Last call to make your voice heard! Our 2022 Developer Survey closes in less than a week. Take survey.

# Tag Info

### Understanding Learning Rate in depth

It depends on the loss function, The loss function of data set 1 might have a different shape than the loss function of a different dataset, for example, they might one might be a simple convex ...

### How do you visualize neural network architectures?

I have found one amazing website. You just need to upload your h5 model, Then you will get a beautiful visualization within a few seconds. Check it out!
• 101

### Why does hyperparameter tuning occur on validation dataset and not at the very beginning?

You perform hyperparameter tuning using train dataset. Validation dataset is used to make sure the model you trained is not overfit. The issue here is that the model has already "seen" the ...

### Deep Learning in a Camera

It is possible to do image/object detection in real time using your camera and specialized libraries like Yolo. See: https://pjreddie.com/darknet/yolo/ However, you will want to train your recognition ...
• 1,335
1 vote

### Deep learning techniques for concept similarity?

You have plenty of existing NLP models already fulfilling this task. For instance, the bert-base-nli-mean-tokens: This model is too general, you will want to use a more adapted one, or build your own ...
• 1,335

### Training neural network for regression with gaussian output layer

You can optimize for the likelihood of the data under the predicted Gaussian distribution. In other words You have a predicted mean (u) and standard deviation (s) for each datapoint (x) Calculate the ...
• 101
1 vote
Accepted

### how to calculate loss function?

As the image says, n represents the number of data points in the batch for which you are currently calculating the loss/performing backpropagation. In the case of batch gradient descent this would be ...
• 5,676

### How to analyse the accuracy and loss graphs of model history?

A basic principle in supervised evaluation is to evaluate on a different data than the training set. This is because the model can overfit, i.e. learn details which happen by chance in the training ...
• 21k

### extract features from low resolution

IIUC, the question is "would blur images affect the features extracted from VGG?" Yes and no, depends on your application (i.e. "what do I want to ...
• 328

### Exploratory data analysis (EDA) on large dataset

You mentioned data is added daily. A lot of this has to do with how your data is structured and if recent data is more important than older data. It might be easier to take a random sample from recent ...

### Why Deep Learning / Neural Networs don't achieve state of the art results in tabular data problems?

In my opinion, deep learning methods are best for (but not only for) representation learning on very generic and homogenous data formats: sound, images, text, videos etc. For most of these formats ...
• 89

### Cat2Vec implementation X = categorical and y = categorical

Cat2Vec only encodes categorically features / X values, cat2Vec does not use the target / Y value.
• 16.3k

### Exploratory data analysis (EDA) on large dataset

Generally it's easier to manipulate a subset of the data, but it's important to take a random sample in order to have a representative sample.
• 21k

### How can i deal with this overfitting?

Here are some proposal. I would need to see the code to be more specific. Did you randomize your data and split to train and validation parts? Have you applied any dropout to your learning process? ...
• 1,335

### Time series prediction using ARIMA vs LSTM

I think, you are misusing MachineLearning & Deep Learning when trying to predict tomorrow... - any statistical Approximations of the chaos can show just averaged Tendency & its borders (...

### Time series prediction using ARIMA vs LSTM

ARIMA gives Trend (or Regression) - can see its slope (that reflects speed of change dx/dy)... LSTM gives MovingAverage - posessing curvature & slope in each moment - that can characterize not ...
1 vote

### Derivative of MSE Cost Function

Any term $f$ that is not a function of $\theta_j$ in any equation will have a partial derivative $\frac{\partial}{\partial\theta_j}(f) = 0$. Importantly, no $x_i$, $y$ or $\theta_{i \ne j}$ depend in ...
• 27.2k
Accepted

### NameError: name 'librosa' is not defined

problem solved by creating a new virtual env and installing all the packages using pip install instead of conda

### CNN not learning properly

Late answer, but I was running into a similar issue. I set the learning rate of my Adam optimizer to a lower value (e.g. 3e-5) and voila! The model started fitting.
• 101

### I am getting (loss: nan - accuracy: 0.0000e+00) for all epochs after training the model

you should one_hot target(y) before model.fit & give it in training in such a form y= tf.one_hot(y,10,) you've got ok, just display results of each batch ...
1 vote

### How to solve MemoryError problem

You could try the following: 1.) Convert to greyscale images instead of RGB if your application does not need RGB. Colored images consume relatively more memory than greyscale ones. 2.) Resize the ...
• 1,241

### What are the differences between Knowledge Graph Embeddings (KGE) and Graph Neural Network (GNN)

Knowledge graph (KG) is a different structure then Graph Neural Network (GNN). Both are indeed graphs but where KG differs is that it is not a Machine learning (ML) model, its just a way to "...
• 135

### What is Typical Variation Normalization?

I suggest going straight to the source and giving the Google paper a read on it (including the TVN paragraph in the appendix), as well as the CORAL paper which underlies it: https://www.biorxiv.org/...

### Why should the data be shuffled for machine learning tasks

Most of the ML algorithms are likely to pick up patterns even in the order you feed data, to avoid picking inadvertently pattern which does not exist, it is important to feed the model with shuffled ...
1 vote

### What to do if the model is not performing well on a validation dataset

Just a few comments: You probably don't need SMOTE if your data is only slightly imbalanced. It can be a source of bias. 30% accuracy is extremely low: there's probably a mistake somewhere, your ...
• 21k

### Using the first 3 layers of a pretrained network in Keras

Defining a new network using a part of a pretained network in Keras is best done layer by layer: ...
• 16.3k

### How to deal with one output for multiple inputs?

If you don't want any relationship between words of different sentences during encoding, you can encode your sentences separately (in this way you don't have that relationship, because each sentence ...
• 68

### What kind of neural network am I using? How can I build a specific kind of network?

Here you have designed a simple ANN architecture. Anyhow if you want to build CNN architecture(Refer https://keras.io/api/layers/convolution_layers/) RNN architecture(Refer https://keras.io/api/layers/...

### How gradients are flown back to Network in siamese architecture? How weights of all CNN models are same even when using different models

You do three forward passes for the three inputs and calculate one loss. So some modules (maybe all) are used three times. As the gradients depend on the inputs, three gradients get calculated and ...
• 101

### NameError: name 'librosa' is not defined

You have imported librosa but haven't specified how to use it. Try importing it as a keyword. For example: import librosa as lib ...
• 1,241

### Class token in ViT and BERT

My question is — why does this token exist as input in all the transformer blocks and is treated the same as the word / patches tokens? The transformers, by default are sequence to sequence networks. ...
Accepted

### Genetic Algorithms (Specifically with Keras)

You have many 1000s of neural network parameters that need to be set up correctly to generate a policy function for your game. In addition, many of the parameters are co-dependent - a "good" ...
• 27.2k

### How to use text as an input for a neural network - regression problem? How many likes/claps an article will get

A key issue in NLP is to encode text into a numerical representation. Embeddings are used for this purpose. word embeddings [...] is a method used when attempting to predict the next word in a text ...

### Genetic Algorithms (Specifically with Keras)

this algorythm is not the best for solving a Snake Game (best proven are RLs), however, I think this can actually work if your model is well built. As far I understand, your knowledge about DGA is ...
Here is a paragraph from Speech and Language Processing Ch 9. to make effective use of these scores, we’ll normalize them with a softmax to create a vector of weights, $\alpha_{ij}$, that indicates ...