如果需要修改模型规范的参数，可以使用update()代替从头开始重新创建对象。
用法
# S3 method for bag_mars
update(
object,
parameters = NULL,
num_terms = NULL,
prod_degree = NULL,
prune_method = NULL,
fresh = FALSE,
...
)
# S3 method for bag_mlp
update(
object,
parameters = NULL,
hidden_units = NULL,
penalty = NULL,
epochs = NULL,
fresh = FALSE,
...
)
# S3 method for bag_tree
update(
object,
parameters = NULL,
cost_complexity = NULL,
tree_depth = NULL,
min_n = NULL,
class_cost = NULL,
fresh = FALSE,
...
)
# S3 method for bart
update(
object,
parameters = NULL,
trees = NULL,
prior_terminal_node_coef = NULL,
prior_terminal_node_expo = NULL,
prior_outcome_range = NULL,
fresh = FALSE,
...
)
# S3 method for boost_tree
update(
object,
parameters = NULL,
mtry = NULL,
trees = NULL,
min_n = NULL,
tree_depth = NULL,
learn_rate = NULL,
loss_reduction = NULL,
sample_size = NULL,
stop_iter = NULL,
fresh = FALSE,
...
)
# S3 method for C5_rules
update(
object,
parameters = NULL,
trees = NULL,
min_n = NULL,
fresh = FALSE,
...
)
# S3 method for cubist_rules
update(
object,
parameters = NULL,
committees = NULL,
neighbors = NULL,
max_rules = NULL,
fresh = FALSE,
...
)
# S3 method for decision_tree
update(
object,
parameters = NULL,
cost_complexity = NULL,
tree_depth = NULL,
min_n = NULL,
fresh = FALSE,
...
)
# S3 method for discrim_flexible
update(
object,
num_terms = NULL,
prod_degree = NULL,
prune_method = NULL,
fresh = FALSE,
...
)
# S3 method for discrim_linear
update(
object,
penalty = NULL,
regularization_method = NULL,
fresh = FALSE,
...
)
# S3 method for discrim_quad
update(object, regularization_method = NULL, fresh = FALSE, ...)
# S3 method for discrim_regularized
update(
object,
frac_common_cov = NULL,
frac_identity = NULL,
fresh = FALSE,
...
)
# S3 method for gen_additive_mod
update(
object,
select_features = NULL,
adjust_deg_free = NULL,
parameters = NULL,
fresh = FALSE,
...
)
# S3 method for linear_reg
update(
object,
parameters = NULL,
penalty = NULL,
mixture = NULL,
fresh = FALSE,
...
)
# S3 method for logistic_reg
update(
object,
parameters = NULL,
penalty = NULL,
mixture = NULL,
fresh = FALSE,
...
)
# S3 method for mars
update(
object,
parameters = NULL,
num_terms = NULL,
prod_degree = NULL,
prune_method = NULL,
fresh = FALSE,
...
)
# S3 method for mlp
update(
object,
parameters = NULL,
hidden_units = NULL,
penalty = NULL,
dropout = NULL,
epochs = NULL,
activation = NULL,
learn_rate = NULL,
fresh = FALSE,
...
)
# S3 method for multinom_reg
update(
object,
parameters = NULL,
penalty = NULL,
mixture = NULL,
fresh = FALSE,
...
)
# S3 method for naive_Bayes
update(object, smoothness = NULL, Laplace = NULL, fresh = FALSE, ...)
# S3 method for nearest_neighbor
update(
object,
parameters = NULL,
neighbors = NULL,
weight_func = NULL,
dist_power = NULL,
fresh = FALSE,
...
)
# S3 method for pls
update(
object,
parameters = NULL,
predictor_prop = NULL,
num_comp = NULL,
fresh = FALSE,
...
)
# S3 method for poisson_reg
update(
object,
parameters = NULL,
penalty = NULL,
mixture = NULL,
fresh = FALSE,
...
)
# S3 method for proportional_hazards
update(
object,
parameters = NULL,
penalty = NULL,
mixture = NULL,
fresh = FALSE,
...
)
# S3 method for rand_forest
update(
object,
parameters = NULL,
mtry = NULL,
trees = NULL,
min_n = NULL,
fresh = FALSE,
...
)
# S3 method for rule_fit
update(
object,
parameters = NULL,
mtry = NULL,
trees = NULL,
min_n = NULL,
tree_depth = NULL,
learn_rate = NULL,
loss_reduction = NULL,
sample_size = NULL,
penalty = NULL,
fresh = FALSE,
...
)
# S3 method for surv_reg
update(object, parameters = NULL, dist = NULL, fresh = FALSE, ...)
# S3 method for survival_reg
update(object, parameters = NULL, dist = NULL, fresh = FALSE, ...)
# S3 method for svm_linear
update(
object,
parameters = NULL,
cost = NULL,
margin = NULL,
fresh = FALSE,
...
)
# S3 method for svm_poly
update(
object,
parameters = NULL,
cost = NULL,
degree = NULL,
scale_factor = NULL,
margin = NULL,
fresh = FALSE,
...
)
# S3 method for svm_rbf
update(
object,
parameters = NULL,
cost = NULL,
rbf_sigma = NULL,
margin = NULL,
fresh = FALSE,
...
)参数
- object
-
型号规格。
- parameters
-
包含要更新的主要参数的 1 行 tibble 或命名列表。更新时直接使用
parameters或主要参数。如果使用主要参数，这些参数将取代parameters中的值。此外，在此对象中使用引擎参数将导致错误。 - num_terms
-
最终模型中将保留的特征数量，包括截距。
- prod_degree
-
尽可能高的交互程度。
- prune_method
-
修剪方法。
- fresh
-
是否应该就地修改参数或批量替换参数的逻辑。
- ...
-
不用于
update()。 - hidden_units
-
隐藏模型中单元数量的整数。
- penalty
-
一个非负数，表示某些引擎使用的正则化量。
- epochs
-
训练迭代次数的整数。
- cost_complexity
-
CART 模型(仅限特定引擎)使用的成本/复杂性参数(也称为
Cp)的正数。 - tree_depth
-
树的最大深度的整数。
- min_n
-
节点中进一步拆分节点所需的最小数据点数量的整数。
- class_cost
-
类成本的非负标量(其中成本为 1 意味着没有额外成本)。当结果因子的第一级是少数群体时，这非常有用。如果不是这种情况，则可以使用零到一之间的值来偏置到因子的第二级。
- trees
-
集合中包含的树数的整数。
- prior_terminal_node_coef
-
节点是终端节点的先验概率的系数。
- prior_terminal_node_expo
-
节点是终端节点的先验概率的指数。
- prior_outcome_range
-
一个正值，定义预测结果在一定范围内的先验宽度。对于回归来说，它与观察到的数据范围有关；先验是由数据观测范围定义的高斯分布的标准差数。对于分类，它被定义为+/-3的范围(假设在logit标度上)。默认值为 2。
- mtry
-
创建树模型时将在每次分割时随机采样的预测变量的数量(或比例)的数字(仅限特定引擎)
- learn_rate
-
提升算法从迭代到迭代的适应速率的数字(仅限特定引擎)。这有时称为收缩参数。
- loss_reduction
-
进一步分割所需的损失函数减少的数字(仅限特定引擎)。
- sample_size
-
一个数字，表示暴露给拟合例程的数据数量(或比例)。对于
xgboost，在每次迭代时进行采样，而C5.0在训练期间采样一次。 - stop_iter
-
停止前没有改进的迭代次数(仅限特定引擎)。
- committees
-
集合成员数量的非负整数(不大于 100)。
- neighbors
-
用于调整基于模型的预测的训练集实例数量的 0 到 9 之间的整数。
- max_rules
-
规则数量最多。
- regularization_method
-
正则化估计类型的字符串。可能的值为：“
diagonal”、“min_distance”、“shrink_cov”和“shrink_mean”(仅限sparsediscrim引擎)。 - frac_common_cov, frac_identity
-
零到一之间的数值。
- select_features
-
TRUE或FALSE.如果是TRUE，则模型有能力消除预测变量(通过惩罚)。增加adjust_deg_free将增加删除预测变量的可能性。 - adjust_deg_free
-
如果是
select_features = TRUE，则充当平滑度的乘数。将此值增加到 1 以上以生成更平滑的模型。 - mixture
-
0 到 1(含)之间的数字，表示模型中 L1 正则化(即 lasso)的比例。
-
mixture = 1指定纯套索模型， -
mixture = 0指定岭回归模型，并且 -
0 < mixture < 1指定弹性网络模型，插值套索和岭。
仅适用于特定发动机。
-
- dropout
-
0(含)到 1 之间的数字，表示模型训练期间随机设置为零的模型参数的比例。
- activation
-
表示原始预测变量和隐藏单元层之间关系类型的单个字符串。隐藏层和输出层之间的激活函数根据结果类型自动设置为 "linear" 或 "softmax"。可能的值为："linear"、"softmax"、"relu" 和 "elu"
- smoothness
-
一个非负数，表示类边界的相对平滑度。较小的示例会导致模型灵活的边界，较大的值会生成适应性较差的类边界
- Laplace
-
用于平滑 low-frequency 计数的拉普拉斯校正的非负值。
- weight_func
-
用于对样本之间的距离进行加权的核函数类型的单个字符。有效的选择是：
"rectangular","triangular","epanechnikov","biweight","triweight","cos","inv","gaussian","rank"或"optimal"。 - dist_power
-
用于计算 Minkowski 距离的参数的单个数字。
- predictor_prop
-
每个 PLS 分量可以具有非零系数的原始预测变量的最大比例(通过正则化)。该值用于 X 的所有 PLS 分量。
- num_comp
-
要保留的 PLS 组件的数量。
- dist
-
结果概率分布的字符串。默认为"weibull"。
- cost
-
预测样本在边以内或错误一侧的成本的正数
- margin
-
SVM 不敏感损失函数中 epsilon 的正数(仅回归)
- degree
-
多项式次数为正数。
- scale_factor
-
多项式比例因子的正数。
- rbf_sigma
-
径向基函数的正数。
例子
# ------------------------------------------------------------------------------
model <- C5_rules(trees = 10, min_n = 2)
model
#> ! parsnip could not locate an implementation for `C5_rules` model
#> specifications.
#> ℹ The parsnip extension package rules implements support for this
#> specification.
#> ℹ Please install (if needed) and load to continue.
#> C5.0 Model Specification (classification)
#>
#> Main Arguments:
#> trees = 10
#> min_n = 2
#>
#> Computational engine: C5.0
#>
update(model, trees = 1)
#> ! parsnip could not locate an implementation for `C5_rules` model
#> specifications.
#> ℹ The parsnip extension package rules implements support for this
#> specification.
#> ℹ Please install (if needed) and load to continue.
#> C5.0 Model Specification (classification)
#>
#> Main Arguments:
#> trees = 1
#> min_n = 2
#>
#> Computational engine: C5.0
#>
update(model, trees = 1, fresh = TRUE)
#> ! parsnip could not locate an implementation for `C5_rules` model
#> specifications.
#> ℹ The parsnip extension package rules implements support for this
#> specification.
#> ℹ Please install (if needed) and load to continue.
#> C5.0 Model Specification (classification)
#>
#> Main Arguments:
#> trees = 1
#>
#> Computational engine: C5.0
#>
# ------------------------------------------------------------------------------
model <- cubist_rules(committees = 10, neighbors = 2)
model
#> ! parsnip could not locate an implementation for `cubist_rules` model
#> specifications.
#> ℹ The parsnip extension package rules implements support for this
#> specification.
#> ℹ Please install (if needed) and load to continue.
#> Cubist Model Specification (regression)
#>
#> Main Arguments:
#> committees = 10
#> neighbors = 2
#>
#> Computational engine: Cubist
#>
update(model, committees = 1)
#> ! parsnip could not locate an implementation for `cubist_rules` model
#> specifications.
#> ℹ The parsnip extension package rules implements support for this
#> specification.
#> ℹ Please install (if needed) and load to continue.
#> Cubist Model Specification (regression)
#>
#> Main Arguments:
#> committees = 1
#> neighbors = 2
#>
#> Computational engine: Cubist
#>
update(model, committees = 1, fresh = TRUE)
#> ! parsnip could not locate an implementation for `cubist_rules` model
#> specifications.
#> ℹ The parsnip extension package rules implements support for this
#> specification.
#> ℹ Please install (if needed) and load to continue.
#> Cubist Model Specification (regression)
#>
#> Main Arguments:
#> committees = 1
#>
#> Computational engine: Cubist
#>
model <- pls(predictor_prop = 0.1)
model
#> ! parsnip could not locate an implementation for `pls` model
#> specifications.
#> ℹ The parsnip extension package plsmod implements support for this
#> specification.
#> ℹ Please install (if needed) and load to continue.
#> PLS Model Specification (unknown mode)
#>
#> Main Arguments:
#> predictor_prop = 0.1
#>
#> Computational engine: mixOmics
#>
update(model, predictor_prop = 1)
#> ! parsnip could not locate an implementation for `pls` model
#> specifications.
#> ℹ The parsnip extension package plsmod implements support for this
#> specification.
#> ℹ Please install (if needed) and load to continue.
#> PLS Model Specification (unknown mode)
#>
#> Main Arguments:
#> predictor_prop = 1
#>
#> Computational engine: mixOmics
#>
update(model, predictor_prop = 1, fresh = TRUE)
#> ! parsnip could not locate an implementation for `pls` model
#> specifications.
#> ℹ The parsnip extension package plsmod implements support for this
#> specification.
#> ℹ Please install (if needed) and load to continue.
#> PLS Model Specification (unknown mode)
#>
#> Main Arguments:
#> predictor_prop = 1
#>
#> Computational engine: mixOmics
#>
# ------------------------------------------------------------------------------
model <- rule_fit(trees = 10, min_n = 2)
model
#> ! parsnip could not locate an implementation for `rule_fit` model
#> specifications.
#> ℹ The parsnip extension packages agua and rules implement support for
#> this specification.
#> ℹ Please install (if needed) and load to continue.
#> RuleFit Model Specification (unknown mode)
#>
#> Main Arguments:
#> trees = 10
#> min_n = 2
#>
#> Computational engine: xrf
#>
update(model, trees = 1)
#> ! parsnip could not locate an implementation for `rule_fit` model
#> specifications.
#> ℹ The parsnip extension packages agua and rules implement support for
#> this specification.
#> ℹ Please install (if needed) and load to continue.
#> RuleFit Model Specification (unknown mode)
#>
#> Main Arguments:
#> trees = 1
#> min_n = 2
#>
#> Computational engine: xrf
#>
update(model, trees = 1, fresh = TRUE)
#> ! parsnip could not locate an implementation for `rule_fit` model
#> specifications.
#> ℹ The parsnip extension packages agua and rules implement support for
#> this specification.
#> ℹ Please install (if needed) and load to continue.
#> RuleFit Model Specification (unknown mode)
#>
#> Main Arguments:
#> trees = 1
#>
#> Computational engine: xrf
#>
model <- boost_tree(mtry = 10, min_n = 3)
model
#> Boosted Tree Model Specification (unknown mode)
#>
#> Main Arguments:
#> mtry = 10
#> min_n = 3
#>
#> Computational engine: xgboost
#>
update(model, mtry = 1)
#> Boosted Tree Model Specification (unknown mode)
#>
#> Main Arguments:
#> mtry = 1
#> min_n = 3
#>
#> Computational engine: xgboost
#>
update(model, mtry = 1, fresh = TRUE)
#> Boosted Tree Model Specification (unknown mode)
#>
#> Main Arguments:
#> mtry = 1
#>
#> Computational engine: xgboost
#>
param_values <- tibble::tibble(mtry = 10, tree_depth = 5)
model %>% update(param_values)
#> Boosted Tree Model Specification (unknown mode)
#>
#> Main Arguments:
#> mtry = 10
#> min_n = 3
#> tree_depth = 5
#>
#> Computational engine: xgboost
#>
model %>% update(param_values, mtry = 3)
#> Boosted Tree Model Specification (unknown mode)
#>
#> Main Arguments:
#> mtry = 10
#> min_n = 3
#> tree_depth = 5
#>
#> Computational engine: xgboost
#>
param_values$verbose <- 0
# Fails due to engine argument
# model %>% update(param_values)
model <- linear_reg(penalty = 10, mixture = 0.1)
model
#> Linear Regression Model Specification (regression)
#>
#> Main Arguments:
#> penalty = 10
#> mixture = 0.1
#>
#> Computational engine: lm
#>
update(model, penalty = 1)
#> Linear Regression Model Specification (regression)
#>
#> Main Arguments:
#> penalty = 1
#> mixture = 0.1
#>
#> Computational engine: lm
#>
update(model, penalty = 1, fresh = TRUE)
#> Linear Regression Model Specification (regression)
#>
#> Main Arguments:
#> penalty = 1
#>
#> Computational engine: lm
#>
相关用法
- R parsnip predict.model_fit 模型预测
- R parsnip proportional_hazards 比例风险回归
- R parsnip logistic_reg 逻辑回归
- R parsnip linear_reg 线性回归
- R parsnip C5_rules C5.0 基于规则的分类模型
- R parsnip set_engine 声明计算引擎和特定参数
- R parsnip condense_control 将控制对象压缩为更小的控制对象
- R parsnip control_parsnip 控制拟合函数
- R parsnip augment 通过预测增强数据
- R parsnip repair_call 修复模型调用对象
- R parsnip dot-model_param_name_key 翻译模型调整参数的名称
- R parsnip glm_grouped 将数据集中的分组二项式结果与个案权重拟合
- R parsnip rule_fit 规则拟合模型
- R parsnip svm_rbf 径向基函数支持向量机
- R parsnip set_args 更改模型规范的元素
- R parsnip translate 解决计算引擎的模型规范
- R parsnip max_mtry_formula 根据公式确定 mtry 的最大值。此函数可能会根据公式和数据集限制 mtry 的值。对于生存和/或多变量模型来说，这是一种安全的方法。
- R parsnip svm_linear 线性支持向量机
- R parsnip set_new_model 注册模型的工具
- R parsnip rand_forest 随机森林
- R parsnip mlp 单层神经网络
- R parsnip nearest_neighbor K-最近邻
- R parsnip fit 将模型规范拟合到数据集
- R parsnip boost_tree 增强树
- R parsnip bart 贝叶斯加性回归树 (BART)
注：本文由纯净天空筛选整理自Max Kuhn等大神的英文原创作品 Updating a model specification。非经特殊声明，原始代码版权归原作者所有，本译文未经允许或授权，请勿转载或复制。