resizable_all2all.py

# -*- coding: utf-8 -*-
"""
.. invisible:
     _   _ _____ _     _____ _____
    | | | |  ___| |   |  ___/  ___|
    | | | | |__ | |   | |__ \ `--.
    | | | |  __|| |   |  __| `--. \
    \ \_/ / |___| |___| |___/\__/ /
     \___/\____/\_____|____/\____/

Created on Feb 4, 2015

███████████████████████████████████████████████████████████████████████████████

Licensed to the Apache Software Foundation (ASF) under one
or more contributor license agreements.  See the NOTICE file
distributed with this work for additional information
regarding copyright ownership.  The ASF licenses this file
to you under the Apache License, Version 2.0 (the
"License"); you may not use this file except in compliance
with the License.  You may obtain a copy of the License at

  http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing,
software distributed under the License is distributed on an
"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
KIND, either express or implied.  See the License for the
specific language governing permissions and limitations
under the License.

███████████████████████████████████████████████████████████████████████████████
"""


import numpy

from veles.znicz.all2all import All2All


class ResizableAll2All(All2All):
    MAPPING = {"all2all_resizable"}

    @All2All.output_sample_shape.setter
    def output_sample_shape(self, value):
        old_neurons_number = self.neurons_number if self.is_initialized else 0
        self._set_output_sample_shape(value)
        if not self.is_initialized:
            return
        if self.neurons_number <= 0:
            raise ValueError(
                "Neurons number must be greater than 0 (got %d)" % value)
        self._adjust_neurons_number(self.neurons_number - old_neurons_number)

    def _adjust_neurons_number(self, delta):
        if not self.weights_transposed:
            old_nn = self.weights.shape[0]
            new_weights = numpy.zeros((old_nn + delta, self.weights.shape[1]),
                                      self.weights.dtype)
            if delta > 0:
                new_weights[:old_nn] = self.weights.mem
                self.fill_array(self.weights_filling, new_weights[old_nn:],
                                self.weights_stddev)
            else:
                new_weights[:] = self.weights.mem[:new_weights.shape[0]]
        else:
            old_nn = self.weights.shape[1]
            new_weights = numpy.zeros((self.weights.shape[0], old_nn + delta),
                                      self.weights.dtype)
            if delta > 0:
                new_weights[:, :old_nn] = self.weights.mem
                self.fill_array(self.weights_filling, new_weights[:, old_nn:],
                                self.weights_stddev)
            else:
                new_weights[:] = self.weights.mem[:, :new_weights.shape[1]]
        self.weights.reset(new_weights)
        self.output.reset()
        self._create_output()
        self.init_vectors(self.weights, self.output)
        self._backend_init_()