> For the complete documentation index, see [llms.txt](https://deeplearning4j.konduit.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://deeplearning4j.konduit.ai/en-1.0.0-rewrite/model-import/overview-1/losses.md).

# Losses

### Keras Loss Functions

DL4J supports all standard Keras loss functions except `logcosh`. Loss function mapping is implemented in [KerasLossUtils](https://github.com/eclipse/deeplearning4j/blob/master/deeplearning4j/deeplearning4j-modelimport/src/main/java/org/deeplearning4j/nn/modelimport/keras/utils/KerasLossUtils.java).

***

### Loss Mapping Table

| Keras Loss                                | DL4J Loss Class        | Supported |
| ----------------------------------------- | ---------------------- | --------- |
| `mean_squared_error` / `mse`              | `LossMSE`              | Yes       |
| `mean_absolute_error` / `mae`             | `LossMAE`              | Yes       |
| `mean_absolute_percentage_error` / `mape` | `LossMAPE`             | Yes       |
| `mean_squared_logarithmic_error` / `msle` | `LossMSLE`             | Yes       |
| `squared_hinge`                           | `LossSquaredHinge`     | Yes       |
| `hinge`                                   | `LossHinge`            | Yes       |
| `categorical_hinge`                       | `LossCategoricalHinge` | Yes       |
| `logcosh`                                 | —                      | No        |
| `categorical_crossentropy`                | `LossMCXENT`           | Yes       |
| `sparse_categorical_crossentropy`         | `LossSparseMCXENT`     | Yes       |
| `binary_crossentropy`                     | `LossBinaryXENT`       | Yes       |
| `kullback_leibler_divergence` / `kld`     | `LossKLD`              | Yes       |
| `poisson`                                 | `LossPoisson`          | Yes       |
| `cosine_proximity`                        | `LossCosineProximity`  | Yes       |

***

### Loss Descriptions

#### mean\_squared\_error

Mean of squared differences between predictions and targets. Standard regression loss.

```
MSE = mean((y_true - y_pred)^2)
```

Use for regression tasks with normally distributed errors.

#### mean\_absolute\_error

Mean of absolute differences. More robust to outliers than MSE.

```
MAE = mean(|y_true - y_pred|)
```

#### mean\_absolute\_percentage\_error

Scale-independent percentage error.

```
MAPE = mean(|y_true - y_pred| / |y_true|) * 100
```

#### mean\_squared\_logarithmic\_error

Useful when targets span several orders of magnitude.

```
MSLE = mean((log(1 + y_pred) - log(1 + y_true))^2)
```

#### squared\_hinge

Squared hinge loss for binary classification with labels in {-1, +1}.

```
squared_hinge = mean(max(0, 1 - y_true * y_pred)^2)
```

#### hinge

Standard SVM-style hinge loss.

```
hinge = mean(max(0, 1 - y_true * y_pred))
```

#### categorical\_hinge

Multi-class hinge loss variant.

#### categorical\_crossentropy

Standard multi-class cross-entropy. Requires one-hot encoded targets and softmax output.

```
categorical_crossentropy = -sum(y_true * log(y_pred))
```

#### sparse\_categorical\_crossentropy

Same as categorical\_crossentropy but accepts integer class indices rather than one-hot vectors.

#### binary\_crossentropy

Binary cross-entropy for binary or multi-label classification with sigmoid output.

```
binary_crossentropy = -y_true * log(y_pred) - (1 - y_true) * log(1 - y_pred)
```

#### kullback\_leibler\_divergence

Measures divergence between two distributions.

```
KLD = sum(y_true * log(y_true / y_pred))
```

Use for variational autoencoders and distributional matching.

#### poisson

Poisson loss for count-based predictions.

```
poisson = mean(y_pred - y_true * log(y_pred))
```

#### cosine\_proximity

Negative cosine similarity. Used for metric learning and similarity tasks.

```
cosine_proximity = -sum(y_true * y_pred) / (||y_true|| * ||y_pred||)
```

***

### logcosh (Not Supported)

`logcosh` (log of the hyperbolic cosine of the prediction error) is not implemented in DL4J. Use `mean_absolute_error` as an alternative for outlier-robust regression tasks.

***

### Notes

Loss functions are read from the Keras training configuration embedded in the HDF5 file. They are only present when the model was compiled (`model.compile(...)`) and saved with `model.save()`. When loading for inference only, pass `enforceTrainingConfig=false` to bypass training configuration parsing.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://deeplearning4j.konduit.ai/en-1.0.0-rewrite/model-import/overview-1/losses.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.