Edoardo Guerriero
  • Member for 1 year, 10 months
  • Last seen this week
Why does the transformer do better than RNN and LSTM in long-range context dependencies?
Accepted answer
34 votes

I'll list some bullet points of the main innovations introduced by transformers , followed by bullet points of the main characteristics of the other architectures you mentioned, so we can then ...

View answer
What are the main differences between skip-gram and continuous bag of words?
17 votes

So as you're probably already aware of, CBOW and Skip-gram are just mirrored versions of each other. CBOW is trained to predict a single word from a fixed window size of context words, whereas Skip-...

View answer
What is the intuition behind the dot product attention?
13 votes

Let's start with a bit of notation and a couple of important clarifications. $\mathbf{Q}$ refers to the query vectors matrix, $q_i$ being a single query vector associated with a single input word. $\...

View answer
What are pros and cons of Bi-LSTM as compared to LSTM?
6 votes

I would say that the logic behind the introduction was more empirical than technical. The only difference between LSTM and Bi-LSTM is the possibility for Bi-LSTM to leverage future context chunks to ...

View answer
Why are reinforcement learning methods sample inefficient?
5 votes

I will try to give a broad answer, if it's not helpful I'll remove it. When we talk about sampling we are actually talking about the number of interaction required to an agent to learn a good model ...

View answer
Why do we need both the validation set and test set?
Accepted answer
5 votes

The difference between the validation and test set in my opinion should be explained in this way: the validation set is meant to be used multiple times. the test set is meant to be used only once. ...

View answer
What are some online courses for deep reinforcement learning?
5 votes

For the programming part I suggest this YouTube channel by Phil Tabor (he also has a website: neuralnet.ai. I found his videos really useful while I was attending reinforcement learning classes at the ...

View answer
How should I read a deep learning paper?
4 votes

Adding something to nbro answer, from my personal experience there are also some hints that can quickly tell you if you're dealing with a good machine learning paper, i.e. worth to read in its ...

View answer
When can I call an entity a hyperparameter?
4 votes

In older machine learning literature the given definition of hyperparameters was explicitly the same used in Bayesian statistics, i.e. a hyperparameter is a parameter of a prior distribution For ...

View answer
Loss function for choosing a subset of objects
4 votes

The choice of the loss function depends primarily on the type of task you're tackling: classification or regression. Your problem is clearly a classification one since you have classes to which a ...

View answer
What are some well-known problems where neural networks don't do very well?
4 votes

I don't know if it might be of use, but many areas of NLP are still hard to tackle, and even if deep models achieve the state of the art results, they usually beat baseline shallow models by very few ...

View answer
How can the input order of pairs into a neural network not matter (i.e. symmetry)?
3 votes

The problem you're describing is related to (if not a subset of) Shift Invariance. Shift invariance refers to any geometric translation of an input, but concatenation of a pair of tenors in 2 ...

View answer
Why do we multipy context_size with embedding_dim? (PyTorch)
Accepted answer
3 votes

An n-gram language model is a language model trained with n context words. This means you're not feeding the model a single word but n. This is why the dimension of the input layer is "...

View answer
In layman terms, what does "attention" do in a transformer?
3 votes

Let's start by stressing out that in the literature unfortunately the term attention is still used widely without any precise consensus around the technical details, the only constant across papers is ...

View answer
How is the convolution layer is usually implemented in practice?
Accepted answer
3 votes

I don't think that to understand convolution you need to dig into the nested code of huge libraries, since the code becomes quickly really hard to understand and convoluted (ba dum tsss!). Joking ...

View answer
How to choosing the random value for parameter w in deep learning network?
Accepted answer
3 votes

Weights initialisation is strictly related to the vanishing/exploding gradient problem. For a complete explanation, please check this awesome page (also from deeplearning.ai). Here I'll summarise the ...

View answer
Is possible to train a robot or AI to prune fruit trees?
3 votes

I am not an expert in robotics (definitely not in pear trees pruning either) but I will try to give some hints to partially answer and also to help reframe the problem a bit. On overall I'll give ...

View answer
What is the type of problem requiring to rate images on a scale?
Accepted answer
3 votes

The main distinction between tasks is 'classification' vs 'regression'. In classification you would try to identify the presence of a cloud or not in an image, if you want to predict the level of '...

View answer
How is the Markovian property consistent in reinforcement learning based scheduling?
Accepted answer
3 votes

Lot of real tasks are in reality not markovian, but it doesn't mean you can't try to train an agent on these tasks. It's like saying "we assume variable x to be normally distributed", you just assume ...

View answer
Is learning rate the only reason for training loss oscillation after few epochs?
2 votes

The loss graph indicates that the model converged to a local minimum, already after a few epochs, and the weights start to oscillate around it. The learning rate is surely responsible for it, but it's ...

View answer
Custom Tensorflow loss function that disincentivizes all black pixels
Accepted answer
2 votes

The background being an unbalance class is a well known problem in image segmentation. Before digging into custom losses you should take a look to existing ones that address this specific issue like ...

View answer
Best practice for handling letterboxed images for non fully-convolutional deep learning networks?
Accepted answer
2 votes

Padding is indeed the easiest solution. And if no bias is used then masking the extra values during the loss computation is also not necessary, since it's enough to use zero as padding value. You ...

View answer
How can I handle overfitting in reinforcement learning problems?
Accepted answer
2 votes

Overfitting refers to a model being stuck in a local minimum while trying to minimise a loss function. In Reinforcement Learning the aim is to learn an optimal policy by maximising (not always but ...

View answer
Why do I need an initial arbitrary policy to implement value iteration algorithm
2 votes

It seems to me that you're thinking about the parameters a and b as being characteristic of the agent that's moving in the environment (therefore determining the final policy), but they are actually a ...

View answer
Which classifier should I use for a dataset with one feature?
Accepted answer
2 votes

It is not really a metter of what model, but if it is possible at all to predict what you're trying to predict. Let's take a similar dataset from kaggle: California Housing Prices This dataset ...

View answer
Should I always start from the same start state in reinforcement learning?
Accepted answer
2 votes

It depends on the task the agent is trying to learn and of course on the environment constrains. In an Atari game agents have a pre-fixed starting point because that's part of the games rules, so I ...

View answer
Are the labels updated during training in the algorithm presented in "An algorithm for correcting mislabeled data"?
Accepted answer
2 votes

I think that making some draws might help. Below I tried to draw the model architecture. We start with classic feed-forward structure: input represented by a vector I with length f (number of ...

View answer
How should I deal with variable input sizes for a neural network classifier?
2 votes

It is much simpler to process the data in a different way. Since you're using temporal data a common practice is to define a priori a minimum time-step, usually called $\textit{granularity}$, which ...

View answer
Are embeddings in multi-lingual language models comparable across languages?
2 votes

Embeddings generated by transformers like Bert or XLM-R are fundamentally different from embeddings learned through language models like GloVe or Word2Vec. The latter are static, i.e. they are just ...

View answer
Does $(\langle w, x \rangle + b) = ||x - x'||$ hold?
1 votes

I think that your missing the fact that $r$ is not a random constant, but precisely the minimum distance we want to ensure between each point and the separation hyperplane (called margin). And indeed ...

View answer