# Import lsqpy
from lsqpy import Variable,sum_squares,minimize
# Import the test data
from data import x_data,y_data
# Create variables that holds the coefficients
quadratic = Variable()
slope = Variable()
offset = Variable()
# We copy x_data but square the entries
import numpy as np
x_squared = np.power(x_data,2)
# Solve the problem
minimize(sum_squares(offset + x_data*slope + x_squared*quadratic - y_data))
# Create some evenly spaced points for plotting, again replicate powers
t = np.arange(0,5,0.1).reshape(-1,1)
t_squared = np.power(t,2)
# Plot our regressed function
import matplotlib.pyplot as plt
plt.figure(0,(4,4))
plt.plot(x_data,y_data,'ro')
plt.plot(t,offset.value[0,0] + t*slope.value[0,0] + t_squared*quadratic.value[0,0],'b')
plt.xlabel('x')
plt.ylabel('y')
plt.show()
velocity = Variable(2,T) force = Variable(2,T-1) # Create the list of constraints on our variables constraints = [] for i in range(T-1): constraints.append(position[:,i+1] == position[:,i] + h * velocity[:,i]) for i in range(T-1): constraints.append(velocity[:,i+1] == velocity[:,i] + h/mass * force[:,i] - drag*velocity[:,i]) # Add position constraints constraints.append(position[:,0] == 0) constraints.append(position[:,T-1] == final_position) # Add velocity constraints constraints.append(velocity[:,0] == initial_velocity) constraints.append(velocity[:,T-1] == 0) # Solve the problem mu = 1 minimize(mu*sum_squares(velocity)+sum_squares(force),constraints) # Plot the points plt.figure(0,(4,4)) plt.quiver(position.value[0,0:T-1:2],position.value[1,0:T-1:2], force.value[0,::2],force.value[1,::2]) plt.plot(position.value[0,:],position.value[1,:],'r') plt.plot(0,0,'bo') plt.plot(final_position[0],final_position[1],'bo') plt.xlim([-2,12]), plt.ylim([-1,4]) plt.show()
constraints.append(positions[ship][:,0] == 0)
constraints.append(velocities[ship][:,0] == speeds0[:,ship:ship+1])
constraints.append(positions[ship][:,300] == speeds0[:,ship:ship+1])
if ship == 0:
constraints.append(positions[ship][:,formation1_end] == np.array([[20,-20]]).T)
constraints.append(positions[ship][:,400] == np.array([[-20,20]]).T)
constraints.append(positions[ship][:,499] == np.array([[10,-20]]).T)
else:
for t in range(formation1_start,formation1_end):
constraints.append(positions[ship][:,t] - positions[0][:,t] == formation1[:,ship-1:ship])
for t in range(400,500):
constraints.append(positions[ship][:,t] - positions[0][:,t] == formation2[:,ship-1:ship])
objective = sum_squares(forces[0])
for i in range(1,n): objective += sum_squares(forces[i])
minimize(objective,constraints)
# Plot paths
fig = plt.figure(0)
ax = fig.add_subplot(111, autoscale_on=False, xlim=(-25, 25), ylim=(-25, 25))
line, = ax.plot([], [], 'o-', lw=2)
time_template = 'time = %.1fs'
time_text = ax.text(0.05, 0.9, '', transform=ax.transAxes)
def init():
line.set_data([], [])
time_text.set_text('')
return line, time_text
def animate(i):
# Import the test data
from data import x_data, y_data
# Fit a polynomial of this degree to data
num_powers = 20
# Import matplotlib and create the extra variables we need for plotting
# NumPy imported for matrix manipulation
import numpy as np
import matplotlib.pyplot as plt
plt.plot(x_data, y_data, "bo")
t = np.arange(0, 18, 0.1).reshape(-1, 1)
T = np.hstack([np.power(t, i) for i in range(num_powers)])
# We will regress using different powers of x
X = np.hstack([np.power(x_data, i) for i in range(num_powers)])
# Solve the problem
mu = 0
a = Variable(num_powers)
minimize(sum_sq(X * a - y_data) + mu * sum_sq(a))
# Plot our regressed function
plt.plot(t, T.dot(a.value), "r")
plt.xlabel("x")
plt.ylabel("y")
plt.ylim([-5, 5])
plt.show()
"""
A simple linear regression problem
"""
# Import lsqpy
from lsqpy import Variable,sum_squares,minimize
# Import the test data
from data import x_data,y_data
# Solve the problem, and print the result
slope = Variable()
offset = Variable()
minimize(sum_squares(offset + x_data*slope - y_data))
print('slope = '+str(slope.value[0,0])+', offset = '+ str(offset.value[0,0]))
# Print results and plot
import numpy as np
import matplotlib.pyplot as plt
t = np.arange(0, 5.0, 0.1)
plt.figure(0,(4,4))
plt.plot(x_data,y_data,'ro')
plt.plot(t,(slope.value[0,0]*t + offset.value[0,0]))
plt.xlabel('x'), plt.ylabel('y')
plt.show()
from lsqpy import Variable, sum_squares, minimize import matplotlib.pyplot as plt import numpy as np # Import data for the problem from data import temps, n seasonal = Variable(n) eq_constraints = [] for i in range(365,n): eq_constraints.append(seasonal[i] == seasonal[i-365]) smoothing = 0 smooth_objective = sum_squares(seasonal[0:n-1] - seasonal[1:n]) minimize(sum_squares(temps - seasonal) + smoothing*smooth_objective,eq_constraints) residuals = temps - seasonal.value # Generate the residuals matrix matlist = [] ar_len = 5 for i in range(ar_len): matlist.append(residuals[ar_len-i-1:n-i-1]) residuals_mat = np.hstack(matlist) # Solve autoregressive problem ar_coef = Variable(ar_len) minimize(sum_squares(residuals_mat*ar_coef - residuals[ar_len:])) print(ar_coef.value) # Do all plotting plt.figure(0)
from lsqpy import Variable,sum_squares,minimize
import numpy as np
import sys
test_num = int(sys.argv[1])
if test_num == 1:
print('basic problem')
x = Variable(4)
A = np.array([[1,2,3,10],[5,4,6,11],[9,7,8,12]]);
b = np.array([[16,26,36]]).T
minimize(sum_squares(x),[A*x == b])
print(x.value)
elif test_num == 2:
print('another basic problem')
x = Variable(4)
A = np.array([[1,2,3,10],[5,4,6,11],[9,7,8,12]]);
b = np.array([[16,26,36]]).T
objective = sum_squares(A*x+b)+100*sum_squares(x)
const_val = np.array([[100,100]]).T
constraint = const_val == x[2:4]
minimize(objective,[constraint])
print(x.value)
elif test_num == 3:
print('really concise test')
# Generate some data for a small example
A = np.array(range(12)).reshape(4,3)
from lsqpy import Variable, sum_squares, minimize from data import temperatures, num_samples import matplotlib.pyplot as plt import numpy as np yearly_tread = Variable(num_samples) a = Variable() b = Variable() equality_constraints = [] for i in range(12,num_samples): equality_constraints.append(yearly_tread[i] == yearly_tread[i-12]) t = np.array([list(range(0,num_samples))]).T t2 = np.power(t,2) smoothing = 0.0001 objective = sum_squares(yearly_tread + t*b + t2*a - temperatures) objective += smoothing*sum_squares(yearly_tread[0:num_samples-1] - yearly_tread[1:num_samples]) minimize(objective,equality_constraints) plt.figure(0) plt.plot(temperatures) plt.plot(yearly_tread.value + t*b.value + t2*a.value,'r') plt.show() plt.figure(1) plt.plot(yearly_tread.value + t*b.value + t2*a.value - temperatures) plt.show()
""" An example of tomography """ from lsqpy import Variable, sum_squares, minimize from data import img_size, num_pixels, line_mat, line_vals x = Variable(num_pixels) objective = sum_squares(x.__rmul__(line_mat) - line_vals) minimize(objective,method='iterative') import matplotlib.pyplot as plt import matplotlib.cm as cm plt.figure(0) plt.imshow(x.value.reshape(img_size,img_size).T,cmap = cm.Greys_r) plt.show()