Link to text file Hi I am relatively new to Machine learning, I have managed to get a model like in the image attached, I would like to know what I can do more for the model to fit perfectly[model I made],i don't know much about choosing loss function efficiently,following code was made by adding the text file to an another program made to fit a function to data
(https://i.stack.imgur.com/3ldRy.png)
the test file contains noisy voltage measurements
#!/usr/bin/env python3
#
# Fit function to data
import matplotlib.pyplot as plt
import numpy as np
import jax.numpy as jnp
from jax import grad, jit, vmap, random
# load some noisy data
test = np.loadtxt('newe.txt')
N = 200
sigma = 0.05
x = test[:, 0]
y = test[:, 1]
#plt.plot(x,y)
#plt.show()
# Match function to data
def func(params, x):
# Parameterised damped oscillation
l, omega = params
# Note, we "normalise" parameters
y_pred = jnp.exp(l*10 * x) * jnp.sin(2*jnp.pi* omega*10 * x)
return y_pred
def loss(params, x, y):
# Loss function
y_pred = func(params, x)
return jnp.mean((y - y_pred)**2)
# Compile loss and gradient
c_loss = jit(loss)
d_loss = jit(grad(loss))
# One iteration of gradient descent
def update_params(params, x, y):
grads = d_loss(params, x, y)
params = [param - 0.1 * grad for param, grad in zip (params, grads)]
return params
# Initialise parameters
key = random.PRNGKey(0)
params = [random.normal(key, (1,)), random.normal(key, (1,))]
err = []
for epoch in range(100000):
err.append(c_loss(params, x, y))
params = update_params(params, x, y)
err.append(c_loss(params, x, y))
print("Damping: ", params[0]*10)
print("Frequency:", params[1]*10)
y_pred = func(params, x)
# Plot loss and predictions
f, ax = plt.subplots(1,2)
ax[0].semilogy(err)
ax[0].set_title("History")
ax[1].plot(x, y, label="ground truth")
ax[1].plot(x, y_pred, label="predictions")
ax[1].legend()
plt.show()
