Develop Reference

Multiclass logistic regression from scratch

I’m trying to apply multiclass logistic regression from scratch. The dataset is the MNIST.
I built some functions such as hypothesis, sigmoid, cost function, cost function derivate, and gradient descendent. My code is below.
I’m struggling with:
As all images are labeled with the respective digit that they represent. There are a total of 10 classes.
Inside the function gradient descendent, I need to loop through each class, but I do not know how to apply it using the One vs All method.
In other words, what I need to do are:
How to filter each class inside the gradient descendent.
After that, how to build a function to predict the test set.
Here is my code.
import numpy as np
import pandas as pd
# Only training data set
# the test data will be load later.
url='https://drive.google.com/file/d/1-MO8oCfq4KU361QeeL4DdafVBhZePUNT/view?usp=sharing'
url='https://drive.google.com/uc?id=' + url.split('/')[-2]
df = pd.read_csv(url,header = None)
X = df.values[:, 0:-1]
y = df.values[:, -1]
m = np.size(X, 0)
y = np.array(y).reshape(m, 1)
X = np.c_[ np.ones(m), X ] # Bias
def hypothesis(X, thetas):
return sigmoid( X.dot(thetas)) #- 0.0000001
def sigmoid(z):
return 1/(1+np.exp(-z))
def losscost(X, y, m, thetas):
h = hypothesis(X, thetas)
return -(1/m) * ( y.dot(np.log(h)) + (1-y).dot(np.log(1-h)) )
def derivativelosscost(X, y, m, thetas):
h = hypothesis(X, thetas)
return (h-y).dot(X)/m
def descendinggradient(X, y, m, epoch, alpha, thetas):
n = np.size(X, 1)
J_historico = []
for i in range(epoch):
for j in range(0,10): # 10 classes
# How to filter each class inside here (inside this def descendinggradient)?
# 2 lines below are wrong.
#thetas = thetas - alpha * derivativelosscost(X, y, m, thetas)
#J_historico = J_historico + [losscost(X, y, m, thetas)]
return [thetas, J_historico]
alpha = 0.01
epoch = 50
(thetas, J_historico) = descendinggradient(X, y, m, epoch, alpha)
# After that, how to build a function to predict the test set.

Let me explain this problem step-by-step:
First since you code doesn't provides the actual data or a link to it I've created a random dataset followed by the same commands you used to create X and Y:
batch_size = 20
num_classes = 10
rng = np.random.default_rng(seed=42)
df = pd.DataFrame(
4* rng.random((batch_size, num_classes + 1)) - 2, # Create Random Array Between -2, 2
columns=['X0','X1','X2','X3','X4','X5','X6','X7','X8', 'X9','Y']
)
X = df.values[:, 0:-1]
y = df.values[:, -1]
m = np.size(X, 0)
y = np.array(y).reshape(m, 1)
X = np.c_[ np.ones(m), X ] # Bias
Next lets take a look at your hypothesis function. If we would just run hypothesis and take a look at the first sample, we will get a vector with the size (10,1). I also needed to provide the initial thetas for this case:
thetas = rng.random((X.shape[1],num_classes))
h = hypothesis(X, thetas)
print(h[0])
>>>[0.89701729 0.90050806 0.98358408 0.81786334 0.96636732 0.97819512
0.89118488 0.87238045 0.70612173 0.30256924]
Basically the function calculates a "propabilties"[1] for each class.
At this point we got to the first issue in your code. The result of the sigmoid function returns "propabilities" which are not "connected" to each other. So to set those "propabilties" in relation we need a another function: SOFTMAX. You will find plenty implementations about this functions. In short: It will calculate the "propabilites" based on the "sigmoid", so that the sum overall class-"propabilites" results to 1.
So for your second question "How to implement a predict after training", we only need to find the argmax value to determine the class:
h = hypothesis(X, thetas)
p = softmax(h) # needs to be implemented
prediction = np.argmax(p, axis=1)
print(prediction)
>>>[2 5 5 8 3 5 2 1 3 5 2 3 8 3 3 9 5 1 1 8]
Now that we know how to predict a class, we also need to know where to setup the training. We want to do this directly after the softmax function. But instead of using the argmax to determine the winning class, we use the costfunction and its derivative. Your problem in your code: You used the crossentropy loss for a binary problem. The binary problem also don't need to use the softmax function, because the sigmoid function already provides the connection of the two binary classes. So since we are not interested in the result at all of the cross-entropy-loss for multiple classes and only into its derivative, we also want to calculate this directly.
The conversion from binary crossentropy to multiclass is kind of unintuitive in the first view. I recommend to read a bit about it before implementing. After this you basicly use your line:
thetas = thetas - alpha * derivativelosscost(X, y, m, thetas)
for updating the thetas.
[1]These are not actuall propabilities, but this is a complete different topic.

Calculating Batch normalization

Part 1
Im going through this article and wanted to try and calculate a forward and backward pass with batch normalization.
When doing the steps after the first layer I get a batch norm output that are equal for all features.
Here is the code (I have on purpose done it in very small steps):
w = np.array([[0.3, 0.4],[0.5,0.1],[0.2,0.3]])
X = np.array([[0.7,0.1],[0.3,0.8],[0.4,0.6]])
def mu(x,axis=0):
return np.mean(x,axis=axis)
def sigma(z, mu):
Ai = np.sum(z,axis=0)
return np.sqrt((1/len(Ai)) * (Ai-mu)**2)
def Ai(z):
return np.sum(z,axis=0)
def norm(Ai,mu,sigma):
return (Ai-mu)/sigma
z1 = np.dot(w1,X.T)
mu1 = mu(z1)
A1 = Ai(z1)
sigma1 = sigma(z1,mu1)
gamma1 = np.ones(len(A1))
beta1 = np.zeros(len(A1))
Ahat = norm(A1,mu1,sigma1) #since gamma is just ones it does change anything here
The output I get from this is:
[1.73205081 1.73205081 1.73205081]
Part 2
In this image:
Should the sigma_mov and mu_mov be set to zero for the first layer?
EDIT: I think I found what I did wrong. In the normalization step I used A1 and not z1. Also I think I found that its normal to use initlize moving average with zeros for mean and ones for variance. Nice if anyone can confirm this.

RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation?

I am using pytorch-1.5 to do some gan test. My code is very simple gan code which just fit the sin(x) function:
import torch
import torch.nn as nn
import numpy as np
import matplotlib.pyplot as plt
# Hyper Parameters
BATCH_SIZE = 64
LR_G = 0.0001
LR_D = 0.0001
N_IDEAS = 5
ART_COMPONENTS = 15
PAINT_POINTS = np.vstack([np.linspace(-1, 1, ART_COMPONENTS) for _ in range(BATCH_SIZE)])
def artist_works(): # painting from the famous artist (real target)
r = 0.02 * np.random.randn(1, ART_COMPONENTS)
paintings = np.sin(PAINT_POINTS * np.pi) + r
paintings = torch.from_numpy(paintings).float()
return paintings
G = nn.Sequential( # Generator
nn.Linear(N_IDEAS, 128), # random ideas (could from normal distribution)
nn.ReLU(),
nn.Linear(128, ART_COMPONENTS), # making a painting from these random ideas
)
D = nn.Sequential( # Discriminator
nn.Linear(ART_COMPONENTS, 128), # receive art work either from the famous artist or a newbie like G
nn.ReLU(),
nn.Linear(128, 1),
nn.Sigmoid(), # tell the probability that the art work is made by artist
)
opt_D = torch.optim.Adam(D.parameters(), lr=LR_D)
opt_G = torch.optim.Adam(G.parameters(), lr=LR_G)
for step in range(10000):
artist_paintings = artist_works() # real painting from artist
G_ideas = torch.randn(BATCH_SIZE, N_IDEAS) # random ideas
G_paintings = G(G_ideas) # fake painting from G (random ideas)
prob_artist0 = D(artist_paintings) # D try to increase this prob
prob_artist1 = D(G_paintings) # D try to reduce this prob
D_loss = - torch.mean(torch.log(prob_artist0) + torch.log(1. - prob_artist1))
G_loss = torch.mean(torch.log(1. - prob_artist1))
opt_D.zero_grad()
D_loss.backward(retain_graph=True) # reusing computational graph
opt_D.step()
opt_G.zero_grad()
G_loss.backward()
opt_G.step()
But when i runing it got this error:
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.FloatTensor [128, 1]], which is output 0 of TBackward, is at version 2; expected version 1 instead. Hint: the backtrace further above shows the operation that failed to compute its gradient. The variable in question was changed in there or anywhere later. Good luck!
Is there something wrong with my code?

This happens because the opt_D.step() modifies the parameters of your discriminator inplace. But these parameters are required to compute the gradient for the generator. You can fix this by changing your code to:
for step in range(10000):
artist_paintings = artist_works() # real painting from artist
G_ideas = torch.randn(BATCH_SIZE, N_IDEAS) # random ideas
G_paintings = G(G_ideas) # fake painting from G (random ideas)
prob_artist1 = D(G_paintings) # G tries to fool D
G_loss = torch.mean(torch.log(1. - prob_artist1))
opt_G.zero_grad()
G_loss.backward()
opt_G.step()
prob_artist0 = D(artist_paintings) # D try to increase this prob
# detach here to make sure we don't backprop in G that was already changed.
prob_artist1 = D(G_paintings.detach()) # D try to reduce this prob
D_loss = - torch.mean(torch.log(prob_artist0) + torch.log(1. - prob_artist1))
opt_D.zero_grad()
D_loss.backward(retain_graph=True) # reusing computational graph
opt_D.step()
You can find more about this issue here https://github.com/pytorch/pytorch/issues/39141

A general reason why it works on 1.4 but give an error in 1.5 is because 'Before 1.5, these tests were not working properly for the optimizers. That’s why you didn’t see any error. But the computed gradients were not correct.'
You can check this link for more discussion about the influence of version: https://discuss.pytorch.org/t/solved-pytorch1-5-runtimeerror-one-of-the-variables-needed-for-gradient-computation-has-been-modified-by-an-inplace-operation/90256/4

I had an similar issue and I solved it by changing my torch version to 1.4 and it worked for me

Modifying the cost function in TensorFlow

I want to modify the following cost function in a way that it adds extra weight to the samples where the prediction is higher than the true output!
cost = tf.reduce_sum(tf.pow(logits-Y, 2))/(2*batch_size)
I found it to be tricky in Tensorflow operations! I want to use Tensorflow operations to do the following codes (written by numpy):
batch_szie = 100
label = np.random.normal(size=batch_szie)
cost = (np.sum(np.power((2*label [label >=0]),2)) + np.sum(np.power((2*label [label <0]),2)))/batch_szie
Please note that the first two lines are just for simulating the label = logits-Y.
Any help/suggestion? Thanks :)

Here I found an answer to this question. However, I think there should be easier and more concise ways.
batch_size = 4
labels = tf.constant ([1,-1,2,1])
pos_index = tf.where(tf.greater_equal(labels, 0))
pos_index = tf.reshape(pos_index, [-1])
pos_label = 5 * tf.gather(labels, pos_index)
neg_index = tf.where(tf.less_equal(labels, 0))
neg_index = tf.reshape(neg_index, [-1])
neg_label = tf.gather(labels, neg_index)
cost = (tf.reduce_sum(tf.pow(pos_label, 2)) + tf.reduce_sum(tf.pow(neg_label, 2)))/(2*batch_size)
with tf.Session() as sess:
print(sess.run(cost))

Backprop implementation issue

What I am supposed to do. I have an black and white image (100x100px):
I am supposed to train a backpropagation neural network with this image. The inputs are x, y coordinates of the image (from 0 to 99) and output is either 1 (white color) or 0 (black color).
Once the network has learned, I would like it to reproduce the image based on its weights and get the closest possible image to the original.
Here is my backprop implementation:
import os
import math
import Image
import random
from random import sample
#------------------------------ class definitions
class Weight:
def __init__(self, fromNeuron, toNeuron):
self.value = random.uniform(-0.5, 0.5)
self.fromNeuron = fromNeuron
self.toNeuron = toNeuron
fromNeuron.outputWeights.append(self)
toNeuron.inputWeights.append(self)
self.delta = 0.0 # delta value, this will accumulate and after each training cycle used to adjust the weight value
def calculateDelta(self, network):
self.delta += self.fromNeuron.value * self.toNeuron.error
class Neuron:
def __init__(self):
self.value = 0.0 # the output
self.idealValue = 0.0 # the ideal output
self.error = 0.0 # error between output and ideal output
self.inputWeights = []
self.outputWeights = []
def activate(self, network):
x = 0.0;
for weight in self.inputWeights:
x += weight.value * weight.fromNeuron.value
# sigmoid function
if x < -320:
self.value = 0
elif x > 320:
self.value = 1
else:
self.value = 1 / (1 + math.exp(-x))
class Layer:
def __init__(self, neurons):
self.neurons = neurons
def activate(self, network):
for neuron in self.neurons:
neuron.activate(network)
class Network:
def __init__(self, layers, learningRate):
self.layers = layers
self.learningRate = learningRate # the rate at which the network learns
self.weights = []
for hiddenNeuron in self.layers[1].neurons:
for inputNeuron in self.layers[0].neurons:
self.weights.append(Weight(inputNeuron, hiddenNeuron))
for outputNeuron in self.layers[2].neurons:
self.weights.append(Weight(hiddenNeuron, outputNeuron))
def setInputs(self, inputs):
self.layers[0].neurons[0].value = float(inputs[0])
self.layers[0].neurons[1].value = float(inputs[1])
def setExpectedOutputs(self, expectedOutputs):
self.layers[2].neurons[0].idealValue = expectedOutputs[0]
def calculateOutputs(self, expectedOutputs):
self.setExpectedOutputs(expectedOutputs)
self.layers[1].activate(self) # activation function for hidden layer
self.layers[2].activate(self) # activation function for output layer
def calculateOutputErrors(self):
for neuron in self.layers[2].neurons:
neuron.error = (neuron.idealValue - neuron.value) * neuron.value * (1 - neuron.value)
def calculateHiddenErrors(self):
for neuron in self.layers[1].neurons:
error = 0.0
for weight in neuron.outputWeights:
error += weight.toNeuron.error * weight.value
neuron.error = error * neuron.value * (1 - neuron.value)
def calculateDeltas(self):
for weight in self.weights:
weight.calculateDelta(self)
def train(self, inputs, expectedOutputs):
self.setInputs(inputs)
self.calculateOutputs(expectedOutputs)
self.calculateOutputErrors()
self.calculateHiddenErrors()
self.calculateDeltas()
def learn(self):
for weight in self.weights:
weight.value += self.learningRate * weight.delta
def calculateSingleOutput(self, inputs):
self.setInputs(inputs)
self.layers[1].activate(self)
self.layers[2].activate(self)
#return round(self.layers[2].neurons[0].value, 0)
return self.layers[2].neurons[0].value
#------------------------------ initialize objects etc
inputLayer = Layer([Neuron() for n in range(2)])
hiddenLayer = Layer([Neuron() for n in range(10)])
outputLayer = Layer([Neuron() for n in range(1)])
learningRate = 0.4
network = Network([inputLayer, hiddenLayer, outputLayer], learningRate)
# let's get the training set
os.chdir("D:/stuff")
image = Image.open("backprop-input.gif")
pixels = image.load()
bbox = image.getbbox()
width = 5#bbox[2] # image width
height = 5#bbox[3] # image height
trainingInputs = []
trainingOutputs = []
b = w = 0
for x in range(0, width):
for y in range(0, height):
if (0, 0, 0, 255) == pixels[x, y]:
color = 0
b += 1
elif (255, 255, 255, 255) == pixels[x, y]:
color = 1
w += 1
trainingInputs.append([float(x), float(y)])
trainingOutputs.append([float(color)])
print "\nOriginal image ... Black:"+str(b)+" White:"+str(w)+"\n"
#------------------------------ let's train
for i in range(500):
for j in range(len(trainingOutputs)):
network.train(trainingInputs[j], trainingOutputs[j])
network.learn()
for w in network.weights:
w.delta = 0.0
#------------------------------ let's check
b = w = 0
for x in range(0, width):
for y in range(0, height):
out = network.calculateSingleOutput([float(x), float(y)])
if 0.0 == round(out):
color = (0, 0, 0, 255)
b += 1
elif 1.0 == round(out):
color = (255, 255, 255, 255)
w += 1
pixels[x, y] = color
#print out
print "\nAfter learning the network thinks ... Black:"+str(b)+" White:"+str(w)+"\n"
Obviously, there is some issue with my implementation. The above code returns:
Original image ... Black:21 White:4
After learning the network thinks ...
Black:25 White:0
It does the same thing if I try to use larger training set (I'm testing just 25 pixels from the image above for testing purposes). It returns that all pixels should be black after learning.
Now, if I use a manual training set like this instead:
trainingInputs = [
[0.0,0.0],
[1.0,0.0],
[2.0,0.0],
[0.0,1.0],
[1.0,1.0],
[2.0,1.0],
[0.0,2.0],
[1.0,2.0],
[2.0,2.0]
]
trainingOutputs = [
[0.0],
[1.0],
[1.0],
[0.0],
[1.0],
[0.0],
[0.0],
[0.0],
[1.0]
]
#------------------------------ let's train
for i in range(500):
for j in range(len(trainingOutputs)):
network.train(trainingInputs[j], trainingOutputs[j])
network.learn()
for w in network.weights:
w.delta = 0.0
#------------------------------ let's check
for inputs in trainingInputs:
print network.calculateSingleOutput(inputs)
The output is for example:
0.0330125791296 # this should be 0, OK
0.953539182136 # this should be 1, OK
0.971854575477 # this should be 1, OK
0.00046146137467 # this should be 0, OK
0.896699762781 # this should be 1, OK
0.112909223162 # this should be 0, OK
0.00034058462280 # this should be 0, OK
0.0929886299643 # this should be 0, OK
0.940489647869 # this should be 1, OK
In other words the network guessed all pixels right (both black and white). Why does it say all pixels should be black if I use actual pixels from the image instead of hard coded training set like the above?
I tried changing the amount of neurons in the hidden layers (up to 100 neurons) with no success.
This is a homework.
This is also a continuation of my previous question about backprop.

It's been a while, but I did get my degree in this stuff, so I think hopefully some of it has stuck.
From what I can tell, you're too deeply overloading your middle layer neurons with the input set. That is, your input set consists of 10,000 discrete input values (100 pix x 100 pix); you're attempting to encode those 10,000 values into 10 neurons. This level of encoding is hard (I suspect it's possible, but certainly hard); at the least, you'd need a LOT of training (more than 500 runs) to get it to reproduce reasonably. Even with 100 neurons for the middle layer, you're looking at a relatively dense compression level going on (100 pixels to 1 neuron).
As to what to do about these problems; well, that's tricky. You can increase your number of middle neurons dramatically, and you'll get a reasonable effect, but of course it'll take a long time to train. However, I think there might be a different solution; if possible, you might consider using polar coordinates instead of cartesian coordinates for the input; quick eyeballing of the input pattern indicates a high level of symmetry, and effectively you'd be looking at a linear pattern with a repeated predictable deformation along the angular coordinate, which it seems would encode nicely in a small number of middle layer neurons.
This stuff is tricky; going for a general solution for pattern encoding (as your original solution does) is very complex, and can usually (even with large numbers of middle layer neurons) require a lot of training passes; on the other hand, some advance heuristic task breakdown and a little bit of problem redefinition (i.e. advance converting from cartesian to polar coordinates) can give good solutions for well defined problem sets. Therein, of course, is the perpetual rub; general solutions are hard to come by, but slightly more specified solutions can be quite nice indeed.
Interesting stuff, in any event!