TF学习——TF之API：TensorFlow的高级机器学习API—tf.contrib.learn的简介、使用方法、案例应用之详细攻略

tf.contrib.learn的简介

tf.contrib.learn 是 TensorFlow 提供的一个机器学习高级 API 模块，让用户可以更方便的配置、训练和评估各种各样的机器学习模型，里面内置了很多模型可以直接调用，其实，就是类似最经典的机器学习库 scikit-learn 的 API 。很多调用方法都很相同。

1、常用的模型

线性分类 tf.contrib.learn.LinearClassifier

线性回归 tf.contrib.learn.LinearRegressor

逻辑斯谛回归 tf.contrib.learn.LogisticRegressor

K-means 聚类 tf.contrib.learn.KMeansClustering

神经网络分类器 tf.contrib.learn.DNNClassifier

神经网络回归器 tf.contrib.learn.DNNRegressor

广度深度回归 tf.contrib.learn.DNNLinearCombinedRegressor

广度深度分类 tf.contrib.learn.DNNLinearCombinedClassifier

注意：该模块已被弃用，请仔细查看API说明，

https://github.com/tensorflow/tensorflow/blob/r1.13/tensorflow/contrib/learn/README.md

Many canned estimators (subclasses of Estimator) have equivalents in core exposed under tf.estimator: DNNClassifier, DNNRegressor, DNNEstimator, LinearClassifier, LinearRegressor, LinearEstimator, DNNLinearCombinedClassifier,DNNLinearCombinedRegressor and DNNLinearCombinedEstimator.

To migrate to the new api, users need to take the following steps:

Replace tf.contrib.learn with tf.estimator.

tf.contrib.learn的使用方法

Modules

datasets module: Dataset utilities and synthetic/reference datasets (deprecated).

graph_actions module: High level operations on graphs (deprecated).

head module: Abstractions for the head(s) of a model (deprecated).

io module: Tools to allow different io formats (deprecated).

learn_runner module: Utilities to run and tune an Experiment (deprecated).

models module: Various high level TF models (deprecated).

monitors module: Monitors instrument the training process (deprecated).

ops module: Various TensorFlow Ops (deprecated).

preprocessing module: Preprocessing tools useful for building models (deprecated).

utils module: TensorFlow Learn Utils (deprecated).

Classes

class BaseEstimator: Abstract BaseEstimator class to train and evaluate TensorFlow models.

class DNNClassifier: A classifier for TensorFlow DNN models.

class DNNEstimator: A Estimator for TensorFlow DNN models with user specified _Head.

class DNNLinearCombinedClassifier: A classifier for TensorFlow Linear and DNN joined training models.

class DNNLinearCombinedEstimator: An estimator for TensorFlow Linear and DNN joined training models.

class DNNLinearCombinedRegressor: A regressor for TensorFlow Linear and DNN joined training models.

class DNNRegressor: A regressor for TensorFlow DNN models.

class DynamicRnnEstimator: Dynamically unrolled RNN (deprecated).

class Estimator: Estimator class is the basic TensorFlow model trainer/evaluator.

class Evaluable: Interface for objects that are evaluatable by, e.g., Experiment.

class Experiment: Experiment is a class containing all information needed to train a model.

class ExportStrategy: A class representing a type of model export.

class Head: Interface for the head/top of a model.

class InputFnOps: A return type for an input_fn (deprecated).

class KMeansClustering: An Estimator for K-Means clustering.

class LinearClassifier: Linear classifier model.

class LinearEstimator: Linear model with user specified head.

class LinearRegressor: Linear regressor model.

class MetricSpec: MetricSpec connects a model to metric functions.

class ModeKeys: Standard names for model modes (deprecated).

class ModelFnOps: Ops returned from a model_fn.

class NanLossDuringTrainingError: Unspecified run-time error.

class NotFittedError: Exception class to raise if estimator is used before fitting.

class PredictionKey: THIS CLASS IS DEPRECATED.

class ProblemType: Enum-like values for the type of problem that the model solves.

class RunConfig: This class specifies the configurations for an Estimator run.

class SKCompat: Scikit learn wrapper for TensorFlow Learn Estimator.

class SVM: Support Vector Machine (SVM) model for binary classification.

class TaskType: DEPRECATED CLASS.

class Trainable: Interface for objects that are trainable by, e.g., Experiment.

Functions

LogisticRegressor(...): Builds a logistic regression Estimator for binary classification.

binary_svm_head(...): Creates a Head for binary classification with SVMs. (deprecated)

build_parsing_serving_input_fn(...): Build an input_fn appropriate for serving, expecting fed tf.Examples. (deprecated)

evaluate(...): Evaluate a model loaded from a checkpoint. (deprecated)

extract_dask_data(...): Extract data from dask.Series or dask.DataFrame for predictors. (deprecated)

extract_dask_labels(...): Extract data from dask.Series or dask.DataFrame for labels. (deprecated)

extract_pandas_data(...): Extract data from pandas.DataFrame for predictors. (deprecated)

extract_pandas_labels(...): Extract data from pandas.DataFrame for labels. (deprecated)

extract_pandas_matrix(...): Extracts numpy matrix from pandas DataFrame. (deprecated)

infer(...): Restore graph from restore_checkpoint_path and run output_dict tensors. (deprecated)

infer_real_valued_columns_from_input(...): Creates FeatureColumn objects for inputs defined by input x. (deprecated)

infer_real_valued_columns_from_input_fn(...): Creates FeatureColumn objects for inputs defined by input_fn. (deprecated)

make_export_strategy(...): Create an ExportStrategy for use with Experiment. (deprecated)

multi_class_head(...): Creates a Head for multi class single label classification. (deprecated)

multi_head(...): Creates a MultiHead stemming from same logits/hidden layer. (deprecated)

multi_label_head(...): Creates a Head for multi label classification. (deprecated)

no_op_train_fn(...): DEPRECATED FUNCTION

poisson_regression_head(...): Creates a Head for poisson regression. (deprecated)

read_batch_examples(...): Adds operations to read, queue, batch Example protos. (deprecated)

read_batch_features(...): Adds operations to read, queue, batch and parse Example protos. (deprecated)

read_batch_record_features(...): Reads TFRecord, queues, batches and parses Example proto. (deprecated)

read_keyed_batch_examples(...): Adds operations to read, queue, batch Example protos. (deprecated)

read_keyed_batch_examples_shared_queue(...): Adds operations to read, queue, batch Example protos. (deprecated)

read_keyed_batch_features(...): Adds operations to read, queue, batch and parse Example protos. (deprecated)

read_keyed_batch_features_shared_queue(...): Adds operations to read, queue, batch and parse Exampleprotos.(deprecated)

regression_head(...): Creates a Head for linear regression. (deprecated)

run_feeds(...): See run_feeds_iter(). Returns a list instead of an iterator. (deprecated)

run_n(...): Run output_dict tensors n times, with the same feed_dict each run. (deprecated)

train(...): Train a model. (deprecated)

tf.contrib.learn的案例应用

更新……