Questions tagged [deep-learning]

a new area of Machine Learning research concerned with the technologies used for learning hierarchical representations of data, mainly done with deep neural networks (i.e. networks with two or more hidden layers), but also with some sort of Probabilistic Graphical Models.

Filter by
Sorted by
Tagged with
0
votes
1answer
31 views

Why does a simpler model performs better than a complicated one?

This has happened to me, a complicated model couldn't solve the problem when a simpler one solved it in a few epochs. How is that? I believed that a more complicated model means more number of ...
0
votes
0answers
13 views

How to claim that a CNN model is lightweight?

CNN model has some parameters that can show the a model is lightweight compared to others. The parameters can be Size(after training), Trainable parameters or multiply add. Which parameters are ...
0
votes
0answers
14 views

Training tensorflow Resnet on Pets2009 data with multiple bounded boxes [closed]

I am facing a very trivial problem which might be totally based on my lack of knowledge but I am fed up trying multiple ways and failing. I am using Pets2009 dataset. Images are loaded in dataset. ...
-2
votes
0answers
9 views

How to show bounding boxes from xml file? [closed]

I have a dataset, where I have the images folder and an annotation folder which shows the coordinates of the bounding boxes. How can I show the boxes on the real image? I think OpenCV might be helpful....
0
votes
0answers
16 views

How to get the weight matrices of intermediate layers in bidirectional recurrent neural networks?

I am a newbie in deep learning. I have a doubt regarding the training procedure of bidirectional recurrent neural networks using backpropagation through time. Following the original paper for ...
1
vote
0answers
40 views

How can a neural network get unstuck from a local minimum?

Since, there are so many local minimums in so complex neural network function, it is common for a neural network to get stuck on a local minimum. How will the neural network get unstuck from that ...
1
vote
1answer
41 views

LSTM Target Is Also One of It's Inputs?

I have two input arrays that include both historical and forecasted data, and one input array that is only historical. I'm trying to predict (or "forecast") the latter array given the ...
3
votes
2answers
58 views

Troubles Training a Faster R-CNN RPN using a Resnet 101 backbone in Pytorch

Training Problems for a RPN I am trying to train a network for region proposals as in the anchor box-concept from Faster R-CNN on the Pascal VOC 2012 training data. I am using a pretrained Resnet 101 ...
0
votes
0answers
16 views

Brain Tumor Classification(yes/no) - Model performs extremely well on train and validation set but performs badly on test images from the internet

I am trying to do image classification on brain tumor dataset consisting of 400 normal brain mri images and 1.6k brain tumor mri images. In image preprocessing, I am applying histogram equalization ...
0
votes
0answers
9 views

How to fit the LSTM model correctly with series data

I have a small data set with the transformed series vector data with the music genre. The goal for me is to use the LSTM model to learn the non-linear pattern each row for genre classification. The ...
5
votes
2answers
116 views

Why an increasing validation loss and validation accuracy signifies overfitting?

When I train a neural network, I observe an increasing validation loss, while at the same time, the validation accuracy is also increased. I have read explanations related to the phenomenon, and it ...
2
votes
3answers
101 views

Why do we move in the negative direction of the gradient in Gradient Descent?

It is said that backpropagation, with Gradient Descent, seeks to minimize a cost function using the formula: $$ W_{new} = W_{old} - learningRate \cdot \frac{\partial E}{\partial W} $$ My question is, ...
0
votes
0answers
13 views

Which one is better method and why? Manually Handcrafted sound features vs spectrogram + convolution

I am working on classifying different sounds ( not speech or words exactly something like ambulance alarm, police alarm, cough sounds etc) I read few paper which suggested to extract dsp features such ...
1
vote
0answers
16 views

LSTM keras model architecture interpretation

I would appreciate if anyone could correct my interpretation of the LSTM architecture in keras- for example in this simple case ...
0
votes
0answers
22 views

How to balance dataset using fit_generator() in Keras?

I am trying to use keras to fit a CNN model to classify 2 classes of data . I have imbalanced dataset I want to balance my data equally. How I can do that ?? Any help would be appreciated The main ...
0
votes
0answers
26 views

What algorithmic solution would you use for this scenario?

The Project In a Nutshell Use an algorithmic solution to predict with 70%+ accuracy in as close to real-time as possible the increase and decrease of at least three numeric incremental movements for a ...
2
votes
2answers
249 views

The central idea behind SGD

Pr. Hinton in his popular course on Coursera refers to the following fact: Rprop doesn’t really work when we have very large datasets and need to perform mini-batch weights updates. Why it doesn’t ...
0
votes
0answers
12 views

CNN Image resolution vs size/shape

a common technique to get an image to a particular size is by either resizing it completely which can lead to losing the aspect ratio or e.g. resizing the bigger side and then 0-padding the other. My ...
0
votes
0answers
14 views

how to calculate mean average precision of custom object detection algorithms in python

I know that for calculating mean average precision first we must have ground truth files for each image in dataset. I am following this tutorial to detect whether a person has mask on its image or not....
0
votes
0answers
12 views

Google Trax's GRU layer

I am learning about Trax for the implementation of GRU and LSTMs. Their documentation says that a GRU layer in Trax can only accept a number of hidden units equal to the number of elements in the ...
1
vote
0answers
32 views

ML method suggestion for regression task

I'm working with a brushed DC motor current, voltage, calculated rotations and measured rotations and a couple of state signals(direction of rotation and the operation state). My objective is to ...
0
votes
1answer
139 views

Why sparse features should have bigger learning rates associated? And how Adagrad achieves this? [closed]

I was learning about Adagrad optimizer. I came to know that it has a very helpful functionality which is that we can have lower learning rates for the features that are more common and greater ...
2
votes
1answer
41 views

One Neural network with multiple outputs or multiple neural networks with a single output?

I an building a feed forward deep learning model using tabular data. The inputs are numeric features or categorical features (represented with embeddings). The outputs are the same number of numeric ...
-2
votes
0answers
14 views

Choose proper(best) input size for CNN model?

I have different size images in my dataset for training. How we would choose proper(best) input size for a model? What I do is choosing average width and height of all images. Is there better of ...
1
vote
1answer
58 views

Confusion with Notation in the Book on Deep Learning by Ian Goodfellow et al

In chapter 6.1 on 'Example: Learning XOR', the bottom of page 168 mentions: The activation function $g$ is typically chosen to be a function that is applied element-wise, with $h_i = g(x^TW_{:,i}+c_i)...
1
vote
0answers
22 views

Transfer learning with many small datasets

Context I am working on a NLP-model that can classify documents into one of N categories. I have document data from a number of different customers. The document topics are similar across customers ...
0
votes
1answer
51 views

Wouldn't it make more sense to give less importance to gradient far away in past in AdaGrad? [closed]

This is the update equation of a weight by AdaGrad: $$w_{new} = w_{old} - \frac{lr}{\sqrt{G_{}+E}}.G_{w_{old}}$$ Where $G$ is the sum of the gradients of the same weight at previous iterations, $E$ is ...
-2
votes
1answer
51 views

Why do people prefer $(target-actual)^2$ over $|(target-actual)|$ [duplicate]

When computing loss functions, people use $(target-actual)^2$. They sqaure it to prevent any negative loss. But we can even use $|(target-actual)|$ to prevent any negative loss. So, why do people ...
0
votes
1answer
44 views

Why are we taking the square root of the gradient in Adagrad? [closed]

This is how we update weights with Adagrad: $$w_i = w_i - \frac{lr}{\sqrt{g_i+E}}$$ where, $w_i$ is the $i^{th}$ weight, $lr$ is the learning rate, $g_i$ is the gradient of the $i^{th}$ weight at all ...
0
votes
0answers
5 views

Sending model to client side

I have build the model to detect the distance. So i need image of camera to be sent to my server per 100 millisecond. But If this is not cool with low internet speed. So i use js tool and able to ...
0
votes
0answers
25 views

mathematical representation of General Regression Neural Network (GRNN)

I have little knowledge about GRNN but I read about GRNN from many sources on the internet and papers (https://doi.org/10.1109%2F72.97934). \begin{equation*} y=\frac{\sum\limits ^{n}_{K=1} y_{k} .e^{\...
0
votes
0answers
28 views

CNN model to predict if the shops are open or closed [closed]

I'm planning to train a model used to determine if a shop is open. Images are either shot by my students or scraped from the internet. They have manually cropped them so that only one shop is shown on ...
0
votes
1answer
22 views

Text classification using deep learning with removing name, place address etc

I want to prepare a deep learning model for text classification(Document classification), But in my training content many place have name, address, brand name, etc.. which will do confuse to model as ...
1
vote
0answers
8 views

How does Keras.Model() retrieve the different layers between its passed input and output layers?

I have recently been learning Keras and am trying to understand how the keras.Model function is able to interpolate the different layers between passed input and output layers. Do the input or output ...
2
votes
1answer
27 views

How to continuously train a model with a stream of new incoming data

I have limited experience with machine learning, I trained a few networks, but nothing out of the ordinary. I have the following problem but I am not quite sure how to approach it and I'm hoping to ...
1
vote
1answer
29 views

EarlyStopping based on the loss

When training my CNN model, based on the random initialization of weights, i get the prediction results. In other words, with the same training and test data i get different results every time when i ...
0
votes
1answer
18 views

Back Propagation Vs Learning rate in Neuralnet Optimisation

I was doing some research on how backpropagation works? I read that, backpropagation is used to find the optimal weight of each neuron after every iteration using partial derivates and updates the ...
2
votes
1answer
14 views

How to add Earlystopper in Classifier Model

I have designed the following Binary Classifier Neural Network Model for a task. I want to add an early stopper to the model so that the model stops at an epoch where it has stopped learning ...
0
votes
0answers
233 views

Python: ValueError: Unknown layer: Functional

model_path = './models/VGG16_res.h5' model = load_model(model_path) This is the code which I'm using to load a model TensorFlow version: 2.3.0 I'm not sure ...
0
votes
0answers
21 views

Cat2Vec implementation X = categorical and y = categorical

I am trying to convert categorical values (zipcodes) with Cat2Vec into a matrix which can be used as an input shape for categorical prediction of a target with binary values. After reading several ...
-1
votes
1answer
40 views

Why is it valid to remove a constant factor from the derivative of an error function?

I was reading the book 'Make your own neural network' by Tariq Rashid. In his book, he said: (Note - He's talking about normal feed forward neural networks) The $t_k$ is the target value at node $k$, ...
0
votes
1answer
32 views

How to use 5 by 5 or 7 by 7 kernel size for a deep learning network with 3 by 3 kernels?

I am using a U-Net architecture. The visual area of the segmentation mask is very small and after learning it is giving a lot of false positives. I am thinking of ...
-1
votes
0answers
20 views

Effect of label normalization on optimization?

Let's say in a regression task I have a range of labels 1-60. If I normalize the labels and squeeze those into 0-1 range (by dividing 60) and calculate loss then the calculated loss will be very small ...
0
votes
0answers
20 views

Prediction of output as 1 D vectors from multivariate time vector inputs

I have a data set where data in form of time sequences(which is evenly sampled at 1 sec). For each set I have 4 inputs and 1 output. The length of input and output vectors is same. The entire data has ...
0
votes
0answers
9 views

accuracy and loss NAN for keras multi-label Neural network learning [duplicate]

When I ran a Neural Network modeling for multi-class labeling using Keras, the accuracy, loss, val_accuracy, and val_loss all seems to have nan at some point or other during the training process... ...
0
votes
1answer
14 views

What is the meaning of Face Recognition in wild and in static?

What is meant by when someone says face recognition on wild dataset and on static dataset?
1
vote
1answer
21 views

EfficientNet function composition or Hadamard

In the page 3 of the paper of EfficientNet, there is a equation $$\mathcal{N} = \bigodot_{i=1...s} \mathcal{F}_{i}^{L_i} \big(X_{\langle H_i, W_i, C_i \rangle}\big)$$ where $\mathcal{N}$ is the conv ...
0
votes
0answers
7 views

Preparing training data for crop species classification from drone images

I have high resolution multispectral (R,G,B,NIR,RE bands) images of a field taken from MicaSense RedEdge mounted on a drone. There are various species of crops planted. I want to classify the crops or ...
1
vote
0answers
13 views

Does adding a class to a model increase its performance?

I am experimenting with object detection in images via CNN, to be more specific with yolov5 and faster RCNN. The models I use were pre-trained on the coco dataset before I finetune them. My dataset is ...

1
2
3 4 5
69