Machine learning in trading: theory, models, practice and algo-trading - page 2252

[mytarmailS]

Maxim Dmitrievsky:

What do you mean, you're kidding me?

Well, in that - what are autogenerators I knew before you))

[Maxim Dmitrievsky]

mytarmailS:

Why is there a neuron at all can you explain the block diagram

[mytarmailS]

Maxim Dmitrievsky:

So I said, why did you decide to replace gmm with a coder?

Why did you think it would work better, you had an idea, I'm interested in the idea, you know?

[Maxim Dmitrievsky]

mytarmailS:

So I said, why did you decide to replace the gmm with a coder?

Why did you think it would work better, you had some idea, I'm interested in the idea, you know?

Because it's the same generative model, but customizable.

it works on quotes as well, it's worse on quotes, i don't know why yet

[mytarmailS]

Maxim Dmitrievsky:

because it is the same generative model, but customizable

it works just as well for quotes, it's worse for quotes, I don't know why yet

Do you have access to the network weights and the possibility to change them?

[Vladimir Perervenko]

Maxim Dmitrievsky:

I expected more from them.

The encoder is a neuron.

you can't understand a thing anyway, but here's the structure of it.

So where's the training? It's all functions and classes.

[Maxim Dmitrievsky]

Vladimir Perervenko:

So where's the training? It's all functions and class.

Do you need a training cycle?

[Vladimir Perervenko]

Maxim Dmitrievsky:

do you need a learning cycle?

Well as far as I understand variation AE differs from usual AE by the fact that during training not the hidden but reparameterized value of it is fed to the decoder. I did not see where it happens.

[Maxim Dmitrievsky]

Vladimir Perervenko:

Well as far as I understand the variation AE differs from the usual AE by the fact that during training the decoder is fed not the hidden but the reparameterized value of it. I did not see where it happens.

        z_mu, z_var = self.encoder(x)

        #  sample from the distribution having latent parameters z_mu, z_var
        #  reparameterize
        std = torch.exp(z_var / 2)
        eps = torch.randn_like(std)
        x_sample = eps.mul(std).add_(z_mu)

[Vladimir Perervenko]

Maxim Dmitrievsky:

Well, let's say (not everything is clear to me in Python's code). And where is the training of this BAE?

Is it in pyTorch ?