Solving generalized eigenvalue problem with almost positive definitive matrices?

Question

Assume that we are going to solve:

$$AV = BVD$$

It's the generalized eigenvalue problem. In this case, $A$ and $B$ are symmetrical. But they are not positive definitive because most of its eigenvalues are positive and the rest is negative powered by -18 and more.

Is is still possible to solve this generalized eigenvalue problem, even if $A$ and $B$ is almost positive definitive in practice?

I think -2.2323e-18 can be counted as a zero eigenvalue?

When I solve

$$Sb*V = Sw * V* D$$

I get these eigenvalues. Can we say that $Sw$ and $Sb$ are positive definitive?

The reason why it's 14 positive eigenvalues is because it's 14 different data samples that I compute. So the all eigenvalues is just garbage eigenvalues and I don't need to count them in.

   2.7414e+04
   9.4155e+03
   4.1295e+03
   3.1429e+03
   3.6730e-16
   9.3582e-15
          Inf
          Inf
          Inf
          Inf
          Inf
          Inf
          Inf
          Inf
          Inf
          Inf
   2.0439e-15
  -2.0787e-15
   3.4612e-15
   7.7915e-15
   4.1771e-15
   2.7546e-15
   1.6332e-15
  -1.1891e-15
  -2.4521e-16
   5.3520e-16
   1.1846e-15
   7.4748e-16
   5.2741e-16
   1.5980e-16
   1.6916e-16
  -2.2790e-16
  -4.0109e-16
   1.7067e-17
   1.8431e-16
  -6.6738e-17
  -2.4743e-16
  -1.2387e-16
   1.1155e-16
   6.5867e-17
   1.8417e-16
   7.5477e-17
  -9.3667e-17
  -5.4289e-17
   5.5486e-17
  -4.2705e-17
   6.1318e-18
  -6.8292e-17
  -4.3577e-17
  -1.6952e-17
   4.4937e-17
   9.0261e-17
  -4.4856e-17
  -1.4093e-16
  -1.0877e-17
  -6.1926e-17
   2.8136e-17
  -1.7672e-17
   1.4291e-17
  -1.5689e-17
  -5.2757e-17
  -1.5182e-17
  -5.3647e-18
  -7.8515e-18
  -3.0401e-17
  -5.0176e-17
  -5.6203e-18
   1.3026e-17
  -1.7729e-17
   5.6700e-18
  -3.6998e-18
   2.8389e-17
   5.9863e-17
  -1.9690e-17
  -1.8989e-17
  -3.5953e-17
  -1.0116e-17
  -6.1786e-17
  -5.4509e-17
  -3.6412e-17
  -4.7428e-17
  -7.8577e-18
   4.3425e-17
  -1.7993e-17
  -7.4206e-17
   6.4566e-18
   1.5914e-18
   1.8953e-17
   7.8848e-17
  -1.0132e-17
  -3.1995e-17
   9.0807e-18
  -6.0494e-17
  -2.5929e-17
   5.6330e-17
  -3.6635e-17
   2.0721e-18
   5.2817e-17
   3.2560e-17
   4.3153e-17
   1.0068e-16
   5.6140e-17
   5.4501e-18
  -2.5758e-17
  -1.8913e-17
  -4.2366e-17
   2.6419e-17
   1.0578e-17
  -4.6448e-18
   3.1348e-17
   2.1828e-17
   1.6027e-17
   1.6455e-17
   5.1166e-17
   4.5629e-17
   1.0989e-17
   1.1378e-17
  -2.2593e-17
   7.4640e-18
   1.1364e-17
  -1.1497e-19
  -1.9569e-17
  -3.5304e-17
  -1.1653e-18
  -1.6678e-18
  -9.1828e-18
  -3.7621e-17
  -6.6047e-18
  -9.9503e-18
  -2.8177e-18
   5.0789e-18
   1.7595e-17
  -2.3988e-17
   2.0504e-17
  -2.4144e-18
  -6.7110e-18
   9.4403e-18
   2.7024e-17
   2.0807e-17
  -1.8988e-17
   1.3688e-18
  -1.1622e-17
   1.6796e-17
   4.7089e-18
   1.7252e-17
  -2.7111e-17
   1.1694e-18
   2.5029e-18
  -6.3445e-18
  -2.8082e-17

I have a GNU Octave data file named data that contains $Sw$ and $Sb$

Available for 30 days. https://easyupload.io/4c23se

Do you think I can solve this problem by using regularization on $Sw$ and $Sb$?

$$Sb*V = (Sw + \lambda I) * V* D$$