A simple Python program for an ANN to cover the MNIST dataset – I – a starting point

For beginners both in Python and Machine Learning [ML] the threshold to do some real programming and create your own Artificial Neural Network [ANN] seems to be relatively high. Well, some readers might say: Why program an ANN by yourself at a basic Python level at all when Keras and TensorFlow [TF] are available? Answer: For learning! And eventually to be able to do some things TF has not been made for. And as readers of this blog will see in the future, I have some ideas along this line …

So I thought, just let me set up a small Python3 and Numpy based program to create a simple kind of ANN – a “Multilayer Perceptron” [MLP] – and train it for the MNIST dataset. I expected that my readers and I myself would earn something on various methods used in ML during our numerical experiments. Well, we shall see ..-

Regarding ANN-theory I take a brutal shortcut and assume that my readers are already acquainted with the following topics:

  1. what simple neural networks with “hidden” layers look like and what its weights are,
  2. what a “cost function” is and how the gradient descent method works,
  3. what logistic regression is and what the cost function J for it looks like,
  4. why we need the back propagation of deviations from known values and why it gives you the required partial derivatives of a typical cost function with respect to an ANN’s “weights”,
  5. what a mini-batch approach to weight optimization is.

I cannot spare you the effort of studying most of these topics in advance. Otherwise I would have to write an introductory book on ML myself. [I would do, but you need to give me a sponsor 🙂 ] But even if you are not fully acquainted with all the named topics: I shall briefly comment on each and every of the points in the forthcoming articles. The real basics are, however, much better and more precisely documented in the literature; e.g. in the books of Geron and Rashka (see the references at the end of this article). I recommend to read one of these books whilst we move on with a sequence of steps to build the basic code for our MLP.

We need a relatively well defined first objective for the usage of our ANN. We shall concentrate on classification tasks. As a first example we shall use the conventional MNIST data set. The MNIST data set consists of images of handwritten numbers with 28×28 pixels [px]. It is a standard data set used in many elementary courses on ML. The challenge for our ANN is that it should be able to recognize hand-written digits from a digitized gray-color image after some training.

Note that this task does NOT require the use of a fully fletched multi-layer MLP. “Stochastic Gradient Descent”-approaches for a pure binary classificator to determine (linear) separation surfaces in combination with a “One-versus-All” strategy for multi-category-classification may be sufficient. See chapters 3 to 5 in the book of Geron for more information.

Regarding the build up of the ANN program, I basically follow an approach described by S. Raschka in his book (see the second to last section for a reference). However, at multiple points I take my freedom to organize the code differently and comment in my own way … I am only a beginner in Python; I hope my insights are helpful for others in the same situation.

In any case you should make yourself familiar with numpy arrays and their “shapes”. I assume that you understand the multidimensional structure of Numpy arrays ….


To avoid confusion, I use the following wording and synonyms:

Category: Each input data element is associated with a category to which it belongs. In our case a category
corresponds to a “digit”. The classification algorithm (here: the MLP) may achieve an ability to predict the association of an unknown MNIST like input data sample with its correct category. It should – after some training – detect the (non-linear) separation interfaces for categories in a multidimensional feature space. In the case of MNIST we speak about ten categories corresponding to 10 digits, including zero.

Label: A category may be described by a label. Training data may provide a so called “target label array” _y_train for all input data. We must be prepared to transform target labels for input data into a usable form for an ANN, i.e. into a vectorized form, which selects a specific category out of many. This process is called “label encoding”.

Input data set: A complete “set” of input data. Such a set consists of individual “elements” or “records“. Another term which we I shall frequently use for such an element is a “sample“. The MNIST input set of training data consists of 60000 records or samples – which we provide via an array _X_train. The array is two dimensional as each sample consists of values more multiple properties.

Feature: A sample of the input data set may be equivalent to a mathematical vector, whose elements specify (numerical) values for multiple properties – so called “features” – of a sample. Thus, input samples correspond to points in a multidimensional feature space.

Output data set: A complete set of output data after a so called “propagation” through the ANN for the input data set. “Propagation”means a series of defined mathematical transformations of the original features data of the input sample. The number of samples or records in the output data sets is equal to the number of records in the input data set. The output set will be represented by a Numpy array “_ay_ANN_out“.

A data record or sample of the input data set: One distinct element of the input data set (and its array). Note that such an element itself may be a multidimensional array covering all features in a distinct form. Such arrays represents a so called “tensor”.

A data record of the output data set: One distinct element of the output data set. Note that such an element itself may be an array covering all possible categories in a distinct form. E.g., we may be given a “probability” for each category – which allows us to decide with which of the categories we should associate the output element.

A simple MLP network – layers, nodes, weights

An AN network is composed of a series of horizontally and/or vertically arranged layers with nodes. The nodes represent the artificial neurons. A MLP is an ANN which has a rather simple structure: It consists of an input layer, multiple sequential intermediate “hidden” layers, and an output layer. All nodes of a specific layer are connected with all nodes of neighboring (!) layers, only. We speak of a “dense” or fully “connected layer” structure.

The simplifying sketch below displays an ANN with just three sequentially arranged layers – an input layer, a “hidden” middle layer and an output layer.

Note that in general there can be (many) more hidden layers than just one. Note also that modern ANNs (e.g. Convoluted Networks) may have a much more complicated topological structure with hundreds of layers. There may also be
cascaded networks where a specific layer has connections to many more than just the neighbor layers.

Input layer and its number of nodes
To feed input data into the MLP we need an “input layer” with sufficient input nodes. How many? Well, this depends on the number of features your data set represents. In the MNIST case a sample image contains 28×28 pixels. For each pixel with a gray value (integer number between 0 and 256) given. So a typical image represents 28×28 = 768 different “features” – i.e. 786 numbers for “gray”-values between 0 and 255. We need as many input nodes in our MLP to represent the full image information by the input layer.

For other input data the number of features may be different; in addition features my follow a multidimensional order or organization which first must be “flattened” out in one dimension. The number of input nodes must then be adjusted accordingly. The number of input nodes should, therefore, be a parameter or be derived from information on the type of input data. The way of how you map complicated and structured features to input layers and whether you map all data to a one dimensional input vector is a question one should think about carefully. (Most people today treat e.g. a time dimension of input data as just a special form of a feature – I regard this as questionable in some cases, but this is beyond this article series …)

For our MLP we always assume a mapping of features to a flat one dimensional vector like structure.

Output layer and its number of nodes
We shall use our MLP for classification tasks in the beginning. We, therefore, assume that the output of the ANN should allow for the distinction between “strong>NC” different categories an input data set can belong to. In case of the MNIST dataset we can distinguish between 10 different digits. Thus an output layer must in this case comprise 10 different nodes. To be able to cover other data sets with a different number of categories the number of output nodes must be a parameter of our program, too.

How we indicate the association of a (transformed) sample at the output layer to a category numerically – by a probability number between “0” and “1” or just a “1” at the right category and zeros otherwise – can be a matter of discussion. It is also a question of the cost function we wish to use. We will come back to this point in later articles.

The numbers of “hidden layers” and their nodes
We want the numbers of nodes on “hidden layers” to be parameters for our program. For simple data as MNIST images we do not need big networks, but we want to be able to play around a bit with 1 up to 3 layers. (For an ANN to recognize hand written MNIST digits an input layer “L0” and only one hidden layer “L1” before an output layer “L2” are fully sufficient. Nevertheless in most of our experiments we will actually use 2 hidden layers. There are three reasons: You can approximate any continuous function with two hidden layers (with a special non-linear activation function; see below) and an output layer (with just a linear output function). The other reason is that the full mathematical complexity of “learning” of a MLP appears with two hidden layers (see a later article).

Activation and output functions
The nodes in hidden layers use a so called “activation function” to transform aggregated input from different feeding nodes of the previous layer into one distinct value within a defined interval – e.g. between -1 and 1. Again, we should be prepared to have a program parameter to choose between different “activation functions”.

We should be aware of the fact that the nodes of the output layers need special consideration as the “activation function” there
produces the final output – which in turn must allow for a distinction of categories. This may lead to a special form – e.g. a kind of probability function. So, the type of the “output function” should also be regarded as variable parameter.

A Python class for our ANN and its interface

I develop my code as a Python module in an Eclipse/PyDev IDE, which itself uses a virtual Python3 environment. I described the setup of such a development environment in detail in another previous article of this blog. In the resulting directory structure of the PyDev project I place a module “myann.py” at the location “…../ml_1/mynotebooks/mycode/myann.py”. This file shall contain the code of class “MyANN” for our ANN.

Modules and libraries to import

We need to import some libraries at the head of our Python program first:

Module to create a simple layered neural network for the MNIST data set
Created on 23.08.2019
@author: ramoe
import numpy as np
import math 
import sys
import time
import tensorflow
from sklearn.datasets import fetch_mldata
from sklearn.datasets import fetch_openml
from keras.datasets import mnist as kmnist
from scipy.special import expit  
from matplotlib import pyplot as plt
#from matplotlib.colors import ListedColormap
#import matplotlib.patches as mpat 
#from keras.activations import relu


Why do I import “tensorflow” and “keras”?
Well, only for the purpose to create the input data of MNIST quickly. Sklearn’s “fetchml_data” is doomed to end. The alternative “fetch_openml” does not use caching in some older versions and is also in general terribly slow. But, “keras”, which in turn needs tensorflow as a backend, provides its own tool to provide the MNIST data.

We need “scipy” to get an optimized version of the so called “sigmoid“-function – which is an important version of an activation function. We shall use it most of the time. “numpy” and “math” are required for fast array- and math-operations. “time” is required to measure the run time of program segments and “mathplotlib” will help us to visualize some information gathered during and after training.

The “__init__”-function of our class MyANN

We encapsulate most of the required functionality in a class and its methods. Python provides the “__init__”-function, which we can use as a kind of “constructor” – although it technically is not the same as a constructor in other languages. Anyway, we can use it as an interface to feed in parameters and to initialize variables of a class instance.

We shall build up our “__init__()”-function during the next articles step by step. In the beginning we shall only focus on attributes and methods of our class required to import the MNIST data and put them into Numpy arrays and to create the basic network layers.


class MyANN:
    def __init__(self, 
                 my_data_set = "mnist", 
                 n_hidden_layers = 1, 
                 ay_nodes_layers = [0, 100, 0], # array which should have as much elements as n_hidden + 2
                 n_nodes_layer_out = 10,  # number of nodes in output layer 
                 my_activation_function = "sigmoid", 
                 my_out_function        = "sigmoid",   
                 vect_mode = 'cols', 
                 figs_x1=12.0, figs_x2=8.0, 
                 legend_loc='upper right'
        Initialization of MyANN
            data_set: type of dataset; so far only the "mnist", "mnist_784" and the "mnist_keras" datsets are known. 
                      We use this information to prepare the input data and learn about the feature dimension. 
                      This info is used in preparing the size of the input layer.     
            n_hidden_layers = number of hidden layers => between input layer 0 and output layer n 
            ay_nodes_layers = [0, 100, 0 ] : We set the number of nodes in input layer_0 and the output_layer to zero 
                                             Will be set to real number afterwards by infos from the input dataset. 
                                             All other numbers are used for the node numbers of the hidden layers.
            n_nodes_layer_out = expected number of nodes in the output layer (is checked); 
                                this number corresponds to the number of categories to be distinguished   
            my_activation_function : name of the activation function to use 
            my_out_function : name of the "activation" function of the last layer whcih produces the output values 
            vect_mode: Are 1-dim data arrays (vectors) ordered by columns or rows ?
            figs_x1=12.0, figs_x2=8.0 : Standard sizing of plots , 
            legend_loc='upper right': Position of legends in the plots 


You see that I defined multiple parameters, which are explained in the Python “doc”-string. We use a “string” to choose the dataset to train our ANN on. To be able to work on other data sets later on we assume that specific methods for importing a variety of special input data sets are implemented in our class. This requires that the class knows exactly which kinds of data sets it is capable to handle. We provide an list with this information below. The other parameters should be clear from their inline documentation.

Initialization of class attributes

We first initialize a bunch of class attributes which we shall use to define the network of layers, nodes, weights, to keep our input data and functions.

        # Array (Python list) of known input data sets 
        self.__input_data_sets = ["mnist", "mnist_784", "mnist_keras"]  
        self._my_data_set = my_data_set
        # X, y, X_train, y_train, X_test, y_test  
            # will be set by analyze_input_data 
            # X: Input array (2D) - at present status of MNIST image data, only.    
            # y: result (=classification data) [digits represent categories in the case of Mnist]
        self._X       = None 
        self._X_train = None 
        self._X_test  = None   
        self._y       = None 
        self._y_train = None 
        self._y_test  = None
        # relevant dimensions 
        # from input data information;  will be set in handle_input_data()
        self._dim_sets     = 0  
        self._dim_features = 0  
        self._n_labels     = 0   # number of unique labels - will be extracted from y-data 
        # Img sizes 
        self._dim_img      = 0 # should be sqrt(dim_features) - we assume square like images  
        self._img_h        = 0 
        self._img_w        = 0 
        # Layers
        # ------
        # number of hidden layers 
        self._n_hidden_layers = n_hidden_layers
        # Number of total layers 
        self._n_total_layers = 2 + self._n_hidden_layers  
        # Nodes for hidden layers 
        self._ay_nodes_layers = np.array(ay_nodes_layers)
        # Number of nodes in output layer - will be checked against information from 
target arrays
        self._n_nodes_layer_out = n_nodes_layer_out
        # Weights 
        # --------
        # empty List for all weight-matrices for all layer-connections
        # Numbering : 
        # w[0] contains the weight matrix which connects layer 0 (input layer ) to hidden layer 1 
        # w[1] contains the weight matrix which connects layer 1 (input layer ) to (hidden?) layer 2 
        self._ay_w = []  
        # Known Randomizer methods ( 0: np.random.randint, 1: np.random.uniform )  
        # ------------------
        self.__ay_known_randomizers = [0, 1]

        # Types of activation functions and output functions 
        # ------------------
        self.__ay_activation_functions = ["sigmoid"] # later also relu 
        self.__ay_output_functions     = ["sigmoid"] # later also softmax 
        # the following dictionaries will be used for indirect function calls 
        self.__d_activation_funcs = {
            'sigmoid': self._sigmoid, 
            'relu':    self._relu
        self.__d_output_funcs = { 
            'sigmoid': self._sigmoid, 
            'softmax': self._softmax

        # The following variables will later be set by _check_and set_activation_and_out_functions()            
        self._my_act_func = my_activation_function
        self._my_out_func = my_out_function
        self._act_func = None    
        self._out_func = None    

        # Plot handling 
        # --------------
        # Alternatives to resize plots 
        # 1: just resize figure  2: resize plus create subplots() [figure + axes] 
        self._plot_resize_alternative = 1 
        # Plot-sizing
        self._figs_x1 = figs_x1
        self._figs_x2 = figs_x2
        self._fig = None
        self._ax  = None 
                # alternative 2 does resizing and (!) subplots() 

        # ***********
        # operations 
        # ***********
        # check and handle input data 
        print("\nStopping program regularily")

To make things not more complicated as necessary I omit the usage of “properties” and a full encapsulation of private attributes. For convenience reasons I use only one underscore for some attributes and functions/methods to allow for external usage. This is helpful in a testing phase. However, many items can in the end be switched to really private properties or methods.

List of known input datasets

The list of known input data sets is kept in the variable “self.__input_data_sets”. The variables

self._X, self._X_train, self._X_test, self._y, self._y_train, self._y_test

will be used to keep all sample data of the chosen dataset – i.e. the training data, the test data for checking of the reliability of the algorithm after training and the corresponding target data (y_…) for classification – in distinct array variables during code execution. The target data in the MNIST case contain the digit a specific sample image ( of _X_train or _X_test..) represents.

All of the named attributes will become Numpy arrays. A method called “_handle_input_data(self)” will load the (MNIST) input data and fill the arrays.

The input arrays “X_…” will via their dimensions provide the information on the number of data sets (_dim_sets) and the number of features (_dim_features). Numpy provides the various dimensions of multidimensional arrays in form of a tuple.

The target data arrays “_y_…” provides the number of “categories” (MNIST: 10 digits) the ANN must distinguish after training. We keep this number in the
attribute “_n_labels”. It is also useful to keep the pixel dimensions of input image data. At least for MNIST we assume quadratic images (_img_h = img_w = _dim_img).

Layers and weights

The number of total layers (“_n_total_layers”) is by 2 bigger than the number of hidden layers (_n_hidden_layers).

We take the number of nodes in the layers from the respective list provided as an input parameter “ay_nodes_layers” to our class. We transform the list into a numpy array “_ay_nodes_layers”. The expected number of nodes in the output layer is used for consistency checks and saved in the variable “_n_nodes_layer_out”.

The “weights” of an ANN must be given in form of matrices: A weight describes a connection between two nodes of different adjacent layers. So we have as many connections as there are node combinations (nodex_(N+1), nodey_N), with “nodex_N” meaning a node on layer L_N and nodey_(N+1) a node on layer L_(N+1).

As the number of layers is not fixed, but can be set by the user, I use a Python list “_ay_w” to collect such matrices in the order of layer_0 (input) to layer_n (output).

Weights, i.e. the matrix elements, must initially be set as random numbers. To provide such numbers we have to use randomizer functions. Depending on the kind (floating point numbers, integer numbers) of random numbers to produce we use at least two randomizers (randint, uniform). For the weights we use the uniform-randomizer.

Allowed activation and output function names are listed in Python dictionaries which point to respective methods. This allows for an “indirect addressing” of these functions later on. You may recognize this by the direct reference of the dictionary elements to defined class methods (no strings are used!).

For the time being we work with the “sigmoid” and the “relu” functions for activation and the “sigmoid” and “softmax” functions for output creation. The attributes “self._act_func” and “self._out_func” are used later on to invoke the functions requested by the respective parameters of the classes interface.

The final part of the code segment given above is used for plot-sizing with the help of “matplotlib”; a method “initiate_and_resize_plot()” takes care of this. It can use 2 alternative ways of doing so.

Read and provide the input data

Now let us turn to some methods. We first need to read in and prepare the input data. We use a method “_handle_input_data()” to work on this problem. For the time being we have only three different ways to load the MNIST dataset from different origins.

    # Method to handle different types of input data sets 
    def _handle_input_data(self):    
        Method to deal with the input data: 
        - check if we have a known data set ("mnist" so far)
        - reshape as required 
        - analyze dimensions and extract the feature dimension(s) 
        # check for known dataset 
            if (self._my_data_set not in self._input_data_sets ): 
                raise ValueError
        except ValueError:
            print("The requested input data" + self._my_data_set + " is not known!" )

        # handle the mnist original dataset 
        if ( self._my_data_set == "mnist"): 
            mnist = fetch_mldata('MNIST original')
            self._X, self._y = mnist["data"], mnist["target"]
            print("Input data for dataset " + self._my_data_set + " : \n" + "Original shape of X = " + str(self._X.shape) +
              "\n" + "Original shape of y = " + str(self._y.shape))
            self._X_train, self._X_test, self._y_train, self._y_test = self._X[:60000], self._X[60000:],
 self._y[:60000], self._y[60000:] 
        # handle the mnist_784 dataset 
        if ( self._my_data_set == "mnist_784"): 
            mnist2 = fetch_openml('mnist_784', version=1, cache=True, data_home='~/scikit_learn_data') 
            self._X, self._y = mnist2["data"], mnist2["target"]
            print ("data fetched")
            # the target categories are given as strings not integers 
            self._y = np.array([int(i) for i in self._y])
            print ("data modified")
            print("Input data for dataset " + self._my_data_set + " : \n" + "Original shape of X = " + str(self._X.shape) +
              "\n" + "Original shape of y = " + str(self._y.shape))
            self._X_train, self._X_test, self._y_train, self._y_test = self._X[:60000], self._X[60000:], self._y[:60000], self._y[60000:] 

        # handle the mnist_keras dataset 
        if ( self._my_data_set == "mnist_keras"): 
            (self._X_train, self._y_train), (self._X_test, self._y_test) = kmnist.load_data()
            len_train =  self._X_train.shape[0]
            print("Input data for dataset " + self._my_data_set + " : \n" + "Original shape of X_train = " + str(self._X_train.shape) +
              "\n" + "Original Shape of y_train = " + str(self._y_train.shape))
            len_test =  self._X_test.shape[0]
            print("Original shape of X_test = " + str(self._X_test.shape) +
              "\n" + "Original Shape of y_test = " + str(self._y_test.shape))
            self._X_train = self._X_train.reshape(len_train, 28*28) 
            self._X_test  = self._X_test.reshape(len_test, 28*28) 

        # Common Mnist handling 
        if ( self._my_data_set == "mnist" or self._my_data_set == "mnist_784" or self._my_data_set == "mnist_keras" ): 
        # Other input data sets can not yet be handled 

We first check, whether the input parameter fits a known dataset – and raise an error if otherwise. The data come in different forms for the three sources of MNIST. For each set we want to extract the arrays

self._X_train, self._X_test, self._y_train, self._y_test.

We have to do this a bit differently for the 3 cases. Note that the “mnist_784” set from “fetch_openml” gives the target category values in form of strings and not integers. We correct this directly after loading.

The fastest method for importing the MNIST dataset is based on “keras”; the keras function “kmnist.load_data()” provides already a 60000:10000 ratio for training and test data. However, we get the images in a (60000, 28, 28) array shape; we therefore reshape the “_X_train”-array to (60000, 784) and “_X_test”-array to (10000, 784).

Analysis of the input data and the one-hot-encoding of the target labels

A further handling of the MNIST data requires some common analysis.

    # Method for common input data handling of Mnist data sets
    def _common_handling_of_mnist(self):    
        print("\nFinal input data for dataset " + self._my_data_set + 
              " : \n" + "Shape of X_train = " + str(self._X_train.shape) + 
              "\n" + "Shape of y_train = " + str(self._y_train.shape) + 
              "\n" + "Shape of X_test = " + str(self._X_test.shape) + 
              "\n" + "Shape of y_test = " + str(self._y_test.shape) 
        # mixing the training indices 
        shuffled_index = np.random.permutation(60000)
        self._X_train, self._y_train = self._X_train[shuffled_index], self._y_train[shuffled_index]
        # set 
        self._dim_sets = self._y_train.shape[0]
        self._dim_features = self._X_train.shape[1] 
        self._dim_img = math.sqrt(self._dim_features)
        # we assume square images 
        self._img_h = int(self._dim_img)
        self._img_w = int(self._dim_img)
        # Print dimensions  
        print("\nWe have " + str(self._dim_sets) + " data sets for training") 
        print("Feature dimension is " + str(self._dim_features) + " (= " + str(self._img_w)+ "x" + str(self._img_h) + ")") 

        # we need to encode the digit labels of mnist 

As you see we retrieve some of our class attributes which we shall use during training and do some printing. This is trivial. Not as trivial is, however, the handling of the output data:

What shape do we expect for the “_X_train” and “_y_train”? Each element of the input data set is an array with values for all features. Thus the “_X_train.shape” should be (60000, 784). For _y_train we expect a simple integer describing the digit to which the MNIST input image corresponds. Thus we expect a one dimensional array with _y_train.shape = (60000). So far, so good …

But: The output data of our ANN for one input element will be provided as an array of values for our 10 different categories – and not as a simple number. To account for this we need to encode the “_y_train”-data, i.e. the target labels, into an usable array form. We use two methods to achieve this:

    # Method to get the number of target labels 
    def _get_num_labels(self):
        self._n_labels = len(np.unique(self._y_train))
        print("The number of labels is " + str(self._n_labels))

    # Method to encode all mnist labels 
    def _encode_all_mnist_labels(self, b_print=True):
        We shall use vectorized input and output - i.e. we process a whole batch of input data sets in parallel
        (see article in the Linux blog) 
        The output array will then have the form OUT(i_out_node, idx) where 
            i_out_node enumerates the node of the last layer (i.e. the category)    
            idx enumerates the data set within a batch,
        After training, if y_train[idx] = 6, we would expect an output value of OUT[6,idx] = 1.0 and OUT[i_node, idx]=0.0 otherwise 
        for a categorization decision in the ideal case. Realistically, we will get a distribution of numbers over the nodes 
        with values between 0.0 and 1.0 - with hopefully the maximum value at the right node OUT[6,idx].  
        The following method creates an arrays OneHot[i_out_node, idx] with 
        OneHot[i_node_out, idx] = 1.0, if i_node_out =  int(y[idx])
        OneHot(i_node_out, idx] = 0.0, if i_node_out != int(y[idx])  
        This will allow for a vectorized comparison of calculated values and knwon values during training 
        self._ay_onehot = np.zeros((self._n_labels, self._y_train.shape[0]))
        # ay_oneval is just for convenience and printing purposes 
        self._ay_oneval = np.zeros((self._n_labels, self._y_train.shape[0], 2))
        if b_print: 
            print("\nShape of y_train = " + str(self._y_train.shape))
            print("Shape of ay_onehot = " + str(self._ay_onehot.shape))
        # the next block is just for illustration purposes and a better understanding
        if b_print: 
            values = enumerate(self._y_train[0:12])
            print("\nValues of the enumerate structure for the first 12 elements : = ")

            for iv in values: 
        # here we prepare the array for vectorized 
        print("\nLabels for the first 12 datasets:")
        for idx, val in enumerate(self._y_train):
            self._ay_onehot[val, idx ]   = 1.0
            self._ay_oneval[val, idx, 0] = 1.0 
            self._ay_oneval[val, idx, 1] = val 
        if b_print: 
            print("\nShape of ay_onehot = " + str(self._ay_onehot.shape))
            print(self._ay_onehot[:, 0:12])
            #print("Shape of ay_oneval = " + str(self._ay_oneval.shape))
            #print(self._ay_oneval[:, 0:12, :])

The first method only determines the number of labels (= number of categories). We see from the code of the second method that we encode the target labels in the form of two arrays. The relevant one for our optimization algorithm will be “_ay_onehot”. This array is 2-dimensional. Why?

Working with mini-batches

A big advantage of the weight optimization method we shall use later on during training of our MLP is that we will perform weight adjustment for a whole bunch of training samples in one step. Meaning:

We propagate a whole bunch of test data samples in parallel through the grid to get an array with result data (output array) for all samples.

Such a bunch is called a “batch” and if it is significantly smaller than the whole set of training data – a “mini-batch“.

Working with “mini-batches” during the training and learning phase of an ANN is a compromise between

  • using the full data set (for gradient evaluation) during each training step (“batch approach“)
  • and using just one sample of the training data set for gradient descent and weight corrections (“stochastic approach“).

See chapter 4 of the book of Geron and chapter 2 in the book of Raschka for some thorough information on this topic.

The advantage of mini-batches is that we can use vectorized linear algebra operations over all elements of the batch. Linear Algebra libraries are optimized to perform the resulting vector and matrix operations on modern CPUs and GPUs.

You really should keep the following point in mind to understand the code for the propagation and optimization algorithms discussed in forthcoming articles:
The so called “cost function” will be determined as some peculiar sum over all elements of a batch and the usual evaluation of partial derivatives during gradient descent will be based on matrix operations involving all input elements of a defined batch!

Mini-batches also will help during training in so far as we look at a bunch of multiple selected samples in parallel to achieve bigger steps of our gradient guided descent into an minimum of the cost hyperplane in the beginning – with the disadvantage of making some jumpy stochastic turns on the cost hyperplane instead of a smoother approach.

I probably lost you now 🙂 . The simpler version is: Keep in mind that we later on will work with batches of training data samples in parallel!

However, the separation interface for our categories in the feature space must in the end be adjusted with respect to all given data points of the training set. This means we must perform the training successively for a whole sequence of mini-batches which together cover all available training samples.

What is the shape of the output array?
A single element of the batch is an array of 784 feature values. The corresponding output array is an array with values for 10 categories (here digits). But, what about a whole bunch of test data, i.e. a “batch”?

As I have explained already in my last article
Numpy matrix multiplication for layers of simple feed forward ANNs
the output array for a batch of test data will have the form “_ay_a_Out[i_out_node, idx]” with:

  • i_out_node enumerating the node of the last layer, i.e. a possible category
  • idx enumerating the data sample within a batch of training data

We shall construct the output function such that it provides something like “probability” values within the interval [0,1] for each node of the output layer. We define a perfectly working MLP as one which – after training – produces a “1.0” at the correct category node (i.e. the expected digit) and “0.0” at all other output nodes.

One-hot-encoding of labels
Later on we must compare the real results for the training samples with the expected correct values. To be able to do this we must build up a 2-dim array of the same shape as “_ay_a_out” with correct output values for all test samples of the batch. E.g.: If we expect the digit 7 for the input array of a sample with index idx within the set of training data, we need a 2-dim output array with the element [[0,0,0,0,0,0,0,1,0,0], idx]. The derivation of such an array from a given category label is called “one-hot-encoding“.

By using Numpy’s zero()-function and Pythons “enumerate()”-function we can achieve such an encoding for all data elements of the training data set. See the method “_encode_all_mnist_labels()”. Thus, the array “_ay_onehot” will have a shape of (10, 60000). From this 2-dim array we can later slice out bunches of consecutive test data for mini-batches. The array “_ay_oneval” is provided for convenience and print purposes, only: it provides the expected digit value in addition.

First tests via a Jupyter notebook

Let us test the import of the input data and the label encoding with a Jupyter notebook. In previous articles I have described already how to use such a notebook. I set up a Jupyter notebook called “myANN” (in my present working directory “/projekte/GIT/ai/ml1/mynotebooks”).

I start it with

myself@mytux:/projekte/GIT/ai/ml1> source bin/activate 
(ml1) myself@mytux:/projekte/GIT/ai/ml1> jupyter notebook
[I 15:07:30.953 NotebookApp] Writing notebook server cookie secret to /run/user/21001/jupyter/notebook_cookie_secret
[I 15:07:38.754 NotebookApp] jupyter_tensorboard extension loaded.
[I 15:07:38.754 NotebookApp] Serving notebooks from local directory: /projekte/GIT/ai/ml1
[I 15:07:38.754 NotebookApp] The Jupyter Notebook is running at:
[I 15:07:38.754 NotebookApp] http://localhost:8888/?token=06c2626c8724f65d1e3c4a50457da0d6db414f88a40c7baf
[I 15:07:38.755 NotebookApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).
[C 15:07:38.771 NotebookApp] 

and add two cells. The first one is for the import of libraries.

By the last line I import my present class code. With the second cell I create an instance of my class; the “__init__()”-function is automatically executed ad calls the other methods defined so far:

The output is:

Note that the display of “_ay_onehot” shows the categories in vertical (!) direction (rows) and the index for the input data element in horizontal direction (columns)! You see that the labels in the enumerate structure correspond to the “1”s in the “_ay_onehot”-array.


Importing the MNIST dataset into Numpy arrays via Keras is simple – and has a good performance. We have learned a bit about “one-hot-encoding” and prepared an array “_ay_onehot”, which we shall use during ANN training and weight optimization. It will allow us to calculate a difference between the actual output values of the ANN at the nodes of the output layer and a “1.0” value at the node for the expected sample category and “0.0” otherwise.

In the next article
A simple program for an ANN to cover the Mnist dataset – II
we shall define initial weights for our ANN.

Literature and links

Referenced Books
“Python machine Learning”, Seb. Raschka, 2016, Packt Publishing, Birmingham, UK
“Machine Learning mit Sckit-Learn & TensorFlow”, A. Geron, 2018, O’REILLY, dpunkt.verlag GmbH, Heidelberg, Deutschland

Links regarding cost (or loss) functions and logistic regression
Wikipedia article on Loss functions for classification
uni leipzig on logistic regression

Further articles in this series

A simple Python program for an ANN to cover the MNIST dataset – XIV – cluster detection in feature space
A simple Python program for an ANN to cover the MNIST dataset – XIII – the impact of regularization
A simple Python program for an ANN to cover the MNIST dataset – XII – accuracy evolution, learning
rate, normalization

A simple Python program for an ANN to cover the MNIST dataset – XI – confusion matrix
A simple Python program for an ANN to cover the MNIST dataset – X – mini-batch-shuffling and some more tests
A simple Python program for an ANN to cover the MNIST dataset – IX – First Tests
A simple Python program for an ANN to cover the MNIST dataset – VIII – coding Error Backward Propagation
A simple Python program for an ANN to cover the MNIST dataset – VII – EBP related topics and obstacles
A simple Python program for an ANN to cover the MNIST dataset – VI – the math behind the „error back-propagation“
A simple Python program for an ANN to cover the MNIST dataset – V – coding the loss function
A simple Python program for an ANN to cover the MNIST dataset – IV – the concept of a cost or loss function
A simple Python program for an ANN to cover the MNIST dataset – III – forward propagation
A simple Python program for an ANN to cover the MNIST dataset – II – initial random weight values

Eclipse, PyDev, virtualenv and graphical output of matplotlib on KDE – III

During the last articles of this mini-series

Eclipse, PyDev, virtualenv and graphical output of matplotlib on KDE – I
Eclipse, PyDev, virtualenv and graphical output of matplotlib on KDE – II

we saw how to set up a basic PyDev project in Eclipse which we coupled to a virtual Python environment. We modified the PYTHONPATH and added Python packages to our project with the help of “pip”. In addition we have prepared the PyDev console for interactive Python experiments such that we can use matplotlib in a Qt5 environment. Thus we got a reasonably equipped development environment to start with Python based experiments in “Machine-Learning” [ML] – and build up code much more systematically than just with Jupyter notebooks.

One thing that we expect from an IDE (besides editors and code organization) is a possibility of debugging complicated Python modules and classes. So, in the final article of this series we shall have a brief look at (local) debugging of Python code within PyDEV. To set up a remote debugging server for any clients of our Linux machine is no major problem – but it is beyond the scope of this article.

As with the previous articles experienced Python and PyDev developers won’t learn anything new. My target group is students of ML and people intrested in Python who have some experience with other programming languages.

A “stupid” test example

We need a simple test code example to learn working with the debugging features of PyDev. We open our Eclipse installation and change to the PyDev perspective.

Watch out for the activated symbol at the right upper corner – here a list of different perspectives (which you have used recently) are displayed.

During the last article we had (within the PyDev-Explorer) already created a directory “mytestcode” below our basic project “ml_1”. By default settings of the PyDev environment, actually, we have created not only a directory but a so called Python “package” – which is a collection of modules that have a common context and should be delivered together. (You find more about “modules” and “packages” in any reasonable book on Python).

Now we add a (test) module – i.e. a file with some Python code – to our “package”.

In the next window (not shown we provide a name (without a “.py”-ending), e.g. “stupid” and in the subsequent window we choose a Template of type “<Empty>”. All these actions lead to the creation of a Python file “stupid.py” for which the Python editor of PyDev is opened:

Here we enter some “stupid” code:

test(x=-1, y=-1):
    print("\nhelloworld: let us multiply and divide two POSITIVE INTEGERS < 10")
    if (x < 0):
        a = input("Give an integer value for a: ")
        a = x    
    if (y < 0):
        b = input("Give an integer value for b: ")
        b = y    
    print("a =" + a + ", b = " + b)
    if (a > 10):
    if (b > 10):
    c = int(a) * int(b)
    print("c=" + str(c))
    d = int(a) / int(b)
    print("d = " + str(d))
    return [c, d] 

Of course, you see already traps and lines doomed for failure; some of these traps we want to explore below by debugging :-). Debug As >> Other >> Python Run

By the way, we see a major disadvantage of the Eclipse environment (which does not only affect PyDEV): With “Oxygen” the outline view missed its capability to analyze beyond the first level of the code’s node hierarchy.

Our code does not contain any direct executable commands. We add 2 lines to be able to run it as a “Python Run” at the Python prompt of a console.

res = test()
print("c=" + str(res[0]) + ", d = " + res[1])

After this you may notice that the variable re is now displayed in the “outline view”.

Preparing debugging

Lets try to run our code:
Within the PyDev perspective we open a PyDev console for the interpreter of our virtualenv “ml1” (see the last article about how to do this). There we “import” our “stupid” code and answer the questions by typing in values for the variables a, b and pressing “Enter” each time. We end up with a first error message:

Not unexpected. Ok, lets turn to debugging. Eclipse offers us a special “perspective” which supports debugging. We open it by using the menu “Window >> Perspective >> Open perspective >> Other…“.

In the next window we choose “Debug”:

This leads to a change of the screen layout. The console view shows in the lower part of the screen that we have opened a “Debug console”. (You will see this in some of the following pictures).

There are a lot of new buttons available in the icon bar at the top. One in the middle shows a picture of a “bug” with an arrow besides it: obviously, there are options in what way to debug.

We now first click into our editor window for the “stupid” code and then on the arrow besides the “bug” button; you get something like this:

In my case there is a long list of previous debug runs.
However, in your case the list may be empty as you may never have launched any debug runs before.

As indicated in the screenshot you choose “Debug As >> Other >> Python Run” and click.

Unfortunately, the console view will now display an error message regarding “cython speedups”; this may look similar to the following:

I took the screen shot from another fresh installation – without a virtual Python environment… So the details of the message (especially the path to the interpreter) may look differently.

When I set up my PyDev environment with my virtualenv “ml1” the actual recommendation was to run the command

"/projekte/GIT/ai/ml1/bin/python3.6" "/projekte/eclipse/plugins/org.python.pydev.core_6.4.2.201807021651/pysrc/setup_cython.py" build_ext --inplace

The double quotes around the first 2 parts of the command are important in some command environments! On a Linux command console they do not do any harm.

The reason for this command is that we need to compile and install some additional cython related c-programs to get the “speedups” mentioned in the error message. It is not necessary to understand the details of this in our context.

What you have to do now is to start a Linux terminal window. There you move to the directory of your virtual Python environment – in my case to “/projekte/GIT/ai/ml1”. There you launch the command “source bin/activate” to activate the “virtualenv” with its interpreter.

Then you enter the following commands – Do NOT forget to activate the virtualenv “ml1” with the second command !

myself@mytux:~> cd /projekte/GIT/ai/ml1/
myself@mytux:/projekte/GIT/ai/ml1> source bin/activate
(ml1) myself@mytux:/projekte/GIT/ai/ml1> "/projekte/GIT/ai/ml1/bin/python3.6" "/projekte/eclipse/plugins/org.python.pydev.core_6.4.2.201807021651/pysrc/setup_cython.py" build_ext --inplace
running build_ext
building '_pydevd_bundle.pydevd_cython' extension
creating build
creating build/temp.linux-x86_64-3.6
creating build/temp.linux-x86_64-3.6/_pydevd_bundle
gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -fmessage-length=0 -grecord-gcc-switches -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables -fstack-clash-protection -g -DOPENSSL_LOAD_CONF -fwrapv -fmessage-length=0 -grecord-gcc-switches -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables -fstack-clash-protection -g -fmessage-length=0 -grecord-gcc-switches -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables -fstack-clash-protection -g -fPIC -I/usr/include/python3.6m -c _pydevd_bundle/pydevd_cython.c -o build/temp.linux-x86_64-3.6/_pydevd_bundle/pydevd_cython.o
creating build/lib.linux-x86_64-3.6
creating build/lib.linux-x86_64-3.6/_pydevd_bundle
gcc -pthread -shared -flto -fuse-linker-plugin -ffat-lto-objects -flto-partition=none build/temp.linux-x86_64-3.6/_pydevd_bundle/pydevd_cython.o -L/usr/lib64 -lpython3.6m -o build/lib.linux-x86_64-3.6/_pydevd_bundle/pydevd_cython.cpython-36m-x86_64-linux-gnu.so
copying build/lib.linux-x86_64-3.6/_pydevd_bundle/pydevd_cython.cpython-36m-x86_64-linux-gnu.so -> _pydevd_bundle
running build_ext
building '_pydevd_frame_eval.pydevd_frame_evaluator' extension
creating build/temp.linux-x86_64-3.6/_pydevd_frame_eval
gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -fmessage-length=0 -grecord-gcc-switches -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables -fstack-clash-protection -g -DOPENSSL_LOAD_CONF -fwrapv -fmessage-length=0 -grecord-gcc-switches -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables -fstack-clash-protection -g -fmessage-length=0 -grecord-gcc-switches -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables -fstack-clash-protection -g -fPIC -I/usr/include/python3.6m -c _pydevd_frame_eval/pydevd_frame_evaluator.c -o build/temp.linux-x86_64-3.6/_pydevd_frame_eval/pydevd_frame_evaluator.o
creating build/lib.linux-x86_64-
gcc -pthread -shared -flto -fuse-linker-plugin -ffat-lto-objects -flto-partition=none build/temp.linux-x86_64-3.6/_pydevd_frame_eval/pydevd_frame_evaluator.o -L/usr/lib64 -lpython3.6m -o build/lib.linux-x86_64-3.6/_pydevd_frame_eval/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so                                                             
copying build/lib.linux-x86_64-3.6/_pydevd_frame_eval/pydevd_frame_evaluator.cpython-36m-x86_64-linux-gnu.so -> _pydevd_frame_eval                                            
(ml1) myself@mytux:/projekte/GIT/ai/ml1> 


After this it is reasonable to restart Eclipse via the menu point “File >> Restart“. if you first want to clean up the started Debug runs you could first click on the double red square icon in the console view’s icon bar:

and afterward on the double “x”-icon:

If you looked a bit to the leftmost window “Debug” of the debug perspective during these actions, you would have seen some changes there :-). What you did was to forcefully terminate and remove all launched debug runs from the debug environment!

Then restart by “File >> Restart”. Your Eclipse application should automatically start again – with the perspective and files open that you last worked with.


Now we again click into the editor window for “stupid.py” and move the mouse over the bug-like icon in the top icon bar

and click. Now you should get something like the following – without errors for the debugger:

Our program code has started automatically due to the first of our final two commands (“res=test()”).

Note that the console has two parts – an upper one, where output from the code is shown and a lower part with a Python prompt.
The lower part is for interactive command execution during debugging. We ignore for the time being and only work in the upper part. This is, by the way, the area where the “input” command of Python prompts us to enter values.

There we enter 2 values for our variables a and b and … Of course, we run again into our first error. I have not displayed this as we would learn nothing new from it.

The reason, of course, is: We debugged without having set breakpoints before.

Setting breakpoints is fortunately easy: We move to our PyDev editor view. On the left side we see line numbers (if not: right click on the outer left border stripe of the editor view to get related options!). We right-click on the left side stripe of the line where we want to set a breakpoint to force a stop of the code execution there (i.e. before execution the line’s command).

Hint: Setting and removing breakpoints can also be done by left double-clicking on the leftmost black stripe of the editor.

We add 3 breakpoints for a start – at line 8, 17 and 19 :


Our debugger should be in a state where we can just re-debug the present code. There are 2 possibilities for doing so: In the icon bar of the debug console view we can click on the green “Relaunch”-button:

Or you right-click on the relevant entry on the leftmost view on the last debug runs (which is only one right now) and choose the “Relaunch” option in the context menu:

Code execution stops at our first breakpoint – the correspondent line gets marked in the editor; in the “Variables View” on the upper right we see the values of the variables (x, y) set so far:

How to resume code execution?

This is where the different “player” icons in the top icon menu bar enter the game. You should explore these buttons in detail; for our introduction we just need the green “Resume“-button (and maybe also the red “Terminate“-button):

By clicking on the “Resume”-button code-execution is performed until the next break-point; we enter two integers for a and b again, and the code stops at line 17:

In the variables view we see now that a and b are strings! It is therefore that line 19 will produce an error – as we have learned before. We resume and end at line 19:

Resuming code execution now would again lead to an error. Let us, therefore, change the code, cast a, b to integers and save the modified file. However, this leads to an error message:

The debug environment lost track of our python code file.

Unfortunately, resuming now leads to our old error – despite the file changes:

Whilst debugging as a Python run even reloading the “stupid”-module would not help.

Actually, what we do have to is to relaunch the whole debug process.

Whilst in an un-terminated debug session we can do this by simply pressing the red button with the yellow resume arrow:

Doing so and moving from breakpoint to breakpoint leads again into trouble – however now at a different code line, namely the final one (line 32):

You know,of course, why. But the code of our function test() in between has worked flawlessly now.

Instead of adding another breakpoint or performing the necessary code change for the last line, we try a different way of debugging – namely by linking a standard PyDev console to the debugger.

To clean up our workspace we terminate and remove all running debug runs first. We could use the buttons of the debug console for it as we have learned it above. But this time you could also try and use the leftmost “Debug View” and right-click on the relevant run there to get an option “terminate and Remove”.

Connect a PyDev command console to Debug sessions

We open “Window >> Preferences >> PyDev >> Interactive Console” and activate the checkbox for the point “Connect Console to a debug session?

Then we provoke a new error in our code at line 27 by changing it to:

d = a / int(b)

In addition we remove our last 2 lines

res = test()
print("c=" + str(res[0]) + ", d = " + res[1])

Thereby, importing module “stupid” in a console will NOT cause direct code execution 🙂 .

Let us start a PyDev console for the interpreter of our virtualenv (see the last article if you have forgotten how to do it). This gives us:

As expected we get a PyDev-console with a Python prompt.
The interesting thing, however, is indicated on the left screen side:

The entries in the “Debug View” show that we obviously started some debug run! Actually, our console is now part of a debugging session.

We can type anything at the prompt; we may e.g. define a variable “myvar”. Unfortnately it is not shown in the variables view, but we can get its value at any time when (and if) we have the prompt available :

As we are in an interactive environment we must import our module and do something with the test-function (as no code is executed automatically):

mytestcode import stupid
res = stupid.test()

This brings us to our first break point:

We add a breakpoint to line 27, which will lead to an error – and march from breakpoint to breakpoint until we reach line 27:

Variable “a” is a string and would cause trouble. Let us change the code to “d = int(a) / int(b)” and save the modified file. We again get an error :

pydev debugger: Unable to find module to reload: "stupid".   

If we now resume code execution we will run into the foreseen trouble:

But afterwards we still have an active Python prompt of our Pydev-console for our session. We should be able to reload the code.

For Python 3.6 the following is required to do so:

By using the up-arrow then we move backward through the command history, start “res=stupid.test()” again and move via the break-points to the end – no errors any more (at least for the chosen a- and b-values):

Of course our file “stupid.py” contains a lot more sources for errors. E.g. there is no check or exception handling for a division by zero. And we have not used the function’s parameters, yet. I leave it to the reader to experiment with respective debugging.

How to use the prompt in the normal Debug console?

Now that you know the basics of how to debug, there is one more thing worth mentioning:

When you debug something as a Python Run in the standard “Debug Console”, you may have noticed the (green) Python prompt in the lower part of the console view:

What is it good for?

Well, with it you can interactively change variables and do other things interactively within the context of the debugged object (here of function test()).

Important note:
To avoid inconsistent states you should use the interactive prompt only, when the debugger stopped at a breakpoint. In addition you should enter values for variables requested by “input”-statements only in the upper part of the debug console!

But as soon as the debugger stops you can interactively ask e.g. for values of variables:

As soon as you press Enter in the lower part the
command is reflect in the upper part of the Debug console and results are shown there, too. So, while the debugger stops at a break point, you can do a lot of interactive things; you may set some new variables and use them later on – this is very useful when you want to keep some intermediate results for later purposes during a debug session.

Such new, interactively set variables will, by the way, even be shown in the Variables View (watch out for variables “alp” and “ufo”):

Customization of the debug environment

Eventually, a hint regarding customization of the environment discussed so far.

You should now have a look at all sub-points of the the two following menu points under “Window >> Preferences“:

Of course, there are the sub-points of “PyDev”:
There you can detect a lot more properties of e.g. the “Interpreters” (e.g. of your virtualenv environments) and e.g. the “Interactive Console“.

But Another interesting point is “Run/Debug“:
After a while you may want to use some of the properties there to make life easier during debug sessions. See e.g. for the length of the list of the last debug runs menu point “Run/Debug >> Launching >> Size of recently launched application list”.

Although I cannot discuss it in this article:
Another thing you should become familiar with is the configuration of Debug (and unittest) Runs.
You find a screen for it when you open the combobox of the main debug icon in the top icon bar of Eclipse and click on the point “Debug configurations …” there.

Unit tests

Interactively debugging is nice and useful. Something that is equally important in the long run is, however, “unit testing”. There are two interesting mechanisms that Python provides for doing this: The “doctest”-module and “unittest”. Both are beyond the scope of our introduction to PyDev. You find more about these things on the Internet or in reasonable books on Python.

Note, however, that Python’s “unittest” is especially interesting as the Debug environment of Eclipse/PyDev provides some nice views and tools for it. The following links may give you a first introduction:



In this mini series of articles I have tried to show that setting up Eclipse with PyDev is relatively easy. This gives anybody interested in doing Machine-Learning-experiments with Python, Tensorflow and Keras a reasonable and cost free environment where you can build up and test solid code. A starting point could be module code which you export from a Jupyter notebook after some first trial experiments.

I will provide some examples for a two folded approach with Jupyter notebooks and code refinement via PyDev later on in this

One may ask now, why do we need Jupyter notebooks at all for machine learning experiments. Well, one big advantage of a Jupyter environment is the fact that we can arrange blocks of Python commands in a cell and re-execute the code-blocks in a self-chosen order again. So, it is possible to redo experiments quickly in a modified way; Jupyter gives you – in my opinion – a bit more flexibility in executing preliminary code fragments in a quick and (admittedly) dirty way. In PyDev you would have to work with editors and maybe multiple files to something similar.

Eclipse, PyDev, virtualenv and graphical output of matplotlib on KDE – II

Developing and organizing efficient code after preliminary experiments in machine learning [ML] requires an IDE. This mini-series of articles deals with the setup of a Python environment which supports Eclipse – and Jupyter notebooks. A key ingredient is “virtualenv”: it defines an encapsulated environment for a particular python interpreter together with a specific collection of library packages. In the last article

Eclipse, PyDev, virtualenv and graphical output of matplotlib on KDE – I

we prepared such a virtual Python3 environment “ml1” at a path “/projekte/ai/ml1” and installed some of the basic ML packages there with the help of the “pip3”-mechanism. Within Eclipse we implemented the PyDev plugin. During the setup of a “Python project” we could refer to our “ml1“-environment by defining paths to the Python interpreter and library packages located there.

Changes of the PYTHONPATH from Eclipse/PyDev

To integrate our future own Python modules into interactive experiments we need to add the paths to our own Python file directories into the PYTHONPATH variable. We expect that this should be possible from within Eclipse – and indeed it is on the project level.

In the left Eclipse view of the “PyDev explorer” we add an example directory “mytestcode”; we do this by a right-click on “ml1″ >> New >> folder” and giving the new folder a name in the eventual popup

As soon as the new folder appears we right-click on the root folder of our project “ml_1” in the PyDeev package explorer; in the appearing window we click on “Properties” and get:

There, we choose “PyDev – PYTHONPATH”. By clicking on the button “Add source folder” we can add a folder, e.g. “mytestcode”.

From now on we can import modules in any interactive Python command environment from this directory.

Python console in Eclipse

To perform experiments within an IDE as Eclipse we need some interface to interactively run Python commands and programs. A basic interface for this purpose is a “console”. PyDev, of course, offers a special Python console. How to start it?

If you have chosen a Python perspective within Eclipse you may already see a view area with a console. We start, however, from a perspective where no console view is open, yet:

To add the console view area we use the menu point “Window >> Show View >> Console”.

This gives us:


We got a “Debug console” – not exactly, what we want right now. So, let us open a new console view:

Again a debug console – but we change this now to a PyDev console:

At last, we get a popup where we can choose between a number of defined Python interpreters for command execution. You should at least see 2 items here: A reference to the Linux-system’s Python installation’s interpreter plus a reference to the interpreter configuration of the virtual Python environment, which we had set up in the last article. We had given it the name “python_ml1”.

We chose it; in my case this results in the following view:

Ok, we have a Python prompt (>>>) – but a bunch of error messages, too… The error messages indicate that something to access the graphical environment is missing; PyDEV’s console actually has recognized that it needs an Qt5-based interface to the desktop.

The reason for this is that I had done some customization of the “PyDev” console beforehand; when you look at the choices of “Window >> Preferences” you may find something like this:

here, the setting for “Enable GUI event-loop integration” is interesting: I had chosen the option “PyQt5(qt5)” from the combobox. To me this seemed to be a natural choice on a Qt5-based KDE Linux desktop. Remember, I had the Qt5 python modules installed on my Linux system … Well, error messages nevertheless …

Does the console work at all? Can we use “matplotlib”?

We briefly test whether the Python console works at all:

Yes! And:

We actually do get a reasonable output from “matplotlib”! However, this is NOT based on a “Qt5”-backend, but “TkAgg” (which we can see by the graphical layout of buttons). Where does this come from? And why the complaint of our console about “Qt5”?

Let us try another option from the Combobox : Tkinter(tk).

And then starting yet another console:

Hey, no error messages! This is again a strong indication that some things are missing.

Enable Qt5!

A natural guess is that we need PyQt5 within our virtual environment. Have a look at the Interpreters by choosing
“Window >> Prefrences>> PyDev >> Interpreters >> Python Interpreter”.

There we find no path to the system’s directory for “site-packages”; only the path to thw “ml1”-environments site-package directory is included in the PYTHONPATH. Now, we use “pip” from within Eclipse. This can be done by choosing our “python_ml1” in the upper area and then clicking on “Packages”:

No PyQt5 there – but a button “Install/Uninstall with pip”; we confidently use it:

We terminate all our consoles, we reset the “Interactive console settings” for the GUI event loop integration” (see above) to “PyQt5” and start again a new console for our environment’s “python_ml1” interpreter:

Good! No error messages any more; and:

Yeah, that’s what we want!

Other matplotlib-settings

You should also be aware of the fact that the backend for “matplotlib” may also be defined in a specific configuration file of your environment. In my case we find the relevant file at “/projekte/GIT/ml_1/lib64/python3.6/site-packages/matplotlib/mpl-data/matplotlibrc“.

There you find a commented entry

# backend: :Agg” ,

which you could un-comment and set to a default of “Qt5Agg”. But this is only seldomly required:

Reading the information text in matplotlibrc, we see that Qt5Agg was
automatically chosen as the first working backend of a list of possible backends: MacOSX Qt5Agg Qt4Agg Gtk3Agg TkAgg WxAgg Agg.

By the way this together with the information at https://askubuntu.com/questions/1045720/what-is-a-good-default-backend-for-matplotlibexplains explains why TkAgg worked.

Console colors and command history

Via “Window >&gt: Preferences >> PyDev >> Interactive Console” we can adjust the console colors. I use the following settings to get a dark background:

Command history: The PyDev console, of course, also allows for scrolling through commands but the arrow-up/down-keys. The number of commands can be set via the option “Maximum number of lines to store in global history …”.


A basic Eclipse/PyDev environment which supports a “virtual Python environment (virtualenv) and graphical output in Qt5 quality is set up quickly. We can use it as a tool to collect, rectify and optimize code of experimental Jupyter sessions in Python source files.

In the next article

Eclipse, PyDev, virtualenv and graphical output of matplotlib on KDE – III

we shall have a brief look at debugging local Python code in PyDev.