How to set initial weights in MLPClassifier?

Question

I cannot find a way to set the initial weights of the neural network, could someone tell me how please? I am using python package sklearn.neural_network.MLPClassifier.

Answer 1

Solution: A working solution is to inherit from MLPClassifier and override the _init_coef method. In the _init_coef write the code to set the initial weights. Then use the new class "MLPClassifierOverride" as in the example below instead of "MLPClassifier"

# new class
class MLPClassifierOverride(MLPClassifier):
# Overriding _init_coef method
def _init_coef(self, fan_in, fan_out):
    if self.activation == 'logistic':
        init_bound = np.sqrt(2. / (fan_in + fan_out))
    elif self.activation in ('identity', 'tanh', 'relu'):
        init_bound = np.sqrt(6. / (fan_in + fan_out))
    else:
        raise ValueError("Unknown activation function %s" %
                         self.activation)
    coef_init = ### place your initial values for coef_init here

    intercept_init = ### place your initial values for intercept_init here

    return coef_init, intercept_init

Answer 2

multilayer_perceptron.py initializes the weights based on the nonlinear function used for hidden layers. If you want to try a different initialization, you can take a look at the function _init_coef here and modify as you desire.

Answer 3

The docs show you the attributes in use.

Attributes:
...

coefs_ : list, length n_layers - 1 The ith element in the list represents the weight matrix corresponding to > layer i.

intercepts_ : list, length n_layers - 1 The ith element in the list represents the bias vector corresponding to layer > i + 1.

Just build your classifier clf=MLPClassifier(solver="sgd") and set coefs_ and intercepts_ before calling clf.fit().

The only remaining question is: does sklearn overwrite your inits?

The code looks like:

    if not hasattr(self, 'coefs_') or (not self.warm_start and not
                                       incremental):
        # First time training the model
        self._initialize(y, layer_units)

This looks to me like it won't replace your given coefs_ (you might check biases too).

The packing and unpacking functions further indicates that this should be possible. These are probably used for serialization through pickle internally.