#%%
import pandas as pd
import torch
import numpy as np

#%%
df = pd.read_csv('../temp/2body_data.csv').sort_values(by='c')
df['E'] = df['re_E'] + 1j * df['im_E']
train_data = df[df['re_E'] < 0]
target_data = df[df['re_E'] > 0]

train_cs = train_data['c'].to_numpy()
train_Es = torch.tensor(train_data['E'].to_numpy(), dtype=torch.complex128)

#%%
# hyperparameters
N = 9

# initialize random Hamiltonians
H0 = torch.randn(N, N, dtype=torch.complex128)
H0 = (H0 + torch.transpose(H0, 0, 1)).requires_grad_() # symmetric
H1 = torch.randn(N, N, dtype=torch.complex128)
H1 = (H1 + torch.transpose(H1, 0, 1)).requires_grad_() # symmetric

#%%
# training

# generate a set of c values to follow by subdividing the training cs
subdivisions = 3
c_steps = np.concatenate([np.linspace(start, stop, subdivisions, endpoint=False) for (start, stop) in zip(train_cs, train_cs[1:])])
c_steps = np.append(c_steps, train_cs[-1])

lr = 0.05
epochs = 100000
for epoch in range(epochs):
    Es = torch.empty(len(train_data), dtype=torch.complex128)
    current_E = 0.0 # start at the threshold
    for c in c_steps:
        H = H0 + c * H1
        evals = torch.linalg.eigvals(H)
        current_E = evals[torch.argmin(torch.abs(evals - current_E))]
        if np.any(c == train_cs):
            index = np.where(c == train_cs)[0][0]
            Es[index] = current_E

    loss = ((Es - train_Es).abs() ** 2).sum()

    if epoch % 1000 == 0:
        print(f"Training {(epoch+1)/epochs:.1%} \t Loss: {loss}")

    if H0.grad is not None:
        H0.grad.zero_()
    if H1.grad is not None:
        H1.grad.zero_()
    loss.backward()

    with torch.no_grad():
        H0 -= lr * H0.grad
        H1 -= lr * H1.grad

# %%
# evaluate for all points
all_c = torch.tensor(df['c'].values, dtype=torch.float64)
exact_E = torch.tensor(df['E'].values, dtype=torch.complex128)
pred_Es = torch.empty(len(df), dtype=torch.complex128)
with torch.no_grad():
    for (index, (c, E)) in enumerate(zip(all_c, exact_E)):
        H = H0 + c * H1
        evals = torch.linalg.eigvals(H)
        i = torch.argmin(torch.abs(evals - E)) # TODO: more robust way to identify the eigenvector
        pred_Es[index]= evals[i]

# %%
# plot the results
import matplotlib.pyplot as plt
plt.scatter(train_data['re_E'], train_data['im_E'], label='training')
plt.scatter(target_data['re_E'], target_data['im_E'], label='target')
plt.scatter(pred_Es.real, pred_Es.imag, marker='x', label='predicted')
plt.legend()

# %%