BEAT - tutorial/mlp

This algorithm is a legacy one. The API has changed since its implementation. New versions and forks will need to be updated.

Endpoint Groups 1

Algorithms have at least one input and one output. All algorithm endpoints are organized in groups. Groups are used by the platform to indicate which inputs and outputs are synchronized together. The first group is automatically synchronized with the channel defined by the block in which the algorithm is deployed.

Group: main

Endpoint Name	Data Format	Nature
class_id	system/uint64/1	Input
image	system/array_2d_floats/1	Input
model	tutorial/mlp/1	Output

Parameters 3

Parameters allow users to change the configuration of an algorithm when scheduling an experiment

Name	Type	Default
number-of-iterations	uint32	50
seed	uint32	0
number-of-hidden-units	uint32	10

import bob
import numpy

def mlp_from_data(data):
    """Unmangles a dict of bob.machine.MLP from a BEAT Data object"""

n_inputs  = data.input_subtract.shape[0]
    n_outputs = data.biases[-1].value.shape[0]
    n_hidden  = len(data.biases) - 1
    shape = [n_inputs]
    for i in range(n_hidden): shape.append(data.biases[i].value.shape[0])
    shape.append(n_outputs)
    mlp = bob.machine.MLP(shape)

mlp.input_subtract = data.input_subtract
    mlp.input_divide   = data.input_divide
    weights = []
    for v in data.weights: weights.append(v.value)
    biases = []
    for v in data.biases:  biases.append(v.value)
    mlp.weights = weights
    mlp.biases  = biases

return mlp

class Algorithm:

def __init__(self):
        self.model = None

def process(self, inputs, outputs):

# retrieve the model
      if self.model is None:
        inputs['model'].next()
        self.model = mlp_from_data(inputs['model'].data)

# probe image
      image = inputs['image'].data.value.flatten()

scores = {}
      scores = self.model(image)

# Get the class_id with maximum score
      max_id = scores.argmax()
      outputs['scores'].write({
            'real_identity': inputs['class_id'].data.value,
            'estimated_identity': numpy.uint64(max_id),
            'score': scores[max_id]
         })

return True

The code for this algorithm in Python
The ruler at 80 columns indicate suggested POSIX line breaks (for readability).
The editor will automatically enlarge to accomodate the entirety of your input
Use keyboard shortcuts for search/replace and faster editing. For example, use Ctrl-F (PC) or Cmd-F (Mac) to search through this box

This algorithm implements a scoring procedure for a multi-layer perceptron (MLP) [Bishop] [Duda], a neural network architecture that has some well-defined characteristics such as a feed-forward structure.

This implementation relies on the Bob library.

The inputs are:

image: a two-dimensional array of floats (64 bits), which is flattened for the training procedure
class_id: an identifier for the class of the image, such that supervised training is possible

The output scores is the corresponding set of score values.

[Bishop]

Pattern Recognition and Machine Learning, C.M. Bishop, chapter 5

[Duda]

Pattern Classification, Duda, Hart and Stork, chapter 6

No experiments are using this algorithm.

This algorithm was never executed.

algorithms tutorial mlp_training 3

Endpoint Groups 1

Group: main

Parameters 3

algorithms

tutorial

mlp_training

3