Skip to contents

acc <- sample_weight %*% (y_true == which.max(y_pred))

You can provide logits of classes as y_pred, since argmax of logits and probabilities are same.

This metric creates two local variables, total and count that are used to compute the frequency with which y_pred matches y_true. This frequency is ultimately returned as sparse categorical accuracy: an idempotent operation that simply divides total by count.

If sample_weight is NULL, weights default to 1. Use sample_weight of 0 to mask values.

Usage

metric_sparse_categorical_accuracy(
  y_true,
  y_pred,
  ...,
  name = "sparse_categorical_accuracy",
  dtype = NULL
)

Arguments

y_true

Tensor of true targets.

y_pred

Tensor of predicted targets.

...

For forward/backward compatability.

name

(Optional) string name of the metric instance.

dtype

(Optional) data type of the metric result.

Value

If y_true and y_pred are missing, a Metric instance is returned. The Metric instance that can be passed directly to compile(metrics = ), or used as a standalone object. See ?Metric for example usage. If y_true and y_pred are provided, then a tensor with the computed value is returned.

Usage

Standalone usage:

m <- metric_sparse_categorical_accuracy()
m$update_state(rbind(2L, 1L), rbind(c(0.1, 0.6, 0.3), c(0.05, 0.95, 0)))
m$result()

## tf.Tensor(0.5, shape=(), dtype=float32)

m$reset_state()
m$update_state(rbind(2L, 1L), rbind(c(0.1, 0.6, 0.3), c(0.05, 0.95, 0)),
               sample_weight = c(0.7, 0.3))
m$result()

## tf.Tensor(0.3, shape=(), dtype=float32)

Usage with compile() API:

model %>% compile(optimizer = 'sgd',
                  loss = 'sparse_categorical_crossentropy',
                  metrics = list(metric_sparse_categorical_accuracy()))