python - 在多类分类问题中使用带有 Keras 的 class_weights 参数时出错
问题描述
这个问题已在其他论坛中提出过，我尝试了他们的变体但无济于事：class_weight 用于不平衡数据 - Keras
然而，它似乎陈旧，因为没有人回答这个问题。有谁知道class_weight在使用时如何在 Keras 中实现参数categorical_crossentropy？我一直在尝试class_weight在 Keras 中使用该参数，但不断收到此错误：
ValueError: 预期
class_weight是一个字典，其键从 0 到比类数少一，发现 {'prediction': {0: 1.217169570760731, 1: 5.323420074349443, 2: 0.5023680056130504}
每个样本将被分类为 0、1 或 2（softmax）。该数据集中的偏差很大。我的模型使用 Keras 功能 API。
class_weights 是使用 Sklearn 计算的：
class_weights = class_weight.compute_class_weight('balanced', np.unique(np.array(y_trn_labels_HB_2_pd['labels'])), y_trn_labels_HB_2_pd['labels'])
class_weight_dict = dict(enumerate(class_weights))
class_weight_dict
这是我的最后一层：
prediction = Dense(3, activation="softmax", name = 'prediction')(x)
这是我的模型：
tf.__version__ = 2.3.0
model = Model(inputs = [sequence_input_head, sequence_input_body, semantic_feat,
wordOL_feat, avg_subj_feat], outputs = [prediction])
model.compile(loss = 'categorical_crossentropy',
optimizer='adam',
metrics = ['accuracy'])
model.summary()
这是我的 class_weight 参数：
class_weight= {'prediction': {0:1.217169570760731, 1:5.323420074349443, 2:0.5023680056130504} })
这是完整的错误：
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-271-e3bb78b84171> in <module>()
26 y_val_2_cat),
27 callbacks = [es],
---> 28 class_weight= {'prediction': class_weights})
29
30 modeled = model.save(os.path.join(save_path, path_model))
3 frames
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py in _method_wrapper(self, *args, **kwargs)
106 def _method_wrapper(self, *args, **kwargs):
107 if not self._in_multi_worker_mode(): # pylint: disable=protected-access
--> 108 return method(self, *args, **kwargs)
109
110 # Running inside `run_distribute_coordinator` already.
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, validation_batch_size, validation_freq, max_queue_size, workers, use_multiprocessing)
1061 use_multiprocessing=use_multiprocessing,
1062 model=self,
-> 1063 steps_per_execution=self._steps_per_execution)
1064
1065 # Container that configures and calls `tf.keras.Callback`s.
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in __init__(self, x, y, sample_weight, batch_size, steps_per_epoch, initial_epoch, epochs, shuffle, class_weight, max_queue_size, workers, use_multiprocessing, model, steps_per_execution)
1120 dataset = self._adapter.get_dataset()
1121 if class_weight:
-> 1122 dataset = dataset.map(_make_class_weight_map_fn(class_weight))
1123 self._inferred_steps = self._infer_steps(steps_per_epoch, dataset)
1124
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in _make_class_weight_map_fn(class_weight)
1299 "Expected `class_weight` to be a dict with keys from 0 to one less "
1300 "than the number of classes, found {}").format(class_weight)
-> 1301 raise ValueError(error_msg)
1302
1303 class_weight_tensor = ops.convert_to_tensor_v2(
ValueError: Expected `class_weight` to be a dict with keys from 0 to one less than the number of classes, found {'prediction': {0: 1.217169570760731, 1: 5.323420074349443}}
编辑1：
我试过你的建议@Prateek Bhatt
history = model.fit({'headline': hl_pd_tr, 'articleBody':bd_pd_train, 'semantic': semantic_sim_180_train_x, 'wordOverlap': wrd_OvLp_train_x, 'avg_subjectivity': avg_subj_hb_train_x}, #@param ["model.fit({'headline': hl_pd_tr, 'articleBody':bd_pd_train},", "model.fit({'headline': hl_pd_tr, 'articleBody':bd_pd_train, 'semantic': semantic_x_tr},", "model.fit({'headline': hl_pd_tr, 'articleBody':bd_pd_train, 'semantic': semantic_x_tr, 'wordOverlap': wrd_OvLp_x_tr},", "model.fit({'headline': hl_pd_tr, 'articleBody':bd_pd_train, 'semantic': semantic_x_tr, 'wordOverlap': wrd_OvLp_x_tr, 'avgsubj': avg_subj_x_tr},"] {type:"raw", allow-input: true}
{'prediction':y_train_2_cat},
epochs=100,
batch_size= BATCH__SIZE,
shuffle= True,
validation_data = ([hl_pd_val, bd_pd_val, semantic_sim_180_val_x, wrd_OvLp_val_x, avg_subj_hb_val_x], y_val_2_cat),
callbacks = [es],
class_weight= {0:1.217169570760731, 1:5.323420074349443, 2:0.5023680056130504})
但是，我收到此错误：
ValueError: `class_weight` is only supported for Models with a single output.
完整错误：
ValueError Traceback (most recent call last)
<ipython-input-272-bfbab936a723> in <module>()
26 y_val_2_cat),
27 callbacks = [es],
---> 28 class_weight= {0:1.217169570760731, 1:5.323420074349443, 2:0.5023680056130504})
29
30 modeled = model.save(os.path.join(save_path, path_model))
16 frames
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py in _method_wrapper(self, *args, **kwargs)
106 def _method_wrapper(self, *args, **kwargs):
107 if not self._in_multi_worker_mode(): # pylint: disable=protected-access
--> 108 return method(self, *args, **kwargs)
109
110 # Running inside `run_distribute_coordinator` already.
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, validation_batch_size, validation_freq, max_queue_size, workers, use_multiprocessing)
1061 use_multiprocessing=use_multiprocessing,
1062 model=self,
-> 1063 steps_per_execution=self._steps_per_execution)
1064
1065 # Container that configures and calls `tf.keras.Callback`s.
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in __init__(self, x, y, sample_weight, batch_size, steps_per_epoch, initial_epoch, epochs, shuffle, class_weight, max_queue_size, workers, use_multiprocessing, model, steps_per_execution)
1120 dataset = self._adapter.get_dataset()
1121 if class_weight:
-> 1122 dataset = dataset.map(_make_class_weight_map_fn(class_weight))
1123 self._inferred_steps = self._infer_steps(steps_per_epoch, dataset)
1124
/usr/local/lib/python3.6/dist-packages/tensorflow/python/data/ops/dataset_ops.py in map(self, map_func, num_parallel_calls, deterministic)
1693 """
1694 if num_parallel_calls is None:
-> 1695 return MapDataset(self, map_func, preserve_cardinality=True)
1696 else:
1697 return ParallelMapDataset(
/usr/local/lib/python3.6/dist-packages/tensorflow/python/data/ops/dataset_ops.py in __init__(self, input_dataset, map_func, use_inter_op_parallelism, preserve_cardinality, use_legacy_function)
4043 self._transformation_name(),
4044 dataset=input_dataset,
-> 4045 use_legacy_function=use_legacy_function)
4046 variant_tensor = gen_dataset_ops.map_dataset(
4047 input_dataset._variant_tensor, # pylint: disable=protected-access
/usr/local/lib/python3.6/dist-packages/tensorflow/python/data/ops/dataset_ops.py in __init__(self, func, transformation_name, dataset, input_classes, input_shapes, input_types, input_structure, add_to_graph, use_legacy_function, defun_kwargs)
3369 with tracking.resource_tracker_scope(resource_tracker):
3370 # TODO(b/141462134): Switch to using garbage collection.
-> 3371 self._function = wrapper_fn.get_concrete_function()
3372 if add_to_graph:
3373 self._function.add_to_graph(ops.get_default_graph())
/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/function.py in get_concrete_function(self, *args, **kwargs)
2937 """
2938 graph_function = self._get_concrete_function_garbage_collected(
-> 2939 *args, **kwargs)
2940 graph_function._garbage_collector.release() # pylint: disable=protected-access
2941 return graph_function
/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/function.py in _get_concrete_function_garbage_collected(self, *args, **kwargs)
2904 args, kwargs = None, None
2905 with self._lock:
-> 2906 graph_function, args, kwargs = self._maybe_define_function(args, kwargs)
2907 seen_names = set()
2908 captured = object_identity.ObjectIdentitySet(
/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/function.py in _maybe_define_function(self, args, kwargs)
3211
3212 self._function_cache.missed.add(call_context_key)
-> 3213 graph_function = self._create_graph_function(args, kwargs)
3214 self._function_cache.primary[cache_key] = graph_function
3215 return graph_function, args, kwargs
/usr/local/lib/python3.6/dist-packages/tensorflow/python/eager/function.py in _create_graph_function(self, args, kwargs, override_flat_arg_shapes)
3073 arg_names=arg_names,
3074 override_flat_arg_shapes=override_flat_arg_shapes,
-> 3075 capture_by_value=self._capture_by_value),
3076 self._function_attributes,
3077 function_spec=self.function_spec,
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/func_graph.py in func_graph_from_py_func(name, python_func, args, kwargs, signature, func_graph, autograph, autograph_options, add_control_dependencies, arg_names, op_return_value, collections, capture_by_value, override_flat_arg_shapes)
984 _, original_func = tf_decorator.unwrap(python_func)
985
--> 986 func_outputs = python_func(*func_args, **func_kwargs)
987
988 # invariant: `func_outputs` contains only Tensors, CompositeTensors,
/usr/local/lib/python3.6/dist-packages/tensorflow/python/data/ops/dataset_ops.py in wrapper_fn(*args)
3362 attributes=defun_kwargs)
3363 def wrapper_fn(*args): # pylint: disable=missing-docstring
-> 3364 ret = _wrapper_helper(*args)
3365 ret = structure.to_tensor_list(self._output_structure, ret)
3366 return [ops.convert_to_tensor(t) for t in ret]
/usr/local/lib/python3.6/dist-packages/tensorflow/python/data/ops/dataset_ops.py in _wrapper_helper(*args)
3297 nested_args = (nested_args,)
3298
-> 3299 ret = autograph.tf_convert(func, ag_ctx)(*nested_args)
3300 # If `func` returns a list of tensors, `nest.flatten()` and
3301 # `ops.convert_to_tensor()` would conspire to attempt to stack
/usr/local/lib/python3.6/dist-packages/tensorflow/python/autograph/impl/api.py in wrapper(*args, **kwargs)
253 try:
254 with conversion_ctx:
--> 255 return converted_call(f, args, kwargs, options=options)
256 except Exception as e: # pylint:disable=broad-except
257 if hasattr(e, 'ag_error_metadata'):
/usr/local/lib/python3.6/dist-packages/tensorflow/python/autograph/impl/api.py in converted_call(f, args, kwargs, caller_fn_scope, options)
530
531 if not options.user_requested and conversion.is_whitelisted(f):
--> 532 return _call_unconverted(f, args, kwargs, options)
533
534 # internal_convert_user_code is for example turned off when issuing a dynamic
/usr/local/lib/python3.6/dist-packages/tensorflow/python/autograph/impl/api.py in _call_unconverted(f, args, kwargs, options, update_cache)
337
338 if kwargs is not None:
--> 339 return f(*args, **kwargs)
340 return f(*args)
341
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in _class_weights_map_fn(*data)
1310 if nest.is_sequence(y):
1311 raise ValueError(
-> 1312 "`class_weight` is only supported for Models with a single output.")
1313
1314 if y.shape.rank > 2:
ValueError: `class_weight` is only supported for Models with a single output.
解决方案
如果您有 pandas 数据框，您可以首先使用 compute_class_weight 函数计算 class_weight 参数并传递目标列，例如：
from sklearn.utils import class_weight
class_weights = class_weight.compute_class_weight('balanced',np.unique(df['target']),df['target'])
class_weights = dict(enumerate(class_weights))
然后在拟合时通过class_weights：
history = model.fit(X, Y,epochs=n,class_weight=class_weights)
然后确保使用 categorical_crossentropy 使用 one-hot-encoded 标签进行损失计算。如果您使用索引值，请改用 sparse_categorical_crossentropy。
推荐阅读
- css - 根据某些条件更改 TreeNode 的颜色；使用 antd React 库中的树
- python - 在 Visual Studio 之外，Python 脚本要慢得多
- apache-spark - Spark SQL：有没有办法获得一个大小取决于持续时间而不是项目数量的滑动窗口？
- javascript - 使用 ES6 数组按对象排序的 Javascript 失败
- java - 如何在我的 java 类中将 firebase 值存储为全局整数
- python - CSV转换行列的所有组合
- node.js - 高级调度程序 - Heroku - smtp 服务发送到电子邮件
- ruby - 为什么浮点数除以零不会在 Ruby 中引发异常？
- docker - Docker 卷和绑定
- c - Linux 套接字：有没有办法只接收来自特定环的数据包？