Detecção de anomalias com embeddings

Ver em ai.google.dev Testar um bloco do Colab Veja o notebook no GitHub

Visão geral

Neste tutorial, demonstramos como usar os embeddings da API Gemini para detectar possíveis outliers no seu conjunto de dados. Você vai visualizar um subconjunto do conjunto de dados dos 20 Newsgroup usando t-SNE e detectar outliers fora de um raio específico do ponto central de cada cluster categórico.

Para mais informações sobre como começar a usar embeddings gerados pela API Gemini, confira o guia de início rápido do Python.

Pré-requisitos

É possível executar este guia de início rápido no Google Colab.

Para concluir este guia de início rápido no seu ambiente de desenvolvimento, verifique se ele atende aos seguintes requisitos:

  • Python 3.9 ou superior
  • Uma instalação de jupyter para executar o notebook.

Configuração

Primeiro, baixe e instale a biblioteca Python da API Gemini.

pip install -U -q google.generativeai
import re
import tqdm
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

import google.generativeai as genai

# Used to securely store your API key
from google.colab import userdata

from sklearn.datasets import fetch_20newsgroups
from sklearn.manifold import TSNE

Obter uma chave de API

Antes de usar a API Gemini, você precisa de uma chave de API. Se você ainda não tiver uma, crie uma chave com um clique no Google AI Studio.

Gerar uma chave de API

No Colab, adicione a chave ao gerenciador de secrets no "HELP" no painel esquerdo. Nomeie como API_KEY.

Quando você tiver a chave de API, transmita-a ao SDK. Faça isso de duas maneiras:

  • Coloque a chave na variável de ambiente GOOGLE_API_KEY. O SDK vai selecioná-la automaticamente de lá.
  • Transmita a chave para genai.configure(api_key=...)
.
genai.configure(api_key=GOOGLE_API_KEY)
for m in genai.list_models():
  if 'embedContent' in m.supported_generation_methods:
    print(m.name)
models/embedding-001
models/embedding-001

Preparar conjunto de dados

O conjunto de dados de texto dos 20 grupos de notícias contém 18.000 postagens em grupos de notícias sobre 20 tópicos divididos em conjuntos de treinamento e teste. A divisão entre os conjuntos de dados de treinamento e teste é baseada em mensagens postadas antes e depois de uma data específica. Neste tutorial, usamos o subconjunto de treinamento.

newsgroups_train = fetch_20newsgroups(subset='train')

# View list of class names for dataset
newsgroups_train.target_names
['alt.atheism',
 'comp.graphics',
 'comp.os.ms-windows.misc',
 'comp.sys.ibm.pc.hardware',
 'comp.sys.mac.hardware',
 'comp.windows.x',
 'misc.forsale',
 'rec.autos',
 'rec.motorcycles',
 'rec.sport.baseball',
 'rec.sport.hockey',
 'sci.crypt',
 'sci.electronics',
 'sci.med',
 'sci.space',
 'soc.religion.christian',
 'talk.politics.guns',
 'talk.politics.mideast',
 'talk.politics.misc',
 'talk.religion.misc']

Aqui está o primeiro exemplo no conjunto de treinamento.

idx = newsgroups_train.data[0].index('Lines')
print(newsgroups_train.data[0][idx:])
Lines: 15

 I was wondering if anyone out there could enlighten me on this car I saw
the other day. It was a 2-door sports car, looked to be from the late 60s/
early 70s. It was called a Bricklin. The doors were really small. In addition,
the front bumper was separate from the rest of the body. This is 
all I know. If anyone can tellme a model name, engine specs, years
of production, where this car is made, history, or whatever info you
have on this funky looking car, please e-mail.

Thanks,

- IL
   ---- brought to you by your neighborhood Lerxst ----
# Apply functions to remove names, emails, and extraneous words from data points in newsgroups.data
newsgroups_train.data = [re.sub(r'[\w\.-]+@[\w\.-]+', '', d) for d in newsgroups_train.data] # Remove email
newsgroups_train.data = [re.sub(r"\([^()]*\)", "", d) for d in newsgroups_train.data] # Remove names
newsgroups_train.data = [d.replace("From: ", "") for d in newsgroups_train.data] # Remove "From: "
newsgroups_train.data = [d.replace("\nSubject: ", "") for d in newsgroups_train.data] # Remove "\nSubject: "

# Cut off each text entry after 5,000 characters
newsgroups_train.data = [d[0:5000] if len(d) > 5000 else d for d in newsgroups_train.data]
# Put training points into a dataframe
df_train = pd.DataFrame(newsgroups_train.data, columns=['Text'])
df_train['Label'] = newsgroups_train.target
# Match label to target name index
df_train['Class Name'] = df_train['Label'].map(newsgroups_train.target_names.__getitem__)

df_train

Em seguida, faça uma amostra de alguns dados pegando 150 pontos de dados no conjunto de dados de treinamento e escolhendo algumas categorias. Este tutorial usa as categorias de ciências.

# Take a sample of each label category from df_train
SAMPLE_SIZE = 150
df_train = (df_train.groupby('Label', as_index = False)
                    .apply(lambda x: x.sample(SAMPLE_SIZE))
                    .reset_index(drop=True))

# Choose categories about science
df_train = df_train[df_train['Class Name'].str.contains('sci')]

# Reset the index
df_train = df_train.reset_index()
df_train
df_train['Class Name'].value_counts()
sci.crypt          150
sci.electronics    150
sci.med            150
sci.space          150
Name: Class Name, dtype: int64

Criar os embeddings

Nesta seção, você vai aprender a gerar embeddings para os diferentes textos no DataFrame usando os embeddings da API Gemini.

Mudanças de API em embeddings com modelo embedding-001

Para o novo modelo de embeddings, embedding-001, há um novo parâmetro de tipo de tarefa e o título opcional (válido somente com task_type=RETRIEVAL_DOCUMENT).

Esses novos parâmetros se aplicam apenas aos modelos de embedding mais recentes.Os tipos de tarefa são:

Tipo de tarefa Descrição
RETRIEVAL_QUERY Especifica que o texto é uma consulta em uma configuração de pesquisa/recuperação.
RETRIEVAL_DOCUMENT Especifica que o texto é um documento em uma configuração de pesquisa/recuperação.
SEMANTIC_SIMILARITY Especifica o texto a ser usado para similaridade textual semântica (STS).
CLASSIFICAÇÃO Especifica que os embeddings serão usados para classificação.
CLUSTERING Especifica que os embeddings serão usados para clustering.
from tqdm.auto import tqdm
tqdm.pandas()

from google.api_core import retry

def make_embed_text_fn(model):

  @retry.Retry(timeout=300.0)
  def embed_fn(text: str) -> list[float]:
    # Set the task_type to CLUSTERING.
    embedding = genai.embed_content(model=model,
                                    content=text,
                                    task_type="clustering")['embedding']
    return np.array(embedding)

  return embed_fn

def create_embeddings(df):
  model = 'models/embedding-001'
  df['Embeddings'] = df['Text'].progress_apply(make_embed_text_fn(model))
  return df

df_train = create_embeddings(df_train)
df_train.drop('index', axis=1, inplace=True)
0%|          | 0/600 [00:00<?, ?it/s]

Redução de dimensionalidade

A dimensão do vetor de embedding de documentos é 768. Para visualizar como os documentos incorporados são agrupados, aplique a redução de dimensionalidade, já que só é possível visualizar os embeddings em espaços 2D ou 3D. Os documentos contextualmente semelhantes devem estar mais próximos no espaço, ao contrário dos documentos que não são tão semelhantes.

len(df_train['Embeddings'][0])
768
# Convert df_train['Embeddings'] Pandas series to a np.array of float32
X = np.array(df_train['Embeddings'].to_list(), dtype=np.float32)
X.shape
(600, 768)

Você vai aplicar a abordagem t-Distributed Stochastic Neighbor Embedding (t-SNE) para reduzir a dimensionalidade. Essa técnica reduz o número de dimensões e preserva os clusters (pontos que estão próximos permanecem próximos). Para os dados originais, o modelo tenta construir uma distribuição sobre a qual outros pontos de dados são "vizinhos" Por exemplo, eles compartilham um significado semelhante. Em seguida, ele otimiza uma função objetiva para manter uma distribuição semelhante na visualização.

tsne = TSNE(random_state=0, n_iter=1000)
tsne_results = tsne.fit_transform(X)
df_tsne = pd.DataFrame(tsne_results, columns=['TSNE1', 'TSNE2'])
df_tsne['Class Name'] = df_train['Class Name'] # Add labels column from df_train to df_tsne
df_tsne
fig, ax = plt.subplots(figsize=(8,6)) # Set figsize
sns.set_style('darkgrid', {"grid.color": ".6", "grid.linestyle": ":"})
sns.scatterplot(data=df_tsne, x='TSNE1', y='TSNE2', hue='Class Name', palette='Set2')
sns.move_legend(ax, "upper left", bbox_to_anchor=(1, 1))
plt.title('Scatter plot of news using t-SNE')
plt.xlabel('TSNE1')
plt.ylabel('TSNE2');

png

Detecção de outlier

Para determinar quais pontos são anômalos, determine quais são inliers e outliers. Comece encontrando o centroide, ou local que representa o centro do cluster, e use a distância para determinar os pontos que são outliers.

Comece obtendo o centroide de cada categoria.

def get_centroids(df_tsne):
  # Get the centroid of each cluster
  centroids = df_tsne.groupby('Class Name').mean()
  return centroids

centroids = get_centroids(df_tsne)
centroids
def get_embedding_centroids(df):
  emb_centroids = dict()
  grouped = df.groupby('Class Name')
  for c in grouped.groups:
    sub_df = grouped.get_group(c)
    # Get the centroid value of dimension 768
    emb_centroids[c] = np.mean(sub_df['Embeddings'], axis=0)

  return emb_centroids
emb_c = get_embedding_centroids(df_train)

Representar cada centroide encontrado em relação ao resto dos pontos.

# Plot the centroids against the cluster
fig, ax = plt.subplots(figsize=(8,6)) # Set figsize
sns.set_style('darkgrid', {"grid.color": ".6", "grid.linestyle": ":"})
sns.scatterplot(data=df_tsne, x='TSNE1', y='TSNE2', hue='Class Name', palette='Set2');
sns.scatterplot(data=centroids, x='TSNE1', y='TSNE2', color="black", marker='X', s=100, label='Centroids')
sns.move_legend(ax, "upper left", bbox_to_anchor=(1, 1))
plt.title('Scatter plot of news using t-SNE with centroids')
plt.xlabel('TSNE1')
plt.ylabel('TSNE2');

png

Escolha um raio. Qualquer coisa além desse limite do centroide dessa categoria é considerada uma outlier.

def calculate_euclidean_distance(p1, p2):
  return np.sqrt(np.sum(np.square(p1 - p2)))

def detect_outlier(df, emb_centroids, radius):
  for idx, row in df.iterrows():
    class_name = row['Class Name'] # Get class name of row
    # Compare centroid distances
    dist = calculate_euclidean_distance(row['Embeddings'],
                                        emb_centroids[class_name])
    df.at[idx, 'Outlier'] = dist > radius

  return len(df[df['Outlier'] == True])
range_ = np.arange(0.3, 0.75, 0.02).round(decimals=2).tolist()
num_outliers = []
for i in range_:
  num_outliers.append(detect_outlier(df_train, emb_c, i))
# Plot range_ and num_outliers
fig = plt.figure(figsize = (14, 8))
plt.rcParams.update({'font.size': 12})
plt.bar(list(map(str, range_)), num_outliers)
plt.title("Number of outliers vs. distance of points from centroid")
plt.xlabel("Distance")
plt.ylabel("Number of outliers")
for i in range(len(range_)):
  plt.text(i, num_outliers[i], num_outliers[i], ha = 'center')

plt.show()

png

Dependendo da sensibilidade do detector de anomalias, você pode escolher o raio. Por enquanto, é usado 0,62, mas é possível alterar esse valor.

# View the points that are outliers
RADIUS = 0.62
detect_outlier(df_train, emb_c, RADIUS)
df_outliers = df_train[df_train['Outlier'] == True]
df_outliers.head()
# Use the index to map the outlier points back to the projected TSNE points
outliers_projected = df_tsne.loc[df_outliers['Outlier'].index]

Traçar os pontos fora da curva e denotá-los usando uma cor vermelha transparente.

fig, ax = plt.subplots(figsize=(8,6)) # Set figsize
plt.rcParams.update({'font.size': 10})
sns.set_style('darkgrid', {"grid.color": ".6", "grid.linestyle": ":"})
sns.scatterplot(data=df_tsne, x='TSNE1', y='TSNE2', hue='Class Name', palette='Set2');
sns.scatterplot(data=centroids, x='TSNE1', y='TSNE2', color="black", marker='X', s=100, label='Centroids')
# Draw a red circle around the outliers
sns.scatterplot(data=outliers_projected, x='TSNE1', y='TSNE2', color='red', marker='o', alpha=0.5, s=90, label='Outliers')
sns.move_legend(ax, "upper left", bbox_to_anchor=(1, 1))
plt.title('Scatter plot of news with outliers projected with t-SNE')
plt.xlabel('TSNE1')
plt.ylabel('TSNE2');

png

Use os valores de índice das famílias de dados para exibir alguns exemplos de valores atípicos em cada categoria. Aqui, o primeiro ponto de dados de cada categoria é impresso. Explore outros pontos em cada categoria para ver os dados considerados outliers ou anomalias.

sci_crypt_outliers = df_outliers[df_outliers['Class Name'] == 'sci.crypt']
print(sci_crypt_outliers['Text'].iloc[0])
Re: Source of random bits on a Unix workstation
Lines: 44
Nntp-Posting-Host: sandstorm

>>For your application, what you can do is to encrypt the real-time clock
>>value with a secret key.

Well, almost.... If I only had to solve the problem for myself, and were
willing to have to type in a second password  whenever I
logged in, it could work. However, I'm trying to create a solution that
anyone can use, and which, once installed, is just as effortless to start up
as the non-solution of just using xhost to control access. I've got
religeous problems with storing secret keys on multiuser computers.

>For a good discussion of cryptographically "good" random number
>generators, check out the draft-ietf-security-randomness-00.txt
>Internet Draft, available at your local friendly internet drafts
>repository.

Thanks for the pointer! It was good reading, and I liked the idea of using
several unrelated sources with a strong mixing function. However, unless I
missed something, the only source they suggested 
that seems available, and unguessable by an intruder, when a Unix is
fresh-booted, is I/O buffers related to network traffic. I believe my
solution basically uses that strategy, without requiring me to reach into
the kernel.

>A reasonably source of randomness is the output of a cryptographic
>hash function , when fed with a large amount of
>more-or-less random data. For example, running MD5 on /dev/mem is a
>slow, but random enough, source of random bits; there are bound to be
>128 bits of entropy in the tens  of megabytes of data in
>a modern workstation's memory, as a fair amount of them are system
>timers, i/o buffers, etc.

I heard about this solution, and it sounded good. Then I heard that folks
were experiencing times of 30-60 seconds to run this, on
reasonably-configured workstations. I'm not willing to add that much delay
to someone's login process. My approach  takes
a second or two to run. I'm considering writing the be-all and end-all of
solutions, that launches the MD5, and simultaneously tries to suck bits off
the net, and if the net should be sitting __SO__ idle that it can't get 10K
after compression before MD5 finishes, use the MD5. This way I could have
guaranteed good bits, and a deterministic upper bound on login time, and
still have the common case of login take only a couple of extra seconds.

-Bennett
sci_elec_outliers = df_outliers[df_outliers['Class Name'] == 'sci.electronics']
print(sci_elec_outliers['Text'].iloc[0])
Re: Laser vs Bubblejet?
Reply-To: 
Distribution: world
X-Mailer: cppnews \\(Revision: 1.20 \\)
Organization: null
Lines: 53

Here is a different viewpoint.

> FYI:  The actual horizontal dot placement resoution of an HP
> deskjet is 1/600th inch.  The electronics and dynamics of the ink
> cartridge, however, limit you to generating dots at 300 per inch.
> On almost any paper, the ink wicks more than 1/300th inch anyway.
> 
> The method of depositing and fusing toner of a laster printer
> results in much less spread than ink drop technology.

In practice there is little difference in quality but more care is needed 
with inkjet because smudges etc. can happen.

> It doesn't take much investigation to see that the mechanical and
> electronic complement of a laser printer is more complex than
> inexpensive ink jet printers.  Recall also that laser printers
> offer a much higher throughput:  10 ppm for a laser versus about 1
> ppm for an ink jet printer.

A cheap laser printer does not manage that sort of throughput and on top of 
that how long does the _first_ sheet take to print? Inkjets are faster than 
you say and in both cases the computer often has trouble keeping up with the 
printer. 

A sage said to me: "Do you want one copy or lots of copies?", "One", 
"Inkjet".
 
> Something else to think about is the cost of consumables over the
> life of the printer.  A 3000 page yield toner cartridge is about
> $US 75-80 at discount while HP high capacity 
> cartridges are about $US 22 at discount.  It could be that over the
> life cycle of the printer that consumables for laser printers are
> less than ink jet printers.  It is getting progressively closer
> between the two technologies.  Laser printers are usually desinged
> for higher duty cycles in pages per month and longer product
> replacement cycles.

Paper cost is the same and both can use refills. Long term the laserprinter 
will need some expensive replacement parts  and on top of that 
are the amortisation costs which favour the lowest purchase cost printer.

HP inkjets understand PCL so in many cases a laserjet driver will work if the 
software package has no inkjet driver. 

There is one wild difference between the two printers: a laserprinter is a 
page printer whilst an inkjet is a line printer. This means that a 
laserprinter can rotate graphic images whilst an inkjet cannot. Few drivers 
actually use this facility.


  TC. 
    E-mail:  or
sci_med_outliers = df_outliers[df_outliers['Class Name'] == 'sci.med']
print(sci_med_outliers['Text'].iloc[0])
Re: THE BACK MACHINE - Update
Organization: University of Nebraska--Lincoln 
Lines: 15
Distribution: na
NNTP-Posting-Host: unlinfo.unl.edu

   I have a BACK MACHINE and have had one since January.  While I have not 
found it to be a panacea for my back pain, I think it has helped somewhat. 
It MAINLY acts to stretch muscles in the back and prevent spasms associated
with pain.  I am taking less pain medication than I was previously.  
   The folks at BACK TECHNOLOGIES are VERY reluctant to honor their return 
policy.  They extended my "warranty" period rather than allow me to return 
the machine when, after the first month or so, I was not thrilled with it. 
They encouraged me to continue to use it, abeit less vigourously. 
   Like I said, I can't say it is a cure-all, but it keeps me stretched out
and I am in less pain.
--
***********************************************************************
Dale M. Webb, DVM, PhD           *  97% of the body is water.  The
Veterinary Diagnostic Center     *  other 3% keeps you from drowning.
University of Nebraska, Lincoln  *
sci_space_outliers = df_outliers[df_outliers['Class Name'] == 'sci.space']
print(sci_space_outliers['Text'].iloc[0])
MACH 25 landing site bases?
Article-I.D.: aurora.1993Apr5.193829.1
Organization: University of Alaska Fairbanks
Lines: 7
Nntp-Posting-Host: acad3.alaska.edu

The supersonic booms hear a few months ago over I belive San Fran, heading east
of what I heard, some new super speed Mach 25 aircraft?? What military based
int he direction of flight are there that could handle a Mach 25aircraft on its
landing decent?? Odd question??

==
Michael Adams,  -- I'm not high, just jacked

Próximas etapas

Você criou um detector de anomalias usando embeddings. Tente usar seus próprios dados textuais para visualizá-los como embeddings e escolha algum limite para detectar outliers. É possível realizar a redução de dimensionalidade para concluir a etapa de visualização. Observe que t-SNE é bom para agrupar entradas, mas pode levar mais tempo para convergir ou ficar preso em mínimos locais. Caso você se depare com esse problema, use a análise de componentes principais (PCA, na sigla em inglês).

Para saber mais sobre como usar embeddings, consulte estes outros tutoriais: