Develop Reference

Neural Network Backpropogation code not working

I need to write a simple neural network that consists of 1 output node, one hidden layer of 3 nodes, and 1 input layer (variable size). For now I am just trying to train on the xor data so lets presume that there are 3 input nodes (one node represents the bias and is always 1). The data is labeled 0,1.
I did out the equations for backpropogation and found that despite being so simple, my code does not converge to the xor data being correct.
Let W be the 3x3 matrix of weights connecting the input and hidden layer, and w be the 1x3 matrix that connects the hidden to output layer. Here are some helper functions for my method
def feed_forward_predict(x, W, w):
sigmoid = lambda x: 1/(1+np.exp(-x))
z = np.array(list(map(sigmoid, np.matmul(W, x))))
L = sigmoid(np.matmul(w, z))
return [L, z, x]
this just takes in a value and makes a prediction using the formula sig(w*sig(W*x)). We also have
def calculate_objective(data, labels, W, w):
obj = 0
for point, label in zip(data, labels):
L, z, x = feed_forward_predict(point, W, w)
obj += (label - L)**2
return obj
which calculates the Mean Squared Error for a bunch of given data points. Both of these functions should work as I checked them by hand. Now the problem comes in for the back propogation algorithm
def back_prop(traindata, trainlabels):
sigmoid = lambda x: 1/(1+np.exp(-x))
sigmoid_prime = lambda x: np.exp(-x)/((1+np.exp(-x))**2)
W = np.random.rand(3, len(traindata[0]))
w = np.random.rand(1, 3)
obj = calculate_objective(traindata, trainlabels, W, w)
print(obj)
epochs = 10_000
eta = .01
prevobj = np.inf
i=0
while(i < epochs):
prevobj = obj
dellw = np.zeros((1,3))
for point, label in zip(traindata, trainlabels):
y, z, x = feed_forward_predict(point, W, w)
dellw += 2*(y - label) * sigmoid_prime(np.dot(w, z)) * z
w -= eta * dellw
for point, label in zip(traindata, trainlabels):
y, z, x = feed_forward_predict(point, W, w)
temp = 2 * (y - label) * sigmoid_prime(np.dot(w, z))
# Note that s,u,v represent the hidden node weights. My professor required it this way
dells = temp * w[0][0] * sigmoid_prime(np.matmul(W[0,:], x)) * x
dellu = temp * w[0][1] * sigmoid_prime(np.matmul(W[1,:], x)) * x
dellv = temp * w[0][2] * sigmoid_prime(np.matmul(W[2,:], x)) * x
dellW = np.array([dells, dellu, dellv])
W -= eta*dellW
obj = calculate_objective(traindata, trainlabels, W, w)
i = i + 1
print("i=", i, " Objective=",obj)
return [W, w]
However this code, despite seemingly being correct in terms of the matrix multiplications and derivatives I took, does not converge to anything. In fact the error consistantly bounces: it will fall, then rise, then fall back to the same spot, then rise again. I believe that the problem lies with the W matrix gradient but I do not know what exactly it is.
If you'd like to see for yourself what is happening, the input data I used is
0: 0 0 1
0: 1 1 1
1: 1 0 1
1: 0 1 1
where the first number represents the label. I also set the random seed to np.random.seed(0) just so that I could be consistant with my matrices I'm dealing with.

It appears you are attempting to setup a manual version of stochastic gradient decent with a fixed learning rate (a classic NN problem).
Some notes on your code. It is very difficult to follow all the steps you are doing with so much loops and inconsistencies. In general, it defeats the purpose of using np.array() if you are using loops. Likewise you should know that np.matmul() is * and np.dot() is #. It is unclear how you are using the derivative. You have it explicitly stated at the start for the activation function and then partially derived in the middle of your loop for the MSE. Ugh.
Some other pointers. Explicitly state all your functions and your data, those should be globals. Those should also be derived all at once based on your fixed data as np.array(). In particular, note that while traditional statistics (like finding the line of best fit) means that we are solving for a fixed set of weights given a random variable; in stochastic gradient decent, we are doing the opposite. We are instead fixing the random variable to our data and optimizing our weights. Hence, your functions should only have your weights as "free variables", everything else is fixed. It is important to follow what is being fixed and what is free to update. Your code does not reflect that you know what is being update and what is fixed.
SGD algorithm outline:
Random params.
Update params by moving params a small percentage in the direction of lowest decent.
Run step (2) for a specified amount of time.
Print your params.
Example of SGD code (here is an example of performing SGD to find the line of best fit for some data).
import numpy as np
#Data
X = np.random.random((100,)) #Random points
Y = (2.3*X + 8) + 0.1*np.random.random((100,)) #Linear model + Noise
#Functions (only free variable is the params) (we want the F of best fit under MSE)
F = lambda p : p[0]*X+p[1]
dF = lambda p : np.array([X,np.ones(X.shape)])
MSE = lambda p : (1/Y.shape[0])*((Y-F(p))**2).sum(0)
dMSE = lambda p : (1/Y.shape[0])*(-2*(Y-F(p))*dF(p)).sum(1)
#SGD loop
lr = 0.05
epochs = 1000
params = np.array([0.0,0.0])
for i in range(epochs):
params -= lr*dMSE(params)
print(params)
Hopefully, written this way it is super clear exactly where the subtraction of the gradient is occurring and exactly how it is calculated. Note also, in case it wasn't clear, the derivative in both dF and dMSE is with respect to the params. Obviously this is a toy problem that can be solved explicitly with the scipy module. Hence, SGD is a clearly useless way to optimize two variables.
from scipy.stats import linregress
params = linregress(X,Y)
print(params)

I think I figured it out, in my code I was not summing the hidden node weight derivatives and instead was assigning at every loop iteration. The correct version would be as follow
for point, label in zip(traindata, trainlabels):
y, z, x = feed_forward_predict(point, W, w)
temp = 2 * (y - label) * sigmoid_prime(np.dot(w, z))
# Note that s,u,v represent the hidden node weights. My professor required it this way
dells += temp * w[0][0] * sigmoid_prime(np.matmul(W[0,:], x)) * x
dellu += temp * w[0][1] * sigmoid_prime(np.matmul(W[1,:], x)) * x
dellv += temp * w[0][2] * sigmoid_prime(np.matmul(W[2,:], x)) * x

Difference in result between fmin and fminsearch in Matlab and Python

My objective is to perform an Inverse Laplace Transform on some decay data (NMR T2 decay via CPMG). For that, we were provided with the CONTIN algorithm. This algorithm was adapted to Matlab by Iari-Gabriel Marino, and it works very well. I want to adapt this code into Python. The core of the problem is with scipy.optimize.fmin, which is not minimizing the mean square deviation (MSD) in any way similar to Matlab's fminsearch. The latter results in a good minimization, while the former doesn't.
I have gone through line by line of my adapted code in Python, and the original Matlab. I checked every matrix and every output. I used this to identify that the critical point is in fmin. I also tried scipy.optimize.minimize and other minimization algorithms, but none gave even remotely satisfactory results.
I have made two MWE, for Python and Matlab, to make it reproducible to all. The example data were obtained from the documentation of the matlab function. Apologies if this is long code, but I don't really know how to shorten it without sacrificing readability and clarity. I tried to have the lines match as closely as possible. I am using Python 3.7.3, scipy v1.3.0, numpy 1.16.2, Matlab R2018b, on Windows 8.1. It's a relatively recent Anaconda install (<2 months).
My code:
import numpy as np
from scipy.optimize import fmin
import matplotlib.pyplot as plt
def msd(g, y, A, alpha, R, w, constraints):
""" msd: mean square deviation. This is the function to be minimized by fmin"""
if 'zero_at_extremes' in constraints:
g[0] = 0
g[-1] = 0
if 'g>0' in constraints:
g = np.abs(g)
r = np.diff(g, axis=0, n=2)
yfit = A # g
# Sum of weighted square residuals
VAR = np.sum(w * (y - yfit) ** 2)
# Regularizor
REG = alpha ** 2 * np.sum((r - R # g) ** 2)
# output to be minimized
return VAR + REG
# Objective: match this distribution
g0 = np.array([0, 0, 10.1625, 25.1974, 21.8711, 1.6377, 7.3895, 8.736, 1.4256, 0, 0]).reshape((-1, 1))
s0 = np.logspace(-3, 6, len(g0)).reshape((-1, 1))
t = np.linspace(0.01, 500, 100).reshape((-1, 1))
sM, tM = np.meshgrid(s0, t)
A = np.exp(-tM / sM)
np.random.seed(1)
# Creates data from the initial distribution with some random noise.
data = (A # g0) + 0.07 * np.random.rand(t.size).reshape((-1, 1))
# Parameters and function start
alpha = 1E-2 # regularization parameter
s = np.logspace(-3, 6, 20).reshape((-1, 1)) # x of the ILT
g0 = np.ones(s.size).reshape((-1, 1)) # guess of y of ILT
y = data # noisy data
options = {'maxiter':1e8, 'maxfun':1e8} # for the fmin function
constraints=['g>0', 'zero_at_extremes'] # constraints for the MSD function
R=np.zeros((len(g0) - 2, len(g0)), order='F') # Regularizor
w=np.ones(y.reshape(-1, 1).size).reshape((-1, 1)) # Weights
sM, tM = np.meshgrid(s, t, indexing='xy')
A = np.exp(-tM/sM)
g0 = g0 * y.sum() / (A # g0).sum() # Makes a "better guess" for the distribution, according to algorithm
print('msd of input data:\n', msd(g0, y, A, alpha, R, w, constraints))
for i in range(5): # Just for testing. If this is extremely high, ~1000, it's still bad.
g = fmin(func=msd,
x0 = g0,
args=(y, A, alpha, R, w, constraints),
**options,
disp=True)[:, np.newaxis]
msdfit = msd(g, y, A, alpha, R, w, constraints)
if 'zero_at_extremes' in constraints:
g[0] = 0
g[-1] = 0
if 'g>0' in constraints:
g = np.abs(g)
g0 = g
print('New guess', g)
print('Final msd of g', msdfit)
# Visualize the fit
plt.plot(s, g, label='Initial approximation')
plt.plot(np.logspace(-3, 6, 11), np.array([0, 0, 10.1625, 25.1974, 21.8711, 1.6377, 7.3895, 8.736, 1.4256, 0, 0]), label='Distribution to match')
plt.xscale('log')
plt.legend()
plt.show()
Matlab:
% Objective: match this distribution
g0 = [0 0 10.1625 25.1974 21.8711 1.6377 7.3895 8.736 1.4256 0 0]';
s0 = logspace(-3,6,length(g0))';
t = linspace(0.01,500,100)';
[sM,tM] = meshgrid(s0,t);
A = exp(-tM./sM);
rng(1);
% Creates data from the initial distribution with some random noise.
data = A*g0 + 0.07*rand(size(t));
% Parameters and function start
alpha = 1e-2; % regularization parameter
s = logspace(-3,6,20)'; % x of the ILT
g0 = ones(size(s)); % initial guess of y of ILT
y = data; % noisy data
options = optimset('MaxFunEvals',1e8,'MaxIter',1e8); % constraints for fminsearch
constraints = {'g>0','zero_at_the_extremes'}; % constraints for MSD
R = zeros(length(g0)-2,length(g0));
w = ones(size(y(:)));
[sM,tM] = meshgrid(s,t);
A = exp(-tM./sM);
g0 = g0*sum(y)/sum(A*g0); % Makes a "better guess" for the distribution
disp('msd of input data:')
disp(msd(g0, y, A, alpha, R, w, constraints))
for k = 1:5
[g,msdfit] = fminsearch(#msd,g0,options,y,A,alpha,R,w,constraints);
if ismember('zero_at_the_extremes',constraints)
g(1) = 0;
g(end) = 0;
end
if ismember('g>0',constraints)
g = abs(g);
end
g0 = g;
end
disp('New guess')
disp(g)
disp('Final msd of g')
disp(msdfit)
% Visualize the fit
semilogx(s, g)
hold on
semilogx(logspace(-3,6,11), [0 0 10.1625 25.1974 21.8711 1.6377 7.3895 8.736 1.4256 0 0])
legend('First approximation', 'Distribution to match')
hold off
function out = msd(g,y,A,alpha,R,w,constraints)
% msd: The mean square deviation; this is the function
% that has to be minimized by fminsearch
% Constraints and any 'a priori' knowledge
if ismember('zero_at_the_extremes',constraints)
g(1) = 0;
g(end) = 0;
end
if ismember('g>0',constraints)
g = abs(g); % must be g(i)>=0 for each i
end
r = diff(diff(g(1:end))); % second derivative of g
yfit = A*g;
% Sum of weighted square residuals
VAR = sum(w.*(y-yfit).^2);
% Regularizor
REG = alpha^2 * sum((r-R*g).^2);
% Output to be minimized
out = VAR+REG;
end
Here is the optimization in Python
Here is the optimization in Matlab
I have checked the output of MSD of g0 before starting, and both give the value of 2651. After minimization, Python goes up, to 4547, and Matlab goes down to 0.1381.
I think the problem is one of the following. It's in my implementation, that is, I am using fmin wrong, or there's some other passage I got wrong, but I can't figure out what. The fact the MSD increases when it should have decreased with a minimization function is damning. Reading the documentation, the scipy implementation is different from Matlab's (they use the Nelder Mead method described in Lagarias, per their documentation), while scipy uses the original Nelder Mead). Maybe that affects significantly? Or perhaps my initial guess is too bad for scipy's algorithm?

So, quite a long time since I posted this, but I wanted to share what I ended up learning and doing.
The Inverse Laplace Transform for CPMG data is a bit of a misnomer, and it's more properly called just inversion. The general problem is solving a Fredholm integral of the first kind. One way of doing this is the Tikhonov regularization method. Turns out, you can describe this problem quite easily using numpy, and solve it with a scipy package, so I don't have to "reinvent" the wheel with this.
I used the solution shown in this post, and the names here reflect that solution.
def tikhonov_regularized_inversion(
kernel: np.ndarray, alpha: float, data: np.ndarray
) -> np.ndarray:
data = data.reshape(-1, 1)
I = alpha * np.eye(*kernel.shape)
C = np.concatenate([kernel, I], axis=0)
d = np.concatenate([data, np.zeros_like(data)])
x, _ = nnls(C, d.flatten())
Here, kernel is a matrix containing all the possible exponential decay curves, and my solution judges the contribution of each decay curve in the data I received. First, I stack my data as a column, then pad it with zeros, creating the vector d. I then stack my kernel on top of a diagonal matrix containing the regularization parameter alpha along the diagonal, of the same size as the kernel. Last, I call the convenient nnls, a non negative least square solver in scipy.optimize. This is because there's no reason to have a negative contribution, only no contribution.
This solved my problem, it's quick and convenient.

Logarithm function-approximation algorithm

I created a function to calculate the parameters of a logarithm-function.
My aim is to predict the future results of data points that follow a logarithm function. But what is the most important is that my algorithm fits the last results better than the whole data points as it is the prediction that matters. I currently use Mean Squared Error to optimize my parameters but I do not know how to weight it such as it takes my most recent data points as more important than the first ones.
Here is my equation:
y = C * log( a * x + b )
Here is my code:
import numpy as np
from sklearn.metrics import mean_squared_error
def approximate_log_function(x, y):
C = np.arange(0.01, 1, step = 0.01)
a = np.arange(0.01, 1, step = 0.01)
b = np.arange(0.01, 1, step = 0.01)
min_mse = 9999999999
parameters = [0, 0, 0]
for i in np.array(np.meshgrid(C, a, b)).T.reshape(-1, 3):
y_estimation = i[0] * np.log(i[1] * np.array(x) + i[2])
mse = mean_squared_error(y, y_estimation)
if mse < min_mse:
min_mse = mse
parameters = [i[0], i[1], i[2]]
return (min_mse, parameters)
You can see in the image below the orange curve is the data I have and the blue line is my fitted line. We see that the line stretch a bit away from the line on the end and I would like to avoid that to improve the prediction from my function.
logarithm function graph
My question is twofold:
Is this actually the best way to do it or is it best to use another function (such as the increasing form of an Exponential Decay)? (y = C ( 1 - e-kt ), k > 0)
How can I change my code so that the last values are more important to be fitted than the first ones.

Usually, in non-linear least-squares, the inverse of the y values is taken as weight, that essentially eliminates outliers, you can expand on that idea by adding a function to calculate the weight based on the x position.
def xWeightA(x):
container=[]
for k in range(len(x)):
if k<int(0.9*len(x)):
container.append(1)
else:
container.append(1.2)
return container
def approximate_log_function(x, y):
C = np.arange(0.01, 1, step = 0.01)
a = np.arange(0.01, 1, step = 0.01)
b = np.arange(0.01, 1, step = 0.01)
min_mse = 9999999999
parameters = [0, 0, 0]
LocalWeight=xWeightA(x)
for i in np.array(np.meshgrid(C, a, b)).T.reshape(-1, 3):
y_estimation = LocalWeight*i[0] * np.log(i[1] * np.array(x) + i[2])
mse = mean_squared_error(y, y_estimation)
if mse < min_mse:
min_mse = mse
parameters = [i[0], i[1], i[2]]
return (min_mse, parameters)
Also, it looks like you're evaluating through the complete objective function, that makes the code to take to much time to find the minimum (at least on my machine). You can use curve_fit or polyfit as suggested, but if the goal is to generate the optimizer try adding an early break or a random search through the grid. Hope it helps

LMS batch gradient descent with NumPy

I'm trying to write some very simple LMS batch gradient descent but I believe I'm doing something wrong with the gradient. The ratio between the order of magnitude and the initial values for theta is very different for the elements of theta so either theta[2] doesn't move (e.g. if alpha = 1e-8) or theta[1] shoots off (e.g. if alpha = .01).
import numpy as np
y = np.array([[400], [330], [369], [232], [540]])
x = np.array([[2104,3], [1600,3], [2400,3], [1416,2], [3000,4]])
x = np.concatenate((np.ones((5,1), dtype=np.int), x), axis=1)
theta = np.array([[0.], [.1], [50.]])
alpha = .01
for i in range(1,1000):
h = np.dot(x, theta)
gradient = np.sum((h - y) * x, axis=0, keepdims=True).transpose()
theta -= alpha * gradient
print ((h - y)**2).sum(), theta.squeeze().tolist()

The algorithm as written is completely correct, but without feature scaling, convergence will be extremely slow as one feature will govern the gradient calculation.
You can perform the scaling in various ways; for now, let us just scale the features by their L^1 norms because it's simple
import numpy as np
y = np.array([[400], [330], [369], [232], [540]])
x_orig = np.array([[2104,3], [1600,3], [2400,3], [1416,2], [3000,4]])
x_orig = np.concatenate((np.ones((5,1), dtype=np.int), x_orig), axis=1)
x_norm = np.sum(x_orig, axis=0)
x = x_orig / x_norm
That is, the sum of every column in x is 1. If you want to retain your good guess at the correct parameters, those have to be scaled accordingly.
theta = (x_norm*[0., .1, 50.]).reshape(3, 1)
With this, we may proceed as you did in your original post, where again you will have to play around with the learning rate until you find a sweet spot.
alpha = .1
for i in range(1, 100000):
h = np.dot(x, theta)
gradient = np.sum((h - y) * x, axis=0, keepdims=True).transpose()
theta -= alpha * gradient
Let's see what we get now that we've found something that seems to converge. Again, your parameters will have to be scaled to relate to the original unscaled features.
print (((h - y)**2).sum(), theta.squeeze()/x_norm)
# Prints 1444.14443271 [ -7.04344646e+01 6.38435468e-02 1.03435881e+02]
At this point, let's cheat and check our results
theta, error, _, _ = np.linalg.lstsq(x_orig, y)
print(error, theta)
# Prints [ 1444.1444327] [[ -7.04346018e+01]
# [ 6.38433756e-02]
# [ 1.03436047e+02]]
A general introductory reference on feature scaling is this Stanford lecture.

Specify the shift for numpy.correlate

I wonder if there is a possibility to specify the shift expressed by k variable for the cross-correlation of two 1D arrays. Because with the numpy.correlate function and its mode parameter set to 'full' I will get cross-correlate coefficients for each k shift for whole length of the taken array (assuming that both arrays are the same size). Let me show you what I mean exactly on below example:
import numpy as np
# Define signal 1.
signal_1 = np.array([1, 2 ,3])
# Define signal 2.
signal_2 = np.array([1, 2, 3])
# Other definitions.
Xi = signal_1
Yi = signal_2
N = np.size(Xi)
k = 3
Xs = np.average(Xi)
Ys = np.average(Yi)
# Cross-covariance coefficient function.
def crossCovariance(Xi, Yi, N, k, Xs, Ys, forCorrelation = False):
autoCov = 0
for i in np.arange(0, N-k):
autoCov += ((Xi[i+k])-Xs)*(Yi[i]-Ys)
if forCorrelation == True:
return autoCov/N
else:
return (1/(N-1))*autoCov
# Expected value function.
def E(X, P):
expectedValue = 0
for i in np.arange(0, np.size(X)):
expectedValue += X[i] * (P[i] / np.size(X))
return expectedValue
# Cross-correlation coefficient function.
def crossCorrelation(Xi, Yi, k):
# Calculate the covariance coefficient.
cov = crossCovariance(Xi, Yi, N, k, Xs, Ys, forCorrelation = True)
# Calculate standard deviations.
EX = E(Xi, np.ones(np.size(Xi)))
SDX = (E((Xi - EX) ** 2, np.ones(np.size(Xi)))) ** (1/2)
EY = E(Yi, np.ones(np.size(Yi)))
SDY = (E((Yi - EY) ** 2, np.ones(np.size(Yi)))) ** (1/2)
# Calculate correlation coefficient.
return cov / (SDX * SDY)
# Express cross-covariance or cross-correlation function in a form of a 1D vector.
def array(k, norm = True):
# If norm = True, return array of autocorrelation coefficients.
# If norm = False, return array of autocovariance coefficients.
vector = np.array([])
shifts = np.abs(np.arange(-k, k+1, 1))
for i in shifts:
if norm == True:
vector = np.append(crossCorrelation(Xi, Yi, i), vector)
else:
vector = np.append(crossCovariance(Xi, Yi, N, i, Xs, Ys), vector)
return vector
In my example, calling the method array(k, norm = True) for different values of k will give resuslt as I shown below:
k = 3, [ 0. -0.5 0. 1. 0. -0.5 0. ]
k = 2, [-0.5 0. 1. 0. -0.5]
k = 1, [ 0. 1. 0.]
k = 0, [ 1.]
My approach is good for the learning purposes but I need to move to the native numpy functions in order to speed up my analysis. How one could specify the k shift value while using the native numpy.correlate function? PS k parameter specify the "time" shift between two arrays. Thank you in advance.

Whilst I'm not aware of any built-in function for computing the cross-correlation for a particular range of signal lags, you can speed your version up a lot by vectorization, i.e. performing operations on arrays rather than single elements in an array.
This version uses only a single Python loop over the lags:
import numpy as np
def xcorr(x, y, k, normalize=True):
n = x.shape[0]
# initialize the output array
out = np.empty((2 * k) + 1, dtype=np.double)
lags = np.arange(-k, k + 1)
# pre-compute E(x), E(y)
mu_x = x.mean()
mu_y = y.mean()
# loop over lags
for ii, lag in enumerate(lags):
# use slice indexing to get 'shifted' views of the two input signals
if lag < 0:
xi = x[:lag]
yi = y[-lag:]
elif lag > 0:
xi = x[:-lag]
yi = y[lag:]
else:
xi = x
yi = y
# x - mu_x; y - mu_y
xdiff = xi - mu_x
ydiff = yi - mu_y
# E[(x - mu_x) * (y - mu_y)]
out[ii] = xdiff.dot(ydiff) / n
# NB: xdiff.dot(ydiff) == (xdiff * ydiff).sum()
if normalize:
# E[(x - mu_x) * (y - mu_y)] / (sigma_x * sigma_y)
out /= np.std(x) * np.std(y)
return lags, out
Some more general points of advice:
As I mentioned in the comments, you should try to give your functions names that are informative, and that aren't likely to conflict with other things in your namespace (e.g. array vs np.array).
It's much better to make your functions self-contained. In your version, N, k, Xs and Ys are defined outside the main function. In this situation you might accidentally modify or overwrite one of these variables, and it can get tricky to debug errors caused by this sort of thing.
Appending to numpy arrays (e.g. using np.append or np.concatenate) is slow, so avoid it whenever you can. If, as in this case, you know the size of the output ahead of time, it's much faster to pre-allocate the output array (e.g. using np.empty or np.zeros), then fill in the elements. If you absolutely have to do concatenation, it's often faster to append to a normal Python list, then convert it to a numpy array at the end.

It's available by specifying maxlags:
import matplotlib.pyplot as plt
xcorr = plt.xcorr(signal_1, signal_2, maxlags=1)
Documentation can be found here. This implementation is based on np.correlate.