23 skills found
git-disl / TOGReal-time object detection is one of the key applications of deep neural networks (DNNs) for real-world mission-critical systems. While DNN-powered object detection systems celebrate many life-enriching opportunities, they also open doors for misuse and abuse. This project presents a suite of adversarial objectness gradient attacks, coined as TOG, which can cause the state-of-the-art deep object detection networks to suffer from untargeted random attacks or even targeted attacks with three types of specificity: (1) object-vanishing, (2) object-fabrication, and (3) object-mislabeling. Apart from tailoring an adversarial perturbation for each input image, we further demonstrate TOG as a universal attack, which trains a single adversarial perturbation that can be generalized to effectively craft an unseen input with a negligible attack time cost. Also, we apply TOG as an adversarial patch attack, a form of physical attacks, showing its ability to optimize a visually confined patch filled with malicious patterns, deceiving well-trained object detectors to misbehave purposefully.
bojone / Gan QpGAN-QP: A Novel GAN Framework without Gradient Vanishing and Lipschitz Constraint
dsgiitr / VisualMLInteractive Visual Machine Learning Demos.
shuheng-liu / Siamese Optimization For Neural NetsA technique I proposed and implemented for training DEEP neural networks on SMALL datasets; effectively avoids overfitting and gradient vanishment/explosion
RahulBhalley / Gan Qp.pytorchUnofficial PyTorch implementation of "GAN-QP: A Novel GAN Framework without Gradient Vanishing and Lipschitz Constraint"
reddyprasade / Machine Learning Interview PreparationPrepare to Technical Skills Here are the essential skills that a Machine Learning Engineer needs, as mentioned Read me files. Within each group are topics that you should be familiar with. Study Tip: Copy and paste this list into a document and save to your computer for easy referral. Computer Science Fundamentals and Programming Topics Data structures: Lists, stacks, queues, strings, hash maps, vectors, matrices, classes & objects, trees, graphs, etc. Algorithms: Recursion, searching, sorting, optimization, dynamic programming, etc. Computability and complexity: P vs. NP, NP-complete problems, big-O notation, approximate algorithms, etc. Computer architecture: Memory, cache, bandwidth, threads & processes, deadlocks, etc. Probability and Statistics Topics Basic probability: Conditional probability, Bayes rule, likelihood, independence, etc. Probabilistic models: Bayes Nets, Markov Decision Processes, Hidden Markov Models, etc. Statistical measures: Mean, median, mode, variance, population parameters vs. sample statistics etc. Proximity and error metrics: Cosine similarity, mean-squared error, Manhattan and Euclidean distance, log-loss, etc. Distributions and random sampling: Uniform, normal, binomial, Poisson, etc. Analysis methods: ANOVA, hypothesis testing, factor analysis, etc. Data Modeling and Evaluation Topics Data preprocessing: Munging/wrangling, transforming, aggregating, etc. Pattern recognition: Correlations, clusters, trends, outliers & anomalies, etc. Dimensionality reduction: Eigenvectors, Principal Component Analysis, etc. Prediction: Classification, regression, sequence prediction, etc.; suitable error/accuracy metrics. Evaluation: Training-testing split, sequential vs. randomized cross-validation, etc. Applying Machine Learning Algorithms and Libraries Topics Models: Parametric vs. nonparametric, decision tree, nearest neighbor, neural net, support vector machine, ensemble of multiple models, etc. Learning procedure: Linear regression, gradient descent, genetic algorithms, bagging, boosting, and other model-specific methods; regularization, hyperparameter tuning, etc. Tradeoffs and gotchas: Relative advantages and disadvantages, bias and variance, overfitting and underfitting, vanishing/exploding gradients, missing data, data leakage, etc. Software Engineering and System Design Topics Software interface: Library calls, REST APIs, data collection endpoints, database queries, etc. User interface: Capturing user inputs & application events, displaying results & visualization, etc. Scalability: Map-reduce, distributed processing, etc. Deployment: Cloud hosting, containers & instances, microservices, etc. Move on to the final lesson of this course to find lots of sample practice questions for each topic!
doans / MoDANetIn the paper, a multi-task deep convolutional neural network, namely MoDANet, is proposed to perform modulation classification and DOA estimation simultaneously. In particular, the network architecture is designed with multiple residual modules, which tackle the vanishing gradient problem. The multi-task learning (MTL) efficiency of MoDANet was evaluated with different variants of Y-shaped connection and fine-tuning some hyper-parameters of the deep network. As a result, MoDANet with one shared residual module using more filters, larger filter size, and longer signal length can achieve better performance of modulation classification and DOA estimation, but those might result in higher computational complexity. Therefore, choosing these parameters to attain a good trade-off between accuracy and computational cost is important, especially for resource-constrained devices. The network is investigated with two typical propagation channel models, including Pedestrian A and Vehicular A, to show the affect of those channels on the efficiency of the network. Remarkably, our work is the first DL-based MTL model to handle two unrelated tasks of modulation classification and DOA estimation. Please cite the papar as:
ronitkathuria15 / Handwritten Prescription RecognitionThe Optical Character Recognition (OCR) system consists of a comprehensive neural network built using Python and TensorFlow that was trained on over 115,000 wordimages from the IAM On-Line Handwriting Database (IAM-OnDB). The neural network consists of 5 Convolutional Neural Network (CNN) layers, 2 Recurrent Neural Network (RNN) Layers, and a final Connectionist Temporal Classification (CTC) layer. As the input image is fed into the CNN layers, a non-linear ReLU function is applied to extract relevant features from the image. The ReLU function is preferred due to the lower likelihood of a vanishing gradient (which arises when network parameters and hyperparameters are not properly set) relative to a sigmoid function. In the case of the RNN layers, the Long Short-Term Memory (LSTM) implementation is used due to its ability to propagate information through long distances. The CTC is given the RNN output matrix and the ground truth text to compute the loss value and the mean of the loss values of the batch elements is used to train the OCR system. This means is fed into an RMSProp optimizer which is focused on minimizing the loss, and it does so in a very robust manner. For inference, the CTC layer decodes the RNN output matrix into the final text. The OCR system reports an accuracy rate of 95.7% for the IAM Test Dataset, but this accuracy falls to 89.4% for unseen handwritten doctors’ prescriptions.
harinisuresh / VanishingGradientA demo of the vanishing gradient problem in a simple fully connected network classifying MNIST images.
ZainUlMustafa / Stock Prediction RNN LSTMStock prediction done using RNN and LTSM to resolve vanishing gradient problem. Dataset used is obtained from Pakistan Stock Exchange
dair-iitd / PoolingAnalysis[EMNLP'20][Findings] Official Repository for the paper "Why and when should you pool? Analyzing Pooling in Recurrent Architectures."
YeongHyeon / Compare Activation FunctionCompare vanishing gradient problem case by case.
sleebapaul / Vanishing GradientsThis is a discussion on an old problem, named Vanishing Gradients, which hindered decades of Deep Learning Research.
grasool / Explore GradientsExplore the problem of vanishing and exploding gradients
immaksudalam / A Comparative Study On Suicidal Ideation Detection Using Machine Learning And Deep Learning ApproachDue to different mental, physical and psychological factors, the tendency of attempting suicide among the people who often feel depressed and lonely is increasing in an alarming rate. Depression is a common mental illness that can interfere with daily activities and lead to suicidal thoughts or attempts. Traditional diagnostic approaches used by mental health specialists can aid in determining a person's level of depression. From study it is notable that, the people with this kind of tendency try to express their feelings through various social media platforms as a text. People likes to post in his/her mother language. So, suicidal sentiment detection from text is needed to be done to prevent suicide by informing their relatives and other law & enforcement authorities. Here, we have tried to figure out a comparative study between machine learning and deep learning algorithms in the study of suicidal sentiment analysis. We have used several Machine learning approaches as well as deep learning algorithms. We also tried hyper-parameter tuning to improve the accuracy of our model, yet we have found the best result in default parameter values. We have also tried to develop a sequential Neural Network Model and Long Short-Term Memory model for the purpose of comparative study. Among all other models, We have got 94% accuracy from SVM model and 93.5% accuracy from Logistic Regression model. In deep learning methodology, sequential recurrent neural network has been used to calculate the value loss. Value loss is almost 3% because of vanishing gradient point and exploding gradient. To reduce the value loss and improve the accuracy we have used long short-term memory. The value loss of LSTM model is less than 1% and the accuracy is secured in 91%.
AFAgarap / Vanishing GradientsAvoiding the vanishing gradients problem by adding random noise and batch normalization
antonior92 / Attractors And Smoothness RNNCompanion code to the 2020 AISTATS paper: ''Beyond exploding and vanishing gradients: analysing RNN training using attractors and smoothness"
jzenn / DSMCSCode of "Resampling Gradients Vanish in Differentiable Sequential Monte Carlo Samplers" (TinyPaper@ICLR'23)
T-Sunm / Vanishing GradientPipeline for solving the Vanishing Gradient problem in Deep Learning. Implements methods such as ReLU, BatchNorm, Skip Connection, and gradient optimization to improve deep model training.
AAA-Zheng / RVSEPPOfficial PyTorch implementation of the paper "Selectively Hard Negative Mining for Alleviating Gradient Vanishing in Image-Text Matching"