# Tag Info

3

Yes, it is not unusual to omit the bias by adding a neuron which always outputs a constant 1, which will then be multiplied by an appropriate weight to give the same formula as you would get using an explicit bias. One notable text using this convention is Understanding Machine Learning: From Theory to Algorithms by Shai Shalev-Shwartz and Shai Ben-David. ...

3

Not necessarily. The neural network (or whatever else you use) is a model of what you are trying to do, and usually models are not able to perfectly model reality, as it is too complex. A noise term is generally used to represent that, ie the imperfection of the model's relationship with the actual world.

2

The short answer is no, you shouldn't do that. There is a "distribution shift" thing when you have different x-y relation on the validation set then on the train set. The distribution shift would deteriorate your model performance and you should try to avoid that. The reason it's bad - ok, you find the way to fix the model for validation data, but ...

2

generally the approach is to have a separate head. For example, imagine you have latent vector $z_k$, you would output two values: $h(z_k)$ and $f(z_k)$ where $0 \leq h \leq 1$ and $b_0 \leq f \leq b_1$ where $b_0$ and $b_1$ are your bounds. In thios setup, during inference you would check $h_k$ and if its greater than some threshold (usually .5), youd ...

2

A function is simply a procedure that maps a particular input to a particular output. You put in $X$, and the function computes $Y$. Those $X$ and $Y$ can take many different forms. It could be mapping one number to another number (convert miles to kilometres), mapping sound to text (name that tune), mapping text to text (translate languages), mapping a ...

2

Can residual connections be beneficial when we have a small training dataset? The usual rule of data science investigations applies here: Try it, measure the results, then you will know. It is very hard to tell, a priori, whether a specific architectural or hyperparameter choice will impact the performance of a neural network on a given problem. In this ...

1

Yes, the functionality should is there. But, don't you think you are overdoing the scales. You have at least 18 scales mentioned here. Too much of anything is bad. There is a reason it likes things divisible by 32 because at that increase in size something more meaningful will show up in the image. Spamming sizes like this won't help you at all, it would ...

1

I commonly use softmax for all 2-class or k-class problems, basically, because I always like to have an output node for each class. For sigmoid, i.e., logistic, you cannot estimate MSE for each sample using the relationship $E_i = \sum_c^C (y_c - \hat{y}_c)^2$, where $C$ is the number of classes, $y_c$ is 0 or 1 for true class membership, and $\hat{y}_c$ is ...

1

Sigmoid is used for binary cases and softmax is its generalized version for multiple classes. But, essentially what they do is over exaggerate the distances between the various values. If you have values on a unit sphere, apply sigmoid or softmax on those values would lead to the points going to the poles of the sphere.

1

Yes, it is possible. What you have shown in case of ANN is what happens in a regression model using NNs. What you have shown in case of RNN is what happens when you are doing sequence-to-sequence translation (like French to English). If you want to get single values like in case of ANN, suppose you are doing regression, then, in the end, you will flatten the ...

1

This seems to be a known problem, and intuitively seems reasonable. You might be interested in the paper Adversarial Training Can Hurt Generalization. The authors suggest that this might be because training on the perturbed data requires the model to learn more robust features, which means more samples are required to obtain performance comparable to a model ...

1

How about a Temporal Convolutional Network? It feels like for such a long sequences having the recurrent/memory based approach is not too feasible. But, intuitively, the 1D convolutions should be able to pick out those rare features from your extremely long sequences. There are also claims that TCNs are comparable to RNNs in performance on common tasks, so ...

1

The authors use so-called embeddings, it's a form to represent the images in some meaningful vector form. The procedure to get embedding as follows. First, keep in mind most of the popular convolutional net architectures starts with convolutional layers and then have few fully connected layers. Then do the following. Train the full network with one-hot ...

1

Do I start off with the epsilon value at the end of the previous session, currently I reinitialize that as well? You should probably re-start with $\epsilon$ at the value you left off at. Using high values of epsilon may cause the neural network to forget some of what it learned from close-to-optimal policies in favour of learning possibly useless values of ...

1

In your case the most probable explanation would be the case of overfitting. The model with too many hidden layers have lots of parameters. By means of all these parameters the model is remembering stuff from the training data itself instead of generalizing by learning the useful patterns. As a rule of thumb if you increase the number of hidden layers more ...

1

You are talking about model parallelism. But, that's not the reason RNNs/LSTMs are not in vogue. Imagine your ability to read the first line of a page and going on reading and still making connections to the first line until the end of the page. Can RNNs/LSTMs do that? No. Can Attention (i.e. Transformers) do it? Yes. The reason is simple Attention is ...

1

That is exactly a neural network works like. Suppose you have a 1000 examples. How you train a network is: First, you divide these 1000 into maybe 100 batches (10 each). After that's done, you feed a batch to the network get its output and compare it with the ground truth, whatever is the error gets backpropagated. Then, for the next batch and then another. ...

1

That equation is just an assumption that we make about the relationship between a response variable (aka dependent variable) $y$ and a predictor (aka independent variable) $x$, i.e. the response variable (target) is an unknown function $f$ of the predictor $x$ plus some noise $\epsilon$ due to e.g. measurement errors (caused e.g. by damaged sensors). So, if ...

1

One way to look at intelligence is it's the way to compress the universe. That means we have a short mental representation of meaningful concepts. For example, if I would say "there is a red swan in your building, it's dangerous and can kill you", you already have concepts of "red", "swan", "danger" and this easy ...

1

I'll try to answer on more general questions Is it ok that model performs better on validation, then on train? It's certainly fine if you use techniques like dropout or data augmentation and the difference is not that big. Because in case of dropout for train you use part of the network, and for validation the whole. I'm suspicious my model is too good. ...

1

Okay - the answer is here https://explained.ai/matrix-calculus/#sec6.2 and it is pretty involved. Basically, there is a difference when you derive the equation for one neuron and when you have to do practically for a set of neurons. The answer is matrix calculus. Here goes from what I could make out. Feel free to correct if I am wrong Gradient Vector/Matrix/...

1

Okay, I think it's better if we distinguish loss and accuracy first via Jeremy's answer, and I agree with him with the sentence "low or huge loss is a subjective metric". The loss value is easy to affect by noise from data and significant increase with a few error data points. My advice in this case is to use more evaluation metrics, and understand ...

1

For standard NNs, their extrapolation behavior an important aspect for financial applications cannot be controlled due to complex functional forms typically involved. Neural Networks with Asymptotics Control discuss how they overcome this significant limitation and develop a new type of neural networks that incorporate large-value asymptotics, when known, ...

Only top voted, non community-wiki answers of a minimum length are eligible