.com/google/jax](https://github.com/google/j... | import numpy as np
# numpy
x = np.random.rand()
print('x:', x)
for i in range(10):
x = np.random.rand()
print('x:', x) | x: 0.7742336894342167
x: 0.45615033221654855
x: 0.5684339488686485
x: 0.018789800436355142
x: 0.6176354970758771
x: 0.6120957227224214
x: 0.6169339968747569
x: 0.9437480785146242
x: 0.6818202991034834
x: 0.359507900573786
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
バラバラの結果が出てきました。これを固定するには、このようなコードを書きます。 | for i in range(10):
np.random.seed(0)
x = np.random.rand()
print('x:', x) | x: 0.5488135039273248
x: 0.5488135039273248
x: 0.5488135039273248
x: 0.5488135039273248
x: 0.5488135039273248
x: 0.5488135039273248
x: 0.5488135039273248
x: 0.5488135039273248
x: 0.5488135039273248
x: 0.5488135039273248
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
ところでnumpyでは、`np.random.get_state()` で乱数生成器の状態が確認できます。 | np.random.seed(0)
state = np.random.get_state()
print(state[0])
print('[', *state[1][:10], '...')
print(*state[1][-10:], ']')
np.random.seed(20040304)
state = np.random.get_state()
print(state[0])
print('[', *state[1][:10], '...')
print(*state[1][-10:], ']') | MT19937
[ 20040304 3876245041 2868517820 934780921 2883411521 496831348 4198668490 1502140500 1427494545 3747657433 ...
744972032 1872723303 3654422950 1926579586 2599193113 3757568530 3621035041 2338180567 2885432439 2647019928 ]
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
逆に言えば、Numpyの乱数生成はグローバルな一つの状態に依存しています。このことは次のような弊害を生みます。 並列実行と実行順序、再現性 簡単なゲームを作ってみます。 関数`a`, `b`が乱数を生成するので、大きい数を返した方が勝ちというゲームです。 | a = lambda : np.random.rand()
b = lambda : np.random.rand()
def battle():
if a() > b():
return 'A'
else:
return 'B'
for i in range(10):
print('winner is', battle(), '!') | winner is B !
winner is A !
winner is B !
winner is A !
winner is A !
winner is A !
winner is B !
winner is B !
winner is B !
winner is A !
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
また実行すれば、結果は変化します。 | for i in range(10):
print('winner is', battle(), '!') | winner is B !
winner is A !
winner is A !
winner is B !
winner is B !
winner is B !
winner is A !
winner is A !
winner is A !
winner is B !
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
ではこの結果の再現性を持たせるにはどうすればいいでしょうか。簡単な例はこうなります。 | res1 = []
np.random.seed(0)
for i in range(10):
res1.append(battle())
# もう一回
res2 = []
np.random.seed(0)
for i in range(10):
res2.append(battle())
print('1 | 2')
print('=====')
for i in range(10):
print(res1[i], '|', res2[i]) | 1 | 2
=====
B | B
A | A
B | B
B | B
A | A
A | A
B | B
B | B
B | B
B | B
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
というわけで同じ結果が得られました。しかし、この結果には落とし穴があります。関数`battle`の動作をもう少し詳しく確認してみましょう。`a`と`b`が呼び出されるタイミングを確認してみます。 | def a():
print('a is called!')
return np.random.rand()
def b():
print('b is called!')
return np.random.rand()
for i in range(5):
battle()
print('======') | a is called!
b is called!
======
a is called!
b is called!
======
a is called!
b is called!
======
a is called!
b is called!
======
a is called!
b is called!
======
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
このように、aはbより常に先に呼び出されます。ここまでだと何の問題もないように見えますが、実際にはそうではありません。このコードを高速に動作させたい、つまり並列化を行う時にはどうなるでしょうか。関数`a`, `b`に依存関係はありませんから、これらを並列に動作させても問題ないように感じます。ですが、実際には `a`, `b`が返す関数は呼び出し順序に依存しています！従って、このままではせっかく`np.random.seed`をしても意味がなくなってしまいます。 JAXの乱数生成では、JAXにおける乱数生成を確認してみます。先ほどまでで述べたように、次のような条件を満たす乱数生成器を実装したいです。- 再現性があること- 並列化でき... | key = jax.random.PRNGKey(0)
key | _____no_output_____ | Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
keyは単に二つの実数値からなるオブジェクトで、これを用いることによって、JAXでは乱数を生成します。 | jax.random.normal(key) | _____no_output_____ | Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
そして、keyが同じであれば同じ値が生成されます。 | print(key, jax.random.normal(key))
print(key, jax.random.normal(key))
print(key, jax.random.normal(key))
print(key, jax.random.normal(key))
print(key, jax.random.normal(key))
print(key, jax.random.normal(key))
print(key, jax.random.normal(key))
print(key, jax.random.normal(key))
print(key, jax.random.normal(key)) | [0 0] -0.20584235
[0 0] -0.20584235
[0 0] -0.20584235
[0 0] -0.20584235
[0 0] -0.20584235
[0 0] -0.20584235
[0 0] -0.20584235
[0 0] -0.20584235
[0 0] -0.20584235
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
とはいえこれだけだとひとつの数字しか得ることができません。もっとたくさんの乱数が欲しくなった際には、`jax.random.split`を用います。 | key1, key2 = jax.random.split(key)
print(key, '->', key1, key2) | [0 0] -> [4146024105 967050713] [2718843009 1272950319]
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
`jax.random.split`によって、ひとつのkeyから2つのkeyが作り出されます。このkeyによって、また新しい乱数を生み出します。 ちなみに、この二つのkeyは等価ですが、慣例的に二つ目を新しい乱数生成につかい、一つ目はまた新しいkeyを使うために用いられるようです。(以下のコードを参照) | # 慣例的に二つ目をsub_keyとして新しい乱数生成に、一つ目をまた新しい乱数を作るために使用する(下のように書くことでsplit元の古いkeyも削除できる。keyが残ると誤って同じ乱数を作ってしまうので注意が必要。)
key, sub_key = jax.random.split(key)
key, subsub_key = jax.random.split(key) | _____no_output_____ | Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
また、同じkeyから分割されたkeyは、常に等しくなります。 | def check_split(seed):
key = jax.random.PRNGKey(seed)
key, sub_key = jax.random.split(key)
print(key, '->', key, sub_key)
check_split(0)
check_split(0)
check_split(0)
print('=============================================================================')
check_split(2004)
check_split(2004)
check_split(2004) | [4146024105 967050713] -> [4146024105 967050713] [2718843009 1272950319]
[4146024105 967050713] -> [4146024105 967050713] [2718843009 1272950319]
[4146024105 967050713] -> [4146024105 967050713] [2718843009 1272950319]
=============================================================================
[2965909967 23466... | Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
また、一度に何個にもsplitできます。例えば1つのkeyから次のようにして10個のkeyを得ることができます。 | # 何個にもsplitできる。
key = jax.random.PRNGKey(0)
key, *sub_keys = jax.random.split(key, num=10)
key
sub_keys | _____no_output_____ | Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
sequential-equivalent Numpyではsequential-equivalentが保障されています。(適切な訳語がわからない)簡単にいうと、まとめてN個の乱数を取得することと、ひとつひとつ乱数を取得して連結したものは等価である、ということが保障されています。(以下のコードを見るとわかりやすいです) | # ひとつずつ
np.random.seed(0)
print(np.array([np.random.rand() for i in range(10)]))
print('================================================')
# まとめて
np.random.seed(0)
print(np.random.rand(10)) | [0.5488135 0.71518937 0.60276338 0.54488318 0.4236548 0.64589411
0.43758721 0.891773 0.96366276 0.38344152]
================================================
[0.5488135 0.71518937 0.60276338 0.54488318 0.4236548 0.64589411
0.43758721 0.891773 0.96366276 0.38344152]
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
ところがJAXではその限りではありません。JAXで10個の配列を取得する方法としては、- keyを10個用意する- ひとつのkeyから10個作るということが考えられます。 | # やり方 1: keyを10個用意
key = jax.random.PRNGKey(0)
key, *sub_keys = jax.random.split(key, 11)
print(np.array([jax.random.normal(sub_key) for sub_key in sub_keys]))
# やり方 2: ひとつのkeyから10個作る
key = jax.random.PRNGKey(0)
print(np.array(jax.random.normal(key, shape=(10,)))) | [-0.372111 0.2642311 -0.18252774 -0.7368198 -0.44030386 -0.15214427
-0.6713536 -0.59086424 0.73168874 0.56730247]
| Apache-2.0 | _notebooks/2021-11-14-jax-random.ipynb | abap34/my-website |
Lumped Elements Circuits In this notebook, we construct various network from basic lumped elements (resistor, capacitor, inductor), with the 'classic' and the `Circuit` approach. Generally the `Circuit` approach is more verbose than the 'classic' way for building a circuit. However, as the circuit complexity increases... | import numpy as np # for np.allclose() to check that S-params are similar
import skrf as rf
rf.stylely() | _____no_output_____ | BSD-3-Clause | doc/source/examples/circuit/Lumped Element Circuits.ipynb | nmaterise/scikit-rf |
LC Series Circuit In this section we reproduce a simple equivalent model of a capacitor $C$, as illustrated by the figure below: | # reference LC circuit made in Designer
LC_designer = rf.Network('designer_capacitor_30_80MHz_simple.s2p')
# scikit-rf: manually connecting networks
line = rf.media.DefinedGammaZ0(frequency=LC_designer.frequency, z0=50)
LC_manual = line.inductor(24e-9) ** line.capacitor(70e-12)
# scikit-rf: using Circuit builder
port1... | _____no_output_____ | BSD-3-Clause | doc/source/examples/circuit/Lumped Element Circuits.ipynb | nmaterise/scikit-rf |
A More Advanced Equivalent Model In this section we reproduce an equivalent model of a capacitor $C$, as illustrated by the figure below: | # Reference results from ANSYS Designer
LCC_designer = rf.Network('designer_capacitor_30_80MHz_adv.s2p')
# scikit-rf: usual way, but this time this is more tedious to deal with connection and port number
freq = LCC_designer.frequency
line = rf.media.DefinedGammaZ0(frequency=freq, z0=50)
elements1 = line.resistor(1e-2) ... | _____no_output_____ | BSD-3-Clause | doc/source/examples/circuit/Lumped Element Circuits.ipynb | nmaterise/scikit-rf |
Pass band filter Below we construct a pass-band filter, from an example given in [Microwaves101](https://www.microwaves101.com/encyclopedias/lumped-element-filter-calculator): | # Reference result calculated from Designer
passband_designer = rf.Network('designer_bandpass_filter_450_550MHz.s2p')
# scikit-rf: the filter by cascading all lumped-elements
freq = passband_designer.frequency
passband_manual = line.shunt_capacitor(25.406e-12) ** line.shunt_inductor(4.154e-9) ** \
li... | _____no_output_____ | BSD-3-Clause | doc/source/examples/circuit/Lumped Element Circuits.ipynb | nmaterise/scikit-rf |
Boosting: HyperparametersImport [`GradientBoostingClassifier`](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.GradientBoostingClassifier.html) and [`GradientBoostingRegressor`](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.GradientBoostingRegressor.html) from `sklearn` and expl... | from sklearn.ensemble import GradientBoostingClassifier, GradientBoostingRegressor
print(GradientBoostingClassifier())
print(GradientBoostingRegressor()) | GradientBoostingClassifier(criterion='friedman_mse', init=None,
learning_rate=0.1, loss='deviance', max_depth=3,
max_features=None, max_leaf_nodes=None,
min_impurity_decrease=0.0, min_impurity_split=None,
min_samples_leaf=1, min_samples_split=2,
min_... | MIT | ml_algorithms/06_Boosting/06_03/End/06_03.ipynb | joejoeyjoseph/playground |
Simulations: Dynamic learning with two learners, one oracle, and *heuristic evidence-weighting function* This notebook provides code to simulate 1D boundary learning in two agents learning from each other and from an "oracle" that always tells the truth. Each agent receives labels from the other agent (based on her cu... | get.pwt <- function(d, s=24, o=5, p=1){
#Computes the proportion of weight given to a source
#based on the distance between learner and source boundary
#
#d = vector of distances between learner and source boundary for n sources
#s = slope of HEW curve
#o = offsetof HEW curve---distances less th... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Here is what the weighting function looks like with default parameters from experiment 2, where s = 24 and o = 5: | plot(0:150, get.pwt(0:150, s=24, o=5), type = "l", lwd = 3, pch = 16, ylab = "Source weight", xlab = "Source distance", ylim = c(0,1))
| _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Other possible weighting functions Here we define some other weighting functions to investigate group learning dynamics under different hypotheses about weighting. Equal weight to both sources This function can be used in place of get.pwt to simulate learning where both sources always get equal weight. | get.samewt <-function(d, s=NA, o=NA, p=NA){
#Returns a vector of 0.5 for each element of d
#essentially always giving the same .5 weight to each source
#All parameters ignored except d, only included to work with other code
#############
out<-rep(.5, times=length(d))
out
} | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Rectified linear weighting This function returns the weight of a source based on a linear decline of the source's distance from the learner's curent boundary, rectified at 0 and 1. | get.rlwt <- function(d, s=0.01, o=4.5,p=0){
#Rectified linear weighting function
#d = vector of distances for sources to be weighted
#o = offset; distances less than this get weight 1
#s = slope, rate at which weight diminishes with distance
#p = proportion shrinkage from 1 and 0.
#Returns vecto... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Here is what the rectified weighting function looks like. $o$ shifts it left/right, $s$ changes the slope. | plot(0:300, get.rlwt(0:300, s=.005, o=4.5, p=.0), type="l", ylim = c(0,1),
ylab="Source weight", xlab="Source distance") | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Sigmoid This returns a source weight as the sigmoid of its distance from the learner's source. Like HEW and rectified linear, the function is bounded at ${0,1}$. | get.sigwt <- function(d, s=1, o=4.5,p=NA){
#Sigmoidal weighting function
#d = vector of distances for sources to be weighted
#o = offset, shifts sigmoid left/right
#s = slope of sigmoid
#p = for compatibility, not used
#Returns vector of weights, one for each element in d
#############
d... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Here is the plot: | plot(0:300, get.sigwt(0:300, s=.05, o=-3), type="l", ylim = c(0,1),
ylab="Source weight", xlab="Source distance") | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Update the learner's current boundary according to evidence-weighting function. Note that different results obtain depending on whether you use the curve to compute the close source weight first or the far source weight first. Both lead to stable states where learners disagree, but Experiment 2 shows that a source wit... | update.bound <- function(i, s1, s2, r = 1, weightfirst ="c", closebig=T, f = get.pwt, fpars=c(24, 5, 1)){
##############
#Updates learner's current boundary accourding to nonlinear weighting function
#
#i=learner's initial boundary
#s1, s2 = source 1 and 2 boundaries
#r = rate of boundary change... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Simulate two learners and one static source This simulation involves two learners and one oracle, as reported in the main paper. The following function generates the sequence of belief-states occupied by each learner over the course of learning, given their starting states, the ground truth, one of the evidence-weight... | dynamic.sim <- function(l1, l2, static, nsteps=100, r=1, f = get.pwt, fpars=c(25,5,1)){
#Simulates two learners, learning from each other and from one static source
#l1, l2, static = initial boundaries for learners 1 and 2 and static source
#nsteps = number of learning steps to simulate
#r = updating r... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Plot simulations for a grid of possible initial learning boundaries in the pair The code below runs the 2-learer simulation several times, with the two learners each beginning with a belief about the category boundary lying somewhere between 0 and 300. For each pair of initial beliefs, it computes how the beliefs cha... | gridpts <- 20 #Number of grid points in each dimension
gtruth <- 150 #Location of ground-truth boundary
niter <- 200 #Number of learning iterations
uprate <- 0.1 #Proportional rate at which beliefs are updated on each iteration
upfunc <- get.pwt #Function for computing source weights
upfunc.pars <- c(24,5,1) #Paramete... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
As you can see, when learner beliefs begin in the upper left or lower right quadrants, they converge on the truth. These are cases where each learner begins on a different side of the ground truth--so the oracle and the other source are always pulling in the same direction. The two learners do not perfectly agree until... | firstbound <- 150 #Expected initial boundary
newbound <- update.bound(150, 165, 50, fpars = c(25,4.5,1), weightfirst = "d") #new boundary after updating weight
bshift <- newbound - firstbound #expected amount of shift
dwt <- get.pwt(100) #expected weight given to far source
print(round(c(firstbound, newbound, bshift, d... | [1] 150.00 141.14 -8.86 0.20
| MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
So given an initial boundary at 150 and sources at 165 and 50, the weighting curve from Experiment 2 predicts a final boundary around 141, with a total shift of about 9 units toward the distal source. The observed shift wa 14 +/- 7.5 units toward the far source, a confidence interval that includes this prediction. The ... | tmp <- rep(0, times = 100) #vector of zeros to contain predictions for one model
#for(i1 in c(1:100)) tmp[i1] <- update.bound(100, 150-i1, 150+i1, fpars = c(24,5,1))
#Create empy plot
plot(0,0, type = "n", xaxt = "n", xlab = "Source location", pch = 16,
ylab = "Shift toward midpoint", ylim = c(-10,100), xlim = c... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
The result shows that quantitative predictions about the amount of shift vary quite a bit with the parameters of the trust weighting function, especially as the sources grow further toward the poles. But all parameterizations yield the same U-shape: when the sources are closer to the midpoint than is the learner's init... | plot(0,0,type = "n", xlim = c(0,150), ylim = c(0,1), xlab = "Distance of source", ylab = "Source weight")
for(i1 in c(100:5)) lines(0:150, get.pwt(0:150, i1, 5, 1), col = rainbow(100)[i1-15], lwd = 5)
lines(0:150, get.pwt(0:150, 24,5,1), col=1, lwd = 5) | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Effects of social connections amongst learning pairs In the above simulation of dynamic learners we considered pairs of learners with many different initial beliefs, each learning from the other and from a static oracle. What happens in a group of learners, where the two sources for any single learner are selected acc... | get.srcs <- function(l, s, p = "r", r=NA){
#Function to get two source distances for a learner based on a policy
#l = learner's current boundary
#s = current boundary for all sources
#p = policy for choosing 2 sources:
#r = random
#s = two most similar
#m = mixed ie most similar ... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Check to make sure the code works | get.srcs(5, 1:10, p="n", r=2) | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Functions to simulate a population The following code populates a matrix (out) in which columns indicate learners/sources and rows indicate learning epochs. For each epoch and learner, the code selects two sources according to some policy, as determined by the get.srcs function. The learner then updates her boundary a... | sim.pop <- function(l=c(1:10)*14, o=150, nsteps=300, rate=.1, policy="r", exrad=5){
init<-c(l, o) #Initial boundaries for learners and oracles
nl <- length(l) #number of learners
no <- length(o) #number of oracles
ns <- length(init) #total number of sources
out <- matrix(0, nsteps+1, ns) #Initialize... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
The following code plots the change in learner boundaries over time generated by the preceding code. | plot.popsim <- function(d, nl=10){
no <- dim(d)[2] - nl #Number of oracles
nsteps <- dim(d)[1] -1
#Plot initial boundaries and frame:
plot(rep(0, times = nl), d[1,1:nl], pch=16, col = 3, xlim = c(0,nsteps), ylim = c(0,300),
ylab = "Boundary", xlab = "Time")
#Add lines showing how each learn... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Below code runs simulation with two oracles at 150 and random initial beliefs sampled from 10-200. Change the policy parameter to one of the above to see the result of different source-selection policies. | n <- 10 #Number of simulated agents
no <- 2 #Number of oracles
gt <- 150 #ground truth provided by oracles
ibspan <- 100 #Maximum span of initial learner belief distribution
ibshift <- 140 #Shift from 0 of initial learner belief distribution
p <- "m" #Policy for selecting sources, one of:
#r = random s = two ... | _____no_output_____ | MIT | Models/AgentBasedModels.ipynb | ttrogers/frigo-chen-rogers |
Copyright 2019 The TensorFlow Hub Authors.Licensed under the Apache License, Version 2.0 (the "License"); | # Copyright 2019 The TensorFlow Hub Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by app... | _____no_output_____ | Apache-2.0 | site/ja/hub/tutorials/tf2_arbitrary_image_stylization.ipynb | phoenix-fork-tensorflow/docs-l10n |
任意画風の高速画風変換 TensorFlow.org で表示 Google Colab で実行 GitHub でソースを表示 ノートブックをダウンロード TF Hub モデルを見る [magenta](https://github.com/tensorflow/magenta/tree/master/magenta/models/arbitrary_image_stylization) と次の発表のモデルコードに基づきます。[Exploring the structure of a real-time, arbitrary neural artistic stylization netw... | import functools
import os
from matplotlib import gridspec
import matplotlib.pylab as plt
import numpy as np
import tensorflow as tf
import tensorflow_hub as hub
print("TF Version: ", tf.__version__)
print("TF-Hub version: ", hub.__version__)
print("Eager mode enabled: ", tf.executing_eagerly())
print("GPU available:... | _____no_output_____ | Apache-2.0 | site/ja/hub/tutorials/tf2_arbitrary_image_stylization.ipynb | phoenix-fork-tensorflow/docs-l10n |
使用する画像を取得しましょう。 | # @title Load example images { display-mode: "form" }
content_image_url = 'https://upload.wikimedia.org/wikipedia/commons/thumb/f/fd/Golden_Gate_Bridge_from_Battery_Spencer.jpg/640px-Golden_Gate_Bridge_from_Battery_Spencer.jpg' # @param {type:"string"}
style_image_url = 'https://upload.wikimedia.org/wikipedia/common... | _____no_output_____ | Apache-2.0 | site/ja/hub/tutorials/tf2_arbitrary_image_stylization.ipynb | phoenix-fork-tensorflow/docs-l10n |
TF-Hub モジュールをインポートする | # Load TF-Hub module.
hub_handle = 'https://tfhub.dev/google/magenta/arbitrary-image-stylization-v1-256/2'
hub_module = hub.load(hub_handle) | _____no_output_____ | Apache-2.0 | site/ja/hub/tutorials/tf2_arbitrary_image_stylization.ipynb | phoenix-fork-tensorflow/docs-l10n |
画風に使用する Hub モジュールのシグネチャは、次のとおりです。```outputs = hub_module(content_image, style_image) stylized_image = outputs[0]```上記の `content_image`、`style_image`、および `stylized_image` は、形状 `[batch_size, image_height, image_width, 3]` の 4-D テンソルです。現在の例では 1 つの画像のみを提供するためバッチの次元は 1 ですが、同じモジュールを使用して、同時に複数の画像を処理することができます。画像の入力と出力の値範囲は [0,... | # Stylize content image with given style image.
# This is pretty fast within a few milliseconds on a GPU.
outputs = hub_module(tf.constant(content_image), tf.constant(style_image))
stylized_image = outputs[0]
# Visualize input images and the generated stylized image.
show_n([content_image, style_image, stylized_image... | _____no_output_____ | Apache-2.0 | site/ja/hub/tutorials/tf2_arbitrary_image_stylization.ipynb | phoenix-fork-tensorflow/docs-l10n |
複数の画像で試してみる | # @title To Run: Load more images { display-mode: "form" }
content_urls = dict(
sea_turtle='https://upload.wikimedia.org/wikipedia/commons/d/d7/Green_Sea_Turtle_grazing_seagrass.jpg',
tuebingen='https://upload.wikimedia.org/wikipedia/commons/0/00/Tuebingen_Neckarfront.jpg',
grace_hopper='https://storage.googleap... | _____no_output_____ | Apache-2.0 | site/ja/hub/tutorials/tf2_arbitrary_image_stylization.ipynb | phoenix-fork-tensorflow/docs-l10n |
Data seems imbalance hence need to be balanced basis resampling techniques. | def date_q(date):
"""
Convert Date to Quarter when separated with /
"""
qdate = date.strip().split('/')[1:]
qdate1 = qdate[0]
if qdate1 in ['01','02','03']:
return (str('Q1' + '-' + qdate[1]))
if qdate1 in ['04','05','06']:
return (str('Q2' + '-' + qdate[1]))
if qdate1 i... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
EDA for merged file | campaign_data_DATE = campaign_data.copy()
campaign_data_DATE.head()
campaign_data_DATE['start_date_q'] = campaign_data_DATE['start_date'].map(lambda x: date_q(x))
campaign_data_DATE['end_date_q'] = campaign_data_DATE['end_date'].map(lambda x: date_q(x))
campaign_data_DATE.head()
campaign_data_DATE.drop(['start_date',... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Lets have a look for Customer Id's in terms of using coupons most and least number of times. | '''
Customer ids using coupons at least once with their demographic details
'''
a = pd.DataFrame(train_data_merge_EDA[(train_data_merge_EDA['redemption_status']==1)])
b = pd.DataFrame(a.groupby('customer_id')['redemption_status'].sum()).reset_index()
b.columns = ['customer_id','redeem_count']
b.sort_values(by='redeem_... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Hypothesis to test:1. Is age_range of 36 to 55 is mostly using coupons???2. Is Marital Status as Married are to redeem coupon???3. Is couple (family size of 2) are using coupons mostly???4. Is people not on rent are mostly using coupons???5. Is no_of_children irrelevant to redeem coupon???6. Is income bracket of 5 are ... | '''
1. Is age_range of 36 to 55 is mostly using coupons???
'''
d = pd.DataFrame(train_data_merge_EDA.groupby(['age_range'])['redemption_status'].sum()).reset_index()
d.columns = ['age_range','tot_redeem']
d['percent'] = round(d['tot_redeem']/(d['tot_redeem'].sum())*100,2)
display (d)
%matplotlib notebook
train_data_mer... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference : Out of the age demographic data available, age range of 36 to 55 are mostly redeeming coupons. | '''
2. Is Marital Status as Married are to redeem coupon???
'''
e = pd.DataFrame(train_data_merge_EDA.groupby(['marital_status'])['redemption_status'].sum()).reset_index()
e.columns = ['marital_status','tot_redeem']
e['percent'] = round(e['tot_redeem']/(e['tot_redeem'].sum())*100,2)
display (e)
%matplotlib notebook
tr... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference : As most of the customers haven't mentioned their marital status it is difficult to support that Married customer redeem more. But basis available data, we could see Married people are mostly using coupons to redeem. So at this stage I could think of giving more weightage to a person which has discloses mari... | '''
3. Is couple (family size of 2) are using coupons mostly???
'''
f = pd.DataFrame(train_data_merge_EDA.groupby(['family_size'])['redemption_status'].sum()).reset_index()
f.columns = ['family_size','tot_redeem']
f['percent'] = round(f['tot_redeem']/(f['tot_redeem'].sum())*100,2)
display (f)
%matplotlib notebook
trai... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference : Considering family size of 2 or more as a married couple (and transforming that ratio on to Marital Status data of Unspecified), then we could make an assumption over here that Married couple are mostly using the coupons to redeem. | '''
4. Is people not on rent are mostly using coupons???
'''
g = pd.DataFrame(train_data_merge_EDA.groupby(['rented'])['redemption_status'].sum()).reset_index()
g.columns = ['rented','tot_redeem']
g['percent'] = round(g['tot_redeem']/(g['tot_redeem'].sum())*100,2)
display (g)
%matplotlib notebook
train_data_merge_EDA.... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference : Maximum people have provided status as not rented so we could assume here more weightage to such customers as they have shown greater tendency towards redemption of the coupons. | '''
5. Is no_of_children irrelevant to redeem coupon???
'''
h = pd.DataFrame(train_data_merge_EDA.groupby(['no_of_children'])['redemption_status'].sum()).reset_index()
h.columns = ['no_of_children','tot_redeem']
h['percent'] = round(h['tot_redeem']/(h['tot_redeem'].sum())*100,2)
display (h)
%matplotlib notebook
train_... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference : Since most of the customers preferred not to disclose on number of children, at this point we can assume that this field has no significance with redemption of coupons. | '''
6. Is income bracket of 5 are using coupons mostly???
'''
j = pd.DataFrame(train_data_merge_EDA.groupby(['income_bracket'])['redemption_status'].sum()).reset_index()
j.columns = ['income_bracket','tot_redeem']
j['percent'] = round(j['tot_redeem']/(j['tot_redeem'].sum())*100,2)
display (j)
%matplotlib notebook
trai... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference : Assuming 5 as mid income group, customers in this group clearly shows behaviour towards redemption of coupon. Let's explore from Coupon's perspective, in terms of most redeemed and attributes associated with coupons | '''
Coupon ids getting redeemed very often and attributes associated with it
'''
k = pd.DataFrame(a.groupby('coupon_id')['redemption_status'].sum()).reset_index()
k.columns = ['coupon_id','redeem_count']
k.sort_values(by='redeem_count',ascending=False,inplace=True)
print ('Top 5 Coupon ids redeemed')
display (k.head()... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Basis visualization for top 5 coupons redeemed, we could look for trend towards1. brand 56 is the top selling, let's verify if specific brand shows tendency towards coupon redemption???2. Verify brand type shows tendency towards coupon redemption???3. Verify category shows tendency towards coupon redemption???4. Verify... | '''
1. verify if specific brand shows tendency towards coupon redemption???
'''
m = pd.DataFrame(l1[(l1['brand']==56)|(l1['brand']==133)|(l1['brand']==1337)|(l1['brand']==544)|(l1['brand']==681)][['brand','brand_type','category']])
m.drop_duplicates(subset=['brand','brand_type','category'], keep='first', inplace=True)
... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference: For brand 56, there is a wide range of category available under the umbrella of Food products and displays greater tendency towards coupon redemption as well. Even rest other top 4 brands as well belong to general food product category under Grocery. | '''
2. Verify brand type shows tendency towards coupon redemption???
'''
m1 = pd.DataFrame(train_data_merge_EDA.groupby(['brand_type'])['redemption_status'].sum()).reset_index()
m1.columns = ['brand_type','tot_redeem']
m1['percent'] = round(m1['tot_redeem']/(m1['tot_redeem'].sum())*100,2)
display (m1)
%matplotlib note... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference: Coupon redemption percentage seems high when associated with an Established brand type. | '''
3. Verify category shows tendency towards coupon redemption???
'''
m2 = pd.DataFrame(train_data_merge_EDA.groupby(['category'])['redemption_status'].sum()).reset_index()
m2.columns = ['category','tot_redeem']
m2['percent'] = round(m2['tot_redeem']/(m2['tot_redeem'].sum())*100,2)
m2.sort_values(by='tot_redeem',asce... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference : Grocery, Packaged Meat, Pharmaceutical, Natural Prodcut, Dairy & Juice and Meat are the category seems more associated with coupon redemption. | '''
4. Verify campaign type shows tendency towards coupon redemption???
'''
m3 = pd.DataFrame(train_data_merge_EDA.groupby(['campaign_type'])['redemption_status'].sum()).reset_index()
m3.columns = ['campaign_type','tot_redeem']
m3['percent'] = round(m3['tot_redeem']/(m3['tot_redeem'].sum())*100,2)
display (m3)
%matplo... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference : Campaign type of X seems to have more associated with coupon redemption. Let's explore coupon redemption trend basis campaign start date and transaction date | n = pd.DataFrame(train_data_merge_EDA.groupby(['start_date_q','end_date_q'])['redemption_status'].sum()).reset_index()
n.columns = ['start_date_q','end_date_q','tot_redeem']
n['percent'] = round(n['tot_redeem']/(n['tot_redeem'].sum())*100,2)
n.sort_values(by='tot_redeem',ascending=False,inplace=True)
display (n)
%matpl... | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Inference : Campaign started between Q2-13 to Q3-13 foolwed by Q1-13 to Q2-13, seems experience more association towards coupon redemption trend. So we could assume over here campaign spanned across early quarters of year seem to more association with coupon redemption. | n1 = pd.DataFrame(train_data_merge_EDA.groupby(['tran_date_q'])['redemption_status'].sum()).reset_index()
n1.columns = ['tran_date_q','tot_redeem']
n1['percent'] = round(n1['tot_redeem']/(n1['tot_redeem'].sum())*100,2)
n1.sort_values(by='tot_redeem',ascending=False,inplace=True)
display (n1) | _____no_output_____ | MIT | Hackathon_ML Sample Problems/Amex Dataset/ML on Amex Dataset_EDA.ipynb | girishvankudre/hackathon_ml_sample |
Copyright 2019 The TensorFlow Authors. | #@title Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under... | _____no_output_____ | MIT | Informatics/Deep Learning/TensorFlow - deeplearning.ai/3. NLP/Course_3_Week_1_Lesson_1.ipynb | MarcosSalib/Cocktail_MOOC |
Mask R-CNN - Train on Custom DatasetThis notebook shows how to train Mask R-CNN on your own dataset. You'd still need a GPU, though, because the network backbone is a Resnet101, which would be too slow to train on a CPU. On a GPU, you can start to get okay-ish results in a few minutes, and good results in less than an... | import os
import sys
import random
import math
import re
import time
import numpy as np
import cv2
import matplotlib
import matplotlib.pyplot as plt
from config import Config
import utils
import model as modellib
import visualize
from model import log
%matplotlib inline
# Root directory of the project
ROOT_DIR = os... | c:\users\yaroslav_strontsitsk\appdata\local\continuum\anaconda3\envs\maskrcnn\lib\site-packages\h5py\__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
from ._conv import reg... | MIT | train_t_shirts.ipynb | lumstery/maskrcnn |
Configurations | class ShapesConfig(Config):
"""Configuration for training on the toy shapes dataset.
Derives from the base Config class and overrides values specific
to the toy shapes dataset.
"""
# Give the configuration a recognizable name
NAME = "clothes"
# Train on 1 GPU and 8 images per GPU. We can pu... |
Configurations:
BACKBONE_SHAPES [[32 32]
[16 16]
[ 8 8]
[ 4 4]
[ 2 2]]
BACKBONE_STRIDES [4, 8, 16, 32, 64]
BATCH_SIZE 8
BBOX_STD_DEV [0.1 0.1 0.2 0.2]
DETECTION_MAX_INSTANCES 100
DETECTION_MIN_CONFIDENCE 0.7
DETECTION_NMS_THRESHOLD ... | MIT | train_t_shirts.ipynb | lumstery/maskrcnn |
Notebook Preferences | def get_ax(rows=1, cols=1, size=8):
"""Return a Matplotlib Axes array to be used in
all visualizations in the notebook. Provide a
central point to control graph sizes.
Change the default size attribute to control the size
of rendered images
"""
_, ax = plt.subplots(rows, cols, figsize=(... | _____no_output_____ | MIT | train_t_shirts.ipynb | lumstery/maskrcnn |
DatasetLoad a datasetExtend the Dataset class and add a method to load the shapes dataset, `load_images()`, and override the following methods:* load_image()* load_mask()* image_reference() | class ShapesDataset(utils.Dataset):
def load_images(self, count, prefix):
"""Load the requested number of images.
count: number of images to load.
"""
# Add classes
self.add_class("clothes", 1, "t-shirt")
# Add images
for i in range(count):
... | _____no_output_____ | MIT | train_t_shirts.ipynb | lumstery/maskrcnn |
Ceate Model | # Create model in training mode
model = modellib.MaskRCNN(mode="training", config=config,
model_dir=MODEL_DIR)
# Which weights to start with?
init_with = "coco" # imagenet, coco, or last
if init_with == "imagenet":
model.load_weights(model.get_imagenet_weights(), by_name=True)
elif init_... | _____no_output_____ | MIT | train_t_shirts.ipynb | lumstery/maskrcnn |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.