Downloads & Free Reading Options - Results

Francois Chollet is the creator of Keras, which is an open source deep learning library that is designed to enable fast, user-friendly experimentation with deep neural networks. It serves as an interface to several deep learning libraries, most popular of which is TensorFlow, and it was integrated into TensorFlow main codebase a while back. Aside from creating an exceptionally useful and popular library, Francois is also a world-class AI researcher and software engineer at Google, and is definitely an outspoken, if not controversial, personality in the AI world, especially in the realm of ideas around the future of artificial intelligence.

Datasets for Topological Deep Learning for Enhanced Diagnosis of Blood Disorders

Datasets for Topological Deep Learning for Enhanced Diagnosis of Blood Disorders

Want to master the most powerful deep learning framework in data science? Discover why PyTorch is the preferred choice for AI researchers, developers, and data scientists worldwide. In this video, we break down how PyTorch works, its key features, advantages, and how it compares with other frameworks like TensorFlow and Scikit-learn. Learn how it powers applications in computer vision, NLP, reinforcement learning, generative AI, and more. Ready to upskill with PyTorch? Explore top data science certifications at USDSI® and accelerate your career today!  http://bit.ly/44yq6lg https://youtu.be/fKz3fxM-fk4

This project aims to conduct a scoping review in accordance with the JBI methodology to map and evaluate the types, effectiveness, and implementation challenges of fall prevention and detection technologies in construction. The review will explore AI-powered systems, deep learning models, wearable sensors, and BIM-integrated hazard detection, with a focus on real-time safety applications. It will also examine key barriers, enablers, and performance outcomes such as model accuracy, site feasibility, and worker acceptance.

Understand how Generative Engine Optimization (GEO) helps your brand get cited by AI tools and increase organic traffic in t oday’s AI-powered search landscape.

https://media.ccc.de/v/35c3-9386-introduction_to_deep_learning This talk will teach you the fundamentals of machine learning and give you a sneak peek into the internals of the mystical black box. You'll see how crazy powerful neural networks can be and understand why they sometimes fail horribly. Computers that are able to learn on their own. It might have sounded like science-fiction just a decade ago, but we're getting closer and closer with recent advancements in Deep Learning. Or are we? In this talk, I'll explain the fundamentals of machine-learning in an understandable and entertaining way. I'll also introduce the basic concepts of deep learning. With the current hype of deep learning and giant tech companies spending billions on research, understanding how those methods works, knowing the challenges and limitations is key to seeing the facts behind the often exaggerated headlines. One of the most common applications of deep learning is the interpretation of images, a field that has been transformed significantly in recent years. Applying neural networks to image data helps visualising and understanding many of the faults as well as advantages of machine learning in general. As a research scientist in the field of automated analysis of bio-medical image data, I can give you some insights into these as well as some real-world applications. teubi https://fahrplan.events.ccc.de/congress/2018/Fahrplan/events/9386.html Source: https://www.youtube.com/watch?v=-EGVhkBC8AY Uploader: media.ccc.de

HCPS TV dropped by Deep Run High School for a quick look into Sarah Oliver's AP Research class.

Traditional brain tumor classification methods face limitations regarding time consuming manual analysis and the potential for inaccuracies. The study proposes a novel approach for brain tumor classification using a hybrid deep learning model incorporating U Net and CNNs. The model will be trained on a large dataset of labeled MRI images to learn features and classify tumors automatically. This approach promises faster diagnoses, improved accuracy compared to traditional methods, and reduced workload for radiologists. The research aims to explore the application of other imaging modalities and expand the scope of the tumor detection system. Dr. M. V. Vijaya Saradhi | Gali Lahari Reddy | Ale Laxminarayana | Alladi Yuvaraj Kumar | Tanishq Duddi "Design and Development of Brain Tumor Classification using Hybrid Deep Learning Algorithm" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-9 | Issue-1 , February 2025, URL: https://www.ijtsrd.com/papers/ijtsrd75030.pdf Paper Url: https://www.ijtsrd.com/computer-science/other/75030/design-and-development-of-brain-tumor-classification-using-hybrid-deep-learning-algorithm/dr-m-v-vijaya-saradhi

Introduction Maize is one of the most important cereal crops worldwide, providing staple food for people globally. Counting maize tassels provides essential information about yield prediction, growth status, and plant phenotyping, but traditional manual approaches are expensive and time-consuming. Recent developments in technology, including high-resolution RGB imagery acquired by unmanned aerial vehicles (UAVs) and advanced machine-learning techniques such as deep learning (DL), have been used to analyze genotypes, phenotypes, and crops. In this study, we modified the YOLOv5s single-stage object detection technique based on a deep convolutional neural network and named it MYOLOv5s. We incorporated BottleneckCSP structures, Hardswish activation function, and two-dimensional spatial dropout layers to increase tassel detection accuracy and reduce overfitting. Our method's performance was compared with three state-of-the-art algorithms: Tasselnetv2+, RetinaNet, and Faster R-CNN. The results obtained from our proposed method demonstrate the effectiveness of MYOLOv5s in detecting and counting maize tassels.  Materials and Methods The High-Intensity Phenotyping Site (HIPS) dataset was collected from the large field at the Agronomy Center for Research and Education (ACRE) of Purdue University, located in West Lafayette, Indiana, USA during the 2020 growing season. A Sony Alpha 7R-III RGB camera mounted on a UAV at a 20m altitude captured high-resolution orthophotos with a pixel resolution of 0.25 cm. The dataset consisted of two replications of 22 entries each for hybrids and inbreds, planted on May 12 using a two-row segment plot layout with a plant population of 30,000 per acre. The hybrids and inbreds in this dataset had varying flowering dates, ranging from 20 days between the first and last variety. This article uses orthophotos taken on July 20th and 24th to train and test the proposed deep network "MYOLOv5s." These orthophotos were divided into 15 images (3670×2150) and then cropped to obtain 150 images (608 × 2048) for each date. Three modifications were applied to the original YOLOv5s to form MYOLOv5s: BottleneckCSP structures were added to the neck part of the YOLOv5s, replacing some C3 modules; two-dimensional spatial dropout layers were used in the defect layer; and the Hardswish activation function was utilized in the convolution structures. These modifications improved tassel detection accuracy. MYOLOv5s was implemented in the Pytorch framework, and the Adam algorithm was applied to optimize it. Hyper-parameters such as the number of epochs, batch size, and learning rates were also optimized to increase tassel detection accuracy. Results and Discussion In this study, we first compared the original and modified YOLOv5s techniques, and our results show that MYOLOv5s improved tassel detection accuracy by approximately 2.80%. We then compared MYOLOv5s performance to the counting-based approach TasselNetv2+ and two detection-based techniques: Faster R-CNN and RetinaNet. Our results demonstrated the superiority of MYOLOv5s in terms of both accuracy and inference time. The proposed method achieved an AP value of 95.30% and an RMSE of 1.9% at 84 FPS, making it about 1.4 times faster than the other techniques. Additionally, MYOLOv5s correctly detected the highest number of maize tassels and showed at least a 17.64% improvement in AP value compared to Faster R-CNN and RetinaNet, respectively. Furthermore, our technique had the lowest false positive and false negative values. The regression plots show that MYOLOv5s provided slightly higher fidelity counts than other methods. Finally, we investigated the effect of score values on the performance of detection-based models and calculated the optimal values of hyperparameters. Conclusion The MYOLOv5s technique outperformed other state-of-the-art models in detecting maize tassels, achieving the highest precision, recall, and average precision (AP) values. The MYOLOv5s method had the lowest root mean square error (RMSE) value in the error counting metric, demonstrating its accuracy in detecting and counting maize tassels. We evaluated the correlation between predicted and ground-truth values of maize tassels using the R2 score, and for the MYOLOv5s method, the R2 score was approximately 99.28%, indicating a strong correlation between predicted and actual values. The MYOLOv5s method performed exceptionally well in detecting tassels, even in highly overlapping areas. It accurately distinguished and detected tassels, regardless of their proximity or overlap with other objects. When compared to the counting-based approach TasselNetv2+, our proposed MYOLOv5s method showed faster inference times. This suggests that the MYOLOv5s method is computationally efficient while maintaining accurate tassel detection capabilities.

Ever Wondered How AI Recognizes Faces, Objects & Text? 🤖📸 This high-tech visualization showcases Deep Learning in Computer Vision , where AI detects objects, faces, and text using glowing bounding boxes. From security systems to self-driving cars, AI-powered image recognition is transforming industries! 🚀 🔍 How AI Image Recognition Works: ✅ Deep Learning Algorithms – Training neural networks to "see" patterns 🧠💡 ✅ Object Detection – Identifying cars, animals, and everyday items 🚗🐶 ✅ Facial Recognition – Unlocking devices & improving security 🔐 ✅ Text Detection – AI reading signs, documents & handwritten notes 📝 💡 Want to Master AI & Computer Vision? Learn from top industry experts at the Best Data Science Course Institute in Delhi! For more information visit our website: https://bostoninstituteofanalytics.org/india/delhi/connaught-place/school-of-technology-ai/data-science-and-artificial-intelligence/

This work presents a technique for classifying X-ray images of the chest (CXR) by applying deep learning-based techniques. The CXR will be classified into three different types, i.e. (i) normal, (ii) COVID-19, and (iii) pneumonia. The classification challenge is raised when the X-ray images of COVID-19 and pneumonia are subtle. The CXR images of the chest are first proceeded to be standardized and to improve the visual contrast of the images. Then, the classification is performed by applying a deep learningbased technique that binds two deep learning network architectures, i.e., convolution neural network (CNN) and long short-term memory (LSTM), to generate a hybrid model for the classification problem. The deep features of the images are extracted by CNN before the final classification is performed using LSTM. In addition to the hybrid models, this work explores the validity of image pre-processing methods that improve the quality of the images before the classification is performed. The experiments were conducted on a public image dataset. The experimental results demonstrate that the proposed technique provides promising results and is superior to the baseline techniques. 

Leukocytes, which are created in the bone marrow comprise one percent of all blood cells. When these white blood cells grow uncontrollably it gives rise, to the development of blood cancer. The proposed research presents an approach, for categorizing One of the three kinds of Multiple Myeloma (MM) and Acute Lymphoblastic Leukaemia (ALL) are the two diseases that make use of the SNAM dataset. the malignancy known as acute lymphoblastic leukaemia (ALL), to start with in which an excessively large number of lymphocytes are produced by the bone marrow. Secondly, Multiple myeloma (MM) is a type of cancer that results in the accumulation of malignant cells in bone marrow, rather than their release into the bloodstream. Hence, the growth of blood cells is to be resist and prevent. Beforehand, the procedure was carried out manually evaluated by experienced haematologists. The proposed methodology totally eliminates the chance of human mistake through using deep learning methods, particularly convolutional neural networks. A total of 89 ALL patients 3256 smears of peripheral blood (PBS) pictures were acquired from an online portal. The model undergoes training using modified convolutional neural networks that has been optimized and its ability to predict which type of malignancy is present in the cells is determined. In 96 out of 100 cases, the algorithm strongly replicated every measurement that corresponded to the samples. The accuracy of the system was found to be 97.6%, which is more appropriate than modern techniques like Decision Trees, Random Forests, Naive Bayes, and Support Vector Machines (SVMs), VGG16, VGG19, AlexNet, Google-Net, Mobile-NetV2. The work showcases that Modified CNN performs more accurately

The ability to describe images with natural language sentences is the hallmark for image and language understanding. Such a system has wide ranging applications such as annotating images and using natural sentences to search for images.In this project we focus on the task of bidirectional image retrieval: such asystem is capable of retrieving an image based on a sentence (image search) andretrieve sentence based on an image query (image annotation). We present asystem based on a global ranking objective function which uses a combinationof convolutional neural networks (CNN) and multi layer perceptrons (MLP).It takes a pair of image and sentence and processes them in different channels,finally embedding it into a common multimodal vector space. These embeddingsencode abstract semantic information about the two inputs and can be comparedusing traditional information retrieval approaches. For each such pair, the modelreturns a score which is interpretted as a similarity metric. If this score is high,the image and sentence are likely to convey similar meaning, and if the score is low then they are likely not to. The visual input is modeled via deep convolutional neural network. On theother hand we explore three models for the textual module. The first one isbag of words with an MLP. The second one uses n-grams (bigram, trigrams,and a combination of trigram & skip-grams) with an MLP. The third is morespecialized deep network specific for modeling variable length sequences (SSE).We report comparable performance to recent work in the field, even though ouroverall model is simpler. We also show that the training time choice of how wecan generate our negative samples has a significant impact on performance, and can be used to specialize the bi-directional system in one particular task.

In many image-related tasks, learning expressive and discriminative representations of images is essential, and deep learning has been studied for automating the learning of such representations. Some user-centric tasks, such as image recommendations, call for effective representations of not only images but also preferences and intents of users over images. Such representations are termed \emph{hybrid} and addressed via a deep learning approach in this paper. We design a dual-net deep network, in which the two sub-networks map input images and preferences of users into a same latent semantic space, and then the distances between images and users in the latent space are calculated to make decisions. We further propose a comparative deep learning (CDL) method to train the deep network, using a pair of images compared against one user to learn the pattern of their relative distances. The CDL embraces much more training data than naive deep learning, and thus achieves superior performance than the latter, with no cost of increasing network complexity. Experimental results with real-world data sets for image recommendations have shown the proposed dual-net network and CDL greatly outperform other state-of-the-art image recommendation solutions.

We propose a novel approach for instance-level image retrieval. It produces a global and compact fixed-length representation for each image by aggregating many region-wise descriptors. In contrast to previous works employing pre-trained deep networks as a black box to produce features, our method leverages a deep architecture trained for the specific task of image retrieval. Our contribution is twofold: (i) we leverage a ranking framework to learn convolution and projection weights that are used to build the region features; and (ii) we employ a region proposal network to learn which regions should be pooled to form the final global descriptor. We show that using clean training data is key to the success of our approach. To that aim, we use a large scale but noisy landmark dataset and develop an automatic cleaning approach. The proposed architecture produces a global image representation in a single forward pass. Our approach significantly outperforms previous approaches based on global descriptors on standard datasets. It even surpasses most prior works based on costly local descriptor indexing and spatial verification. Additional material is available at www.xrce.xerox.com/Deep-Image-Retrieval.

Deep learning has attracted great attention recently and yielded the state of the art performance in dimension reduction and classification problems. However, it cannot effectively handle the structured output prediction, e.g. sequential labeling. In this paper, we propose a deep learning structure, which can learn discriminative features for sequential labeling problems. More specifically, we add the inter-relationship between labels in our deep learning structure, in order to incorporate the context information from the sequential data. Thus, our model is more powerful than linear Conditional Random Fields (CRFs) because the objective function learns latent non-linear features so that target labeling can be better predicted. We pretrain the deep structure with stacked restricted Boltzmann machines (RBMs) for feature learning and optimize our objective function with online learning algorithm, a mixture of perceptron training and stochastic gradient descent. We test our model on different challenge tasks, and show that our model outperforms significantly over the completive baselines.

We develop a representation suitable for the unconstrained recognition of words in natural images: the general case of no fixed lexicon and unknown length. To this end we propose a convolutional neural network (CNN) based architecture which incorporates a Conditional Random Field (CRF) graphical model, taking the whole word image as a single input. The unaries of the CRF are provided by a CNN that predicts characters at each position of the output, while higher order terms are provided by another CNN that detects the presence of N-grams. We show that this entire model (CRF, character predictor, N-gram predictor) can be jointly optimised by back-propagating the structured output loss, essentially requiring the system to perform multi-task learning, and training uses purely synthetically generated data. The resulting model is a more accurate system on standard real-world text recognition benchmarks than character prediction alone, setting a benchmark for systems that have not been trained on a particular lexicon. In addition, our model achieves state-of-the-art accuracy in lexicon-constrained scenarios, without being specifically modelled for constrained recognition. To test the generalisation of our model, we also perform experiments with random alpha-numeric strings to evaluate the method when no visual language model is applicable.

We investigate the $\ell_\infty$-constrained representation which demonstrates robustness to quantization errors, utilizing the tool of deep learning. Based on the Alternating Direction Method of Multipliers (ADMM), we formulate the original convex minimization problem as a feed-forward neural network, named \textit{Deep $\ell_\infty$ Encoder}, by introducing the novel Bounded Linear Unit (BLU) neuron and modeling the Lagrange multipliers as network biases. Such a structural prior acts as an effective network regularization, and facilitates the model initialization. We then investigate the effective use of the proposed model in the application of hashing, by coupling the proposed encoders under a supervised pairwise loss, to develop a \textit{Deep Siamese $\ell_\infty$ Network}, which can be optimized from end to end. Extensive experiments demonstrate the impressive performances of the proposed model. We also provide an in-depth analysis of its behaviors against the competitors.

Deep Learning is a very powerful machine learning model. Deep Learning trains a large number of parameters for multiple layers and is very slow when data is in large scale and the architecture size is large. Inspired from the shrinking technique used in accelerating computation of Support Vector Machines (SVM) algorithm and screening technique used in LASSO, we propose a shrinking Deep Learning with recall (sDLr) approach to speed up deep learning computation. We experiment shrinking Deep Learning with recall (sDLr) using Deep Neural Network (DNN), Deep Belief Network (DBN) and Convolution Neural Network (CNN) on 4 data sets. Results show that the speedup using shrinking Deep Learning with recall (sDLr) can reach more than 2.0 while still giving competitive classification performance.

Deep learning models have shown incredible achievement in the field of autonomous driving, covering different aspects ranging from recognizing traffic signs and traffic lighs, vehicle detection, license plate detection, pedestrian detection. Most of the algorithms perrform better when the traffic lights are bigger in size, but the performance degrades in case of small-sized traffic lights. In this paper, the main emphasis is on evaluating two most promising deep learning architectures: single shot detector (SSD) and faster region convolutinal network (Faster R-CNN) on “la route automatisée (LaRA) traffic light dataset” which contains small traffic lights as objects. The strengths and weaknesses are evaluated based on different parameters. The performance is compared in terms of mean average Precision ([email protected]) and average recall. The impact of data augmentation on the two architectures is also analyzed. ResNet50 V1 as feature extractor for Faster R-CNN achieved 96% mAP (mean average precision) which performed better than Original ResNet50 V1 Faster R-CNN pipeline. Also, different parameters such as batch size, learning rate and optimizer are tuned for detecting and classifying small traffic lights into different categories.

We propose an heterogeneous multi-task learning framework for human pose estimation from monocular image with deep convolutional neural network. In particular, we simultaneously learn a pose-joint regressor and a sliding-window body-part detector in a deep network architecture. We show that including the body-part detection task helps to regularize the network, directing it to converge to a good solution. We report competitive and state-of-art results on several data sets. We also empirically show that the learned neurons in the middle layer of our network are tuned to localized body parts.

Many problems in real-world applications involve predicting several random variables which are statistically related. Markov random fields (MRFs) are a great mathematical tool to encode such relationships. The goal of this paper is to combine MRFs with deep learning algorithms to estimate complex representations while taking into account the dependencies between the output random variables. Towards this goal, we propose a training algorithm that is able to learn structured models jointly with deep features that form the MRF potentials. Our approach is efficient as it blends learning and inference and makes use of GPU acceleration. We demonstrate the effectiveness of our algorithm in the tasks of predicting words from noisy images, as well as multi-class classification of Flickr photographs. We show that joint learning of the deep features and the MRF parameters results in significant performance gains.

Deep neural networks that contain many layers of non-linear feature detectors fell out of favor because it was hard to get enough labeled data to train many millions of parameters and difficult to optimize the connection weights really well. Both of these problems can be overcome by first training a multi-layer belief net to form a top-down generative model of unlabeled input data and then using the features discovered by the belief net to initialize a bottom-up neural net. The neural net can then be discriminatively fine-tuned on a smaller set of labelled data. Marc'Aurelio Ranzato has recently used this deep learning method to create a very good generative model of natural images and Navdeep Jaitly has used it to learn features that are derived directly from the raw sound wave and outperform the features that are usually used for phoneme recognition. An alternative way to deal with the difficult optimization problem is to develop a more sophisticated optimizer that works really well for artificial neural networks – something that the optimization community has often suggested but never done. Ilya Sutskever has recently used an excellent "Hessian free" optimizer developed by James Martens to learn a recurrent neural network that predicts the next character in a string. "He was elected President during the Revolutionary War and forgave Opus Paul at Rome" is an example of what this neural net generates after being trained on character strings from Wikipedia. ©2010 Microsoft Corporation. All rights reserved.

This work proposes deep network models and learning algorithms for unsupervised and supervised binary hashing. Our novel network design constrains one hidden layer to directly output the binary codes. This addresses a challenging issue in some previous works: optimizing non-smooth objective functions due to binarization. Moreover, we incorporate independence and balance properties in the direct and strict forms in the learning. Furthermore, we include similarity preserving property in our objective function. Our resulting optimization with these binary, independence, and balance constraints is difficult to solve. We propose to attack it with alternating optimization and careful relaxation. Experimental results on three benchmark datasets show that our proposed methods compare favorably with the state of the art.

Despite significant recent progress, the best available computer vision algorithms still lag far behind human capabilities, even for recognizing individual discrete objects under various poses, illuminations, and backgrounds. Here we present a new approach to using object pose information to improve deep network learning. While existing large-scale datasets, e.g. ImageNet, do not have pose information, we leverage the newly published turntable dataset, iLab-20M, which has ~22M images of 704 object instances shot under different lightings, camera viewpoints and turntable rotations, to do more controlled object recognition experiments. We introduce a new convolutional neural network architecture, what/where CNN (2W-CNN), built on a linear-chain feedforward CNN (e.g., AlexNet), augmented by hierarchical layers regularized by object poses. Pose information is only used as feedback signal during training, in addition to category information; during test, the feedforward network only predicts category. To validate the approach, we train both 2W-CNN and AlexNet using a fraction of the dataset, and 2W-CNN achieves 6% performance improvement in category prediction. We show mathematically that 2W-CNN has inherent advantages over AlexNet under the stochastic gradient descent (SGD) optimization procedure. Further more, we fine-tune object recognition on ImageNet by using the pretrained 2W-CNN and AlexNet features on iLab-20M, results show that significant improvements have been achieved, compared with training AlexNet from scratch. Moreover, fine-tuning 2W-CNN features performs even better than fine-tuning the pretrained AlexNet features. These results show pretrained features on iLab- 20M generalizes well to natural image datasets, and 2WCNN learns even better features for object recognition than AlexNet.

Radiation therapy has emerged as one of the preferred techniques to treat brain cancer patients. During treatment, a very high dose of radiation is delivered to a very narrow area. Prescribed radiation therapy for brain cancer requires precisely defining the target treatment area, as well as delineating vital brain structures which must be spared from radiotoxicity. Nevertheless, delineation task is usually still manually performed, which is inefficient and operator-dependent. Several attempts of automatizing this process have reported. however, marginal results when analyzing organs in the optic region. In this work we present a deep learning classification scheme based on augmented-enhanced features to automatically segment organs at risk (OARs) in the optic region -optic nerves, optic chiasm, pituitary gland and pituitary stalk-. Fifteen MR images with various types of brain tumors were retrospectively collected to undergo manual and automatic segmentation. Mean Dice Similarity coefficients around 0.80 were reported. Incorporation of proposed features yielded to improvements on the segmentation. Compared with support vector machines, our method achieved better performance with less variation on the results, as well as a considerably reduction on the classification time. Performance of the proposed approach was also evaluated with respect to manual contours. In this case, results obtained from the automatic contours mostly lie on the variability of the observers, showing no significant differences with respect to them. These results suggest therefore that the proposed system is more accurate than other presented approaches, up to date, to segment these structures. The speed, reproducibility, and robustness of the process make the proposed deep learning-based classification system a valuable tool for assisting in the delineation task of small OARs in brain cancer.

Reinforcement learning is considered to be a strong AI paradigm which can be used to teach machines through interaction with the environment and learning from their mistakes. Despite its perceived utility, it has not yet been successfully applied in automotive applications. Motivated by the successful demonstrations of learning of Atari games and Go by Google DeepMind, we propose a framework for autonomous driving using deep reinforcement learning. This is of particular relevance as it is difficult to pose autonomous driving as a supervised learning problem due to strong interactions with the environment including other vehicles, pedestrians and roadworks. As it is a relatively new area of research for autonomous driving, we provide a short overview of deep reinforcement learning and then describe our proposed framework. It incorporates Recurrent Neural Networks for information integration, enabling the car to handle partially observable scenarios. It also integrates the recent work on attention models to focus on relevant information, thereby reducing the computational complexity for deployment on embedded hardware. The framework was tested in an open source 3D car racing simulator called TORCS. Our simulation results demonstrate learning of autonomous maneuvering in a scenario of complex road curvatures and simple interaction of other vehicles.

We propose an inverse reinforcement learning (IRL) approach using Deep Q-Networks to extract the rewards in problems with large state spaces. We evaluate the performance of this approach in a simulation-based autonomous driving scenario. Our results resemble the intuitive relation between the reward function and readings of distance sensors mounted at different poses on the car. We also show that, after a few learning rounds, our simulated agent generates collision-free motions and performs human-like lane change behaviour.

Instability and variability of Deep Reinforcement Learning (DRL) algorithms tend to adversely affect their performance. Averaged-DQN is a simple extension to the DQN algorithm, based on averaging previously learned Q-values estimates, which leads to a more stable training procedure and improved performance by reducing approximation error variance in the target values. To understand the effect of the algorithm, we examine the source of value function estimation errors and provide an analytical comparison within a simplified model. We further present experiments on the Arcade Learning Environment benchmark that demonstrate significantly improved stability and performance due to the proposed extension.

The increasing quality of smartphone cameras and variety of photo editing applications, in addition to the rise in popularity of image-centric social media, have all led to a phenomenal growth in mobile-based photography. Advances in computer vision and machine learning techniques provide a large number of cloud-based services with the ability to provide content analysis, face recognition, and object detection facilities to third parties. These inferences and analytics might come with undesired privacy risks to the individuals. In this paper, we address a fundamental challenge: Can we utilize the local processing capabilities of modern smartphones efficiently to provide desired features to approved analytics services, while protecting against undesired inference attacks and preserving privacy on the cloud? We propose a hybrid architecture for a distributed deep learning model between the smartphone and the cloud. We rely on the Siamese network and machine learning approaches for providing privacy based on defined privacy constraints. We also use transfer learning techniques to evaluate the proposed method. Using the latest deep learning models for Face Recognition, Emotion Detection, and Gender Classification techniques, we demonstrate the effectiveness of our technique in providing highly accurate classification results for the desired analytics, while proving strong privacy guarantees.

Using deep neural nets as function approximator for reinforcement learning tasks have recently been shown to be very powerful for solving problems approaching real-world complexity. Using these results as a benchmark, we discuss the role that the discount factor may play in the quality of the learning process of a deep Q-network (DQN). When the discount factor progressively increases up to its final value, we empirically show that it is possible to significantly reduce the number of learning steps. When used in conjunction with a varying learning rate, we empirically show that it outperforms original DQN on several experiments. We relate this phenomenon with the instabilities of neural networks when they are used in an approximate Dynamic Programming setting. We also describe the possibility to fall within a local optimum during the learning process, thus connecting our discussion with the exploration/exploitation dilemma.

The ability to act in multiple environments and transfer previous knowledge to new situations can be considered a critical aspect of any intelligent agent. Towards this goal, we define a novel method of multitask and transfer learning that enables an autonomous agent to learn how to behave in multiple tasks simultaneously, and then generalize its knowledge to new domains. This method, termed "Actor-Mimic", exploits the use of deep reinforcement learning and model compression techniques to train a single policy network that learns how to act in a set of distinct tasks by using the guidance of several expert teachers. We then show that the representations learnt by the deep policy network are capable of generalizing to new tasks with no prior expert guidance, speeding up learning in novel environments. Although our method can in general be applied to a wide range of problems, we use Atari games as a testing environment to demonstrate these methods.

Computational methods that automatically extract knowledge from data are critical for enabling data-driven materials science. A reliable identification of lattice symmetry is a crucial first step for materials characterization and analytics. Current methods require a user-specified threshold, and are unable to detect ``average symmetries'' for defective structures. Here, we propose a new machine-learning-based approach to automatically classify structures by crystal symmetry. First, we represent crystals by calculating a diffraction image, then construct a deep-learning neural-network model for classification.Our approach is able to correctly classify a dataset comprising more than 100,000 simulated crystal structures, including heavily defective ones. The internal operations of the neural network are unraveled through attentive response maps, demonstrating that it uses the same landmarks a materials scientist would use, although never explicitly instructed to do so. Our study paves the way for crystal-structure recognition of - possibly noisy and incomplete - three-dimensional structural data in big-data materials science. CC0 Waiver

Person re-identification (re-id) is the task of matching multiple occurrences of the same person from different cameras, poses, lighting conditions, and a multitude of other factors which alter the visual appearance. Typically, this is achieved by learning either optimal features or matching metrics which are adapted to specific pairs of camera views dictated by the pairwise labelled training datasets. In this work, we formulate a deep learning based novel approach to automatic prototype-domain discovery for domain perceptive (adaptive) person re-id (rather than camera pair specific learning) for any camera views scalable to new unseen scenes without training data. We learn a separate re-id model for each of the discovered prototype-domains and during model deployment, use the person probe image to select automatically the model of the closest prototype domain. Our approach requires neither supervised nor unsupervised domain adaptation learning, i.e. no data available from the target domains. We evaluate extensively our model under realistic re-id conditions using automatically detected bounding boxes with low-resolution and partial occlusion. We show that our approach outperforms most of the state-of-the-art supervised and unsupervised methods on the latest CUHK-SYSU and PRW benchmarks.

Convolutional networks have marked their place over the last few years as the best performing model for various visual tasks. They are, however, most suited for supervised learning from large amounts of labeled data. Previous attempts have been made to use unlabeled data to improve model performance by applying unsupervised techniques. These attempts require different architectures and training methods. In this work we present a novel approach for unsupervised training of Convolutional networks that is based on contrasting between spatial regions within images. This criterion can be employed within conventional neural networks and trained using standard techniques such as SGD and back-propagation, thus complementing supervised methods.

This paper introduces a novel method for learning how to play the most difficult Atari 2600 games from the Arcade Learning Environment using deep reinforcement learning. The proposed method, human checkpoint replay, consists in using checkpoints sampled from human gameplay as starting points for the learning process. This is meant to compensate for the difficulties of current exploration strategies, such as epsilon-greedy, to find successful control policies in games with sparse rewards. Like other deep reinforcement learning architectures, our model uses a convolutional neural network that receives only raw pixel inputs to estimate the state value function. We tested our method on Montezuma's Revenge and Private Eye, two of the most challenging games from the Atari platform. The results we obtained show a substantial improvement compared to previous learning approaches, as well as over a random player. We also propose a method for training deep reinforcement learning agents using human gameplay experience, which we call human experience replay.

In recent years, deep neural networks (DNNs), have yielded strong results on a wide range of applications. Graphics Processing Units (GPUs) have been one key enabling factor leading to the current popularity of DNNs. However, despite increasing hardware flexibility and software programming toolchain maturity, high efficiency GPU programming remains difficult: it suffers from high complexity, low productivity, and low portability. GPU vendors such as NVIDIA have spent enormous effort to write special-purpose DNN libraries. However, on other hardware targets, especially mobile GPUs, such vendor libraries are not generally available. Thus, the development of portable, open, high-performance, energy-efficient GPU code for DNN operations would enable broader deployment of DNN-based algorithms. Toward this end, this work presents a framework to enable productive, high-efficiency GPU programming for DNN computations across hardware platforms and programming models. In particular, the framework provides specific support for metaprogramming, autotuning, and DNN-tailored data types. Using our framework, we explore implementing DNN operations on three different hardware targets: NVIDIA, AMD, and Qualcomm GPUs. On NVIDIA GPUs, we show both portability between OpenCL and CUDA as well competitive performance compared to the vendor library. On Qualcomm GPUs, we show that our framework enables productive development of target-specific optimizations, and achieves reasonable absolute performance. Finally, On AMD GPUs, we show initial results that indicate our framework can yield reasonable performance on a new platform with minimal effort.

We investigate the problem of image retrieval based on visual queries when the latter comprise arbitrary regions-of-interest (ROI) rather than entire images. Our proposal is a compact image descriptor that combines the state-of-the-art in content-based descriptor extraction with a multi-level, Voronoi-based spatial partitioning of each dataset image. The proposed multi-level Voronoi-based encoding uses a spatial hierarchical K-means over interest-point locations, and computes a content-based descriptor over each cell. In order to reduce the matching complexity with minimal or no sacrifice in retrieval performance: (i) we utilize the tree structure of the spatial hierarchical K-means to perform a top-to-bottom pruning for local similarity maxima; (ii) we propose a new image similarity score that combines relevant information from all partition levels into a single measure for similarity; (iii) we combine our proposal with a novel and efficient approach for optimal bit allocation within quantized descriptor representations. By deriving both a Voronoi-based VLAD descriptor (termed as Fast-VVLAD) and a Voronoi-based deep convolutional neural network (CNN) descriptor (termed as Fast-VDCNN), we demonstrate that our Voronoi-based framework is agnostic to the descriptor basis, and can easily be slotted into existing frameworks. Via a range of ROI queries in two standard datasets, it is shown that the Voronoi-based descriptors achieve comparable or higher mean Average Precision against conventional grid-based spatial search, while offering more than two-fold reduction in complexity. Finally, beyond ROI queries, we show that Voronoi partitioning improves the geometric invariance of compact CNN descriptors, thereby resulting in competitive performance to the current state-of-the-art on whole image retrieval.

Due to the rapid increase in both vehicle traffic and urbanization, effective traffic control systems are now essential. This review paper integrates Intelligent Traffic Analysis Systems (ITAS) with Convolutional Neural Networks (CNNs), a deep learning technology, to provide accurate and real-time data analysis. Advanced technologies used by ITAS monitor, analyze, and reduce traffic flow. Specifically designed deep learning models for object detection and tracking are employed to recognize and monitor cars, trucks, and other relevant entities in the recorded data. Transfer learning from previously trained models is used to train the proposed CNN architecture, which is modified for traffic analysis. This approach enhances efficiency and helps avoid road traffic congestion. Artificial Neural Network, Deep Learning technology, Convolutional Neural Networks (C NNs), Machine Learning.

Due to the numerous variables that may influence the soil-machine interaction systems, predicting the mechanical response of soil interacting with off-road traction equipment is challenging. In this study, deep neural networks (DNNs) are chosen as a potential solution for explaining the varying soil sinkage rates because of their ability to model complex, multivariate, and dynamic systems. Plate sinkage tests were carried out using a Bevameter in a fixed-type soil bin with a 24 m length, 2 m width, and 1 m depth. Experimental tests were conducted at three sinkage rates for two plate sizes, with a soil water content of 10%. The provided empirical data on the soil pressure-sinkage relationship served as the basis for an algorithm capable of discerning the soil-machine interaction. From the iterative process, it was determined that a DNN, specifically a feed-forward back-propagation DNN with three hidden layers, is the optimal choice. The optimized DNN architecture is structured as 3-8-15-10-1, as determined by the Grey Wolf Optimization algorithm. While the Bekker equation had traditionally been employed as a widely accepted method for predicting soil pressure-sinkage behavior, it typically disregarded the influence of sinkage velocity of the soil. However, the findings revealed the significant impact of sinkage velocity on the parameters governing the soil deformation response. The trained DNN successfully incorporated the sinkage velocity into its structure and provided accurate results with an MSE value of 0.0871.

Abstract: Detection of fake news based on deep learning techniques is a major issue used to mislead people. For the experiments, several types of datasets, models, and methodologies have been used to detect fake news. Also, most of the datasets contain text id, tweets id, and user-based id and user-based features. To get the proper results and accuracy various models like CNN (Convolution neural network), DEEP CNN, and LSTM (Long short-term memory) are used

Deep Learning A-Z™ Hands-On Artificial Neural Networks(18. Building a Boltzmann Machine)

Deep Learning A-Z™ Hands-On Artificial Neural Networks(8. Part 3 - Recurrent Neural Networks)

Since the outbreak, the novel coronavirus (COVID-19) has infected more than 180 million people and has taken a toll of 3.91 million lives globally as of June 2021. This virus causes symptoms like fever, cold, and fatigue, and can develop into Pneumonia which can be detected using chest X-rays (CXRs). Therefore, early detection of COVID-19 can help get early medical attention. However, a sudden rise in the number of cases in many countries caused by COVID waves increases the burden on their testing facilities. As a result, they sometimes fail to perform enough testing to contain the spread. This work proposes a deep learning model to detect COVID-19 and Pneumonia based on CXRs. The dataset for our COVID model contains a total of 3,400 CXRs images of COVID-19 patients and 3,400 normal CXRs. The dataset for our Pneumonia model contains 1,300 CXR images of Pneumonia patients and 1,300 normal CXRs. We use convolutional neural network provided by TensorFlow to build our model, which gave 94.17% and 93.55% accuracy for COVID model and Pneumonia model, respectively. Finally, we deployed our model on the web and added a web tracker, which gives us the cases, deaths, and recoveries state-wise and nationwide.

Macroeconomic indicator forecasting is a difficult task and the macroeconomy’s complex operations and dynamic nature make it even more difficult. Machine Learning and Deep Learning methodologies have been investigated as alternatives to traditional forecasting methods because of recent developments in computing power and the emergence of data. How the Machine Learning and Deep Learning paradigms apply to a variety of Macro datasets have been examined in this research paper. Few Machine Learning and Deep Learning algorithms have been trained and their forecasting accuracy has been compared with that of traditional statistical method ARIMA.

Deep Neural Networks (DNNs) are currently used in a wide range of critical real-world applications as machine learning technology. Due to the high number of parameters that make up DNNs, learning and prediction tasks require millions of floating-point operations (FLOPs). Implementing DNNs into a cloud computing system with centralized servers and data storage sub-systems equipped with high-speed and high-performance computing capabilities is a more effective strategy. This research presents an updated analysis of the most recent DNNs used in cloud computing. It highlights the necessity of cloud computing while presenting and debating numerous DNN complexity issues related to various architectures. Additionally, it goes into their intricacies and offers a thorough analysis of several cloud computing platforms for DNN deployment. Additionally, it examines the DNN applications that are already running on cloud computing platforms to highlight the advantages of using cloud computing for DNNs. The study highlights the difficulties associated with implementing DNNs in cloud computing systems and provides suggestions for improving both current and future deployments.

It is often stressed that the pedagogic models and approaches of Collaborative Online Learning support a learner's shared knowledge building within collaborating groups of learners, the individual construction of knowledge and the formation of an ongoing learning Community of Practice. Based on a recent case study of a Danish Master's programme, this paper will demonstrate that the emerging collaborative practice displays tendencies contrary to the generally accepted assumptions. The outcome is not only based on the models and their attributes, it is also affected by the emerging practice itself and the interaction among the participants during a course. From this perspective, it is relevant to look at which possibilities and obstacles teachers encounter when they try to detect slowly emerging tendencies that may lead to major misinterpretations of the subject matter and marginalize or even exclude students from participating in the learning Community of Practice. In conclusion, the case study will identify the slowly emerging tendencies that may be detected and observed at an early stage and thus indicate areas in on-line learning environments that require special attention.

Introduction [1] : Development of a predictive model using machine learning can help livestock farmers to increase their understanding of the performance potential of their livestock. It can assist in decision-making processes related to livestock management, elimination and replacement selection, nutrition, reproduction and other matters of livestock management. Predicting insemination outcomes provides valuable insights to improve reproductive performance, breeding processes, milk production and overall livestock efficiency. The integration of predicting models in the existing systems in animal husbandry increases its practical application as a decision support tool for animal farmers. By developing a tool that can determine the reproductive success of livestock, ranchers can optimize their production and breeding strategies and improve overall livestock management practices to increase reproductive efficiency and profitability. In this study Material and Methods: This study utilized data from the Helal Agro-Industry Co., a commercial dairy farm associated with the Iranian Red Crescent Investment Company. The commercial dairy herds in this region primarily consist of Holstein-Friesian cattle. In terms of record-keeping and efficient data management, the agricultural enterprise utilizes the Modiran Farmer software. This software leverages the Microsoft SQL Server database infrastructure to facilitate the storage of pertinent information. The dataset encompasses a diverse array of tables containing entries spanning various aspects such as reproduction, milking, health profiles, genetic insights, and broader characteristic attributes. The temporal scope of the database spans from January 1994 through May 2023, encapsulating a substantial historical period. We executed a SQL query against the database to generate a dataset of insemination records and their corresponding features. For each insemination record, we retrieved 25 features encompassing covariates related to milking, reproduction, management factors, health, and insemination result. The data underwent further pre-processing after the extraction process to make it suitable for the proposed models. We proposed three different models of Long Short-Term Memory, Multi-Layer perceptron, and XGBoost. A distinct set of cow IDs was acquired, and then, it was partitioned into three subsets: 70% for training, 10% for validation, and 20% for testing. In order to work with LSTM model, by identifying the temporal dependencies relations between a cow’s insemination cycles, we stacked these cycles to form sequences that can then be processed by LSTM model. So, the sets of unique cow IDs were then used to generate the sequences for each cow. A data augmentation method was used to generate all possible sub-sequences of cows’ insemination. Then, the sequences were aligned and stacked to achieve a constant length of 20. In total, about 168,000 training sequences, 23,000 validation sequences, and about 46,000 test sequences were generated. We tuned the parameters and hyperparameters of each model and upon finalizing the optimal architectures for our models, we initiated training experiments by fitting the models to the prepared datasets. Results and Discussion: Our experimental findings reveal that the proposed LSTM model significantly improved prediction accuracy compared to the MLP and XGBoost models. The LSTM model, with its archit e cture of three consecutive LSTM layers, was able to demonstrate the best performance across all evaluation metrics on average over the 10 training runs. LSTM networks are designed to handle long time dependencies well. These networks use memory cells to hold important information over time, which makes them suitable for ordinal data such as time series. In contrast, XGBoost and MLP are not specifically designed to handle temporal dependencies and their performance is more limited on this type of data. Also, LSTM network can learn complex dependencies between ordinal data well. This ability is attributed to the unique structure of LSTM and its gate mechanisms, which enable the network to filter out irrelevant information while retaining essential information. In contrast, models based on XGBoost and MLP are less effective in this area, as they primarily focus on direct interactions between features and struggle to capture temporal dependencies. LSTM-based models excel in extracting higher-level features from data due to their deep learning capabilities. These features provide richer information for classification tasks, ultimately improving classification accuracy. Although XGBoost-based models are known for their precision, they are less adept at extracting high-level features. Additionally, the memory structure of LSTM allows it to handle fluctuations and unexpected variations in sequential data, effectively distinguishing critical information from noise. This feature helps LSTM perform better in situations where the data contains a lot of noise and fluctuations. Conclusion: Overall, we presented and tested the performance of different models for predicting the results of artificial insemination of livestock. This prediction can help livestock farmers improve performance, increase fertility, and reduce livestock costs. In the problem of predicting the results of artificial insemination of livestock, the presented LSTM neural network model shows the best performance based on the stated evaluation criteria, and then the XGBoost-based classifier has better performance than MLP.    

The DoD is investing in autonomy, AI, and machine learning. Deep learning, a sub-field of machine learning is increasing due to newer and cheaper hardware, new algorithms, and big data. Deep learning uses a neural network with multiple weighted layers designed to learn hierarchical feature representations. This research uses the technique of transfer learning, which takes the well-constructed architecture of a source model and retrains it to a target data set—in this case, different coastal landscapes. Eight different classes were trained with oblique (≥ 45°) images. An average accuracy of 95% correct identification was achieved through validation testing. Carmel River State Beach is a known morphodynamic site that changes seasonally. Five different stitched together <10° off NADIR mosaics of this site were selected to test the model’s ability to detect and correctly label areas of change over time. The mosaics were broken into four quadrants of equal area to increase homogeneity of the features. The two landward quadrants showed successful label and change detection; the seaward quadrants showed poor results attributed to smearing and gradient distortion from the stitching process. Successful transfer learning was accomplished with high accuracy; angle differences and stitching caused mislabeling. Larger datasets with single images from multiple angles may reduce labeling error. Multi-label and multispectral approach will enhance and broaden the application of this process.