#ConvolutionalNeuralNetworks

[softlabsgroup05]

Unlock the secrets of training Convolutional Neural Networks (CNNs) with our user-friendly flowchart. From data preparation to model fine-tuning, grasp each step in building powerful image classifiers. Simplify your deep learning path with our guide, perfect for beginners and experts alike. Stay tuned to Softlabs Group for more AI insights!

[sifytech]

Unveiling the Depths of Intelligence: A Journey into the World of Deep Learning

Embark on a captivating exploration into the cutting-edge realm of Deep Learning, a revolutionary paradigm within the broader landscape of artificial intelligence. Read More. https://www.sify.com/ai-analytics/unveiling-the-depths-of-intelligence-a-journey-into-the-world-of-deep-learning/

[guillaumelauzier]

Neural Style Transfer (NST)

Neural Style Transfer (NST) is a captivating intersection of artificial intelligence and artistic creativity. This technology leverages the capabilities of deep learning to merge the essence of one image with the aesthetic style of another.

Basic Concept of Neural Style Transfer (NST)

Combining Content and Style: NST works by taking two images - a content image (like a photograph) and a style image (usually a famous painting) - and combining them. The goal is to produce a new image that retains the original content but is rendered in the artistic style of the second image. Deep Learning at its Core: This process is made possible through deep learning techniques, specifically using Convolutional Neural Networks (CNNs). These networks are adept at recognizing and processing visual information. Content Representation: The CNN captures the content of the target image at its deeper layers, where the network understands higher-level features (like objects and their arrangements). Style Representation: The style of the source image is captured from the correlations between different layers of the CNN. These layers encode textural and color patterns characteristic of the artistic style. Image Transformation: The NST algorithm iteratively adjusts a third, initially random image to minimize the differences in content with the target image and in style with the source image. Resulting Image: The result is a fascinating blend that looks like the original photograph (content) 'painted' in the style of the artwork (style).

How Neural Style Transfer Works with Python Example

Content and Style Images: The process begins with two images: a content image (the subject you want to transform) and a style image (the artistic style to be transferred). Using a Pre-Trained CNN: Typically, a pre-trained CNN like VGG19 is used. This network has been trained on a vast dataset of images and can effectively extract and represent features from these images. Feature Extraction: The CNN extracts content features from the content image and style features from the style image. These features are essentially patterns and textures that define the image's content and style. Combining Features: The NST algorithm then creates a new image that combines the content features of the content image with the style features of the style image. Optimization: This new image is gradually refined through an optimization process, minimizing the loss between its content and the content image, and its style and the style image. Result: The final output is a new image that retains the essence of the content image but is rendered in the style of the style image. Python Code Example: import tensorflow as tf import tensorflow_hub as hub import matplotlib.pyplot as plt import numpy as np # Load content and style images content_image = plt.imread('path_to_content_image.jpg') style_image = plt.imread('path_to_style_image.jpg') # Load a style transfer model from TensorFlow Hub hub_model = hub.load('https://tfhub.dev/google/magenta/arbitrary-image-stylization-v1-256/2') # Preprocess images and run the style transfer content_image = tf.image.convert_image_dtype(content_image, tf.float32) style_image = tf.image.convert_image_dtype(style_image, tf.float32) stylized_image = hub_model(tf.constant(content_image), tf.constant(style_image)) # Display the output plt.imshow(np.squeeze(stylized_image)) plt.show() This code snippet uses TensorFlow and TensorFlow Hub to apply a style transfer model, merging the content of one image with the style of another.

Detailed Section on Content and Style Representations in Neural Style Transfer

Feature Extraction Using Pre-Trained CNN: VGG19, a CNN model pre-trained on a large dataset (like ImageNet), is often used. This model effectively extracts features from images. Content Representation: - The content of an image is represented by the feature maps of higher layers in the CNN. - These layers capture the high-level content of the image, such as objects and their spatial arrangement, but not the finer details or style aspects. Style Representation: - The style of an image is captured by examining the correlations across different layers' feature maps. - These correlations are represented as a Gram matrix, which effectively captures the texture and visual patterns that define the image's style. Combining Content and Style: - NST algorithms aim to preserve the content from the content image while adopting the style of the style image. - This is done by minimizing a loss function that measures the difference in content and style representations between the generated image and the respective content and style images. Python Code Example: import numpy as np import tensorflow as tf from tensorflow.keras.applications import vgg19 from tensorflow.keras.preprocessing.image import load_img, img_to_array # Function to preprocess the image for VGG19 def preprocess_image(image_path, target_size=(224, 224)): img = load_img(image_path, target_size=target_size) img = img_to_array(img) img = np.expand_dims(img, axis=0) img = vgg19.preprocess_input(img) return img # Load your content and style images content_image = preprocess_image('path_to_your_content_image.jpg') style_image = preprocess_image('path_to_your_style_image.jpg') # Load the VGG19 model model = vgg19.VGG19(weights='imagenet', include_top=False) # Define a function to get content and style features def get_features(image, model): layers = { 'content': , 'style': } features = {} outputs = + layers] model = tf.keras.Model(, outputs) image_features = model(image) for name, output in zip(layers + layers, image_features): features = output return features # Extract features content_features = get_features(content_image, model) style_features = get_features(style_image, model) This code provides a basic structure for extracting content and style features using VGG19 in Python. Further steps would involve defining and optimizing the loss functions to generate the stylized image.

Applications of Neural Style Transfer

Video Styling: NST can be applied to video content, allowing filmmakers and content creators to impart artistic styles to their videos. This can transform ordinary footage into visually stunning sequences that resemble paintings or other art forms. Website Design: In web design, NST can be used to create unique, visually appealing backgrounds and elements. Designers can apply specific artistic styles to images, aligning them with the overall aesthetic of the website. Fashion and Textile Design: NST has been explored in the fashion industry for designing fabrics and garments. By transferring artistic styles onto textile patterns, designers can create innovative and unique clothing lines. Augmented Reality (AR) and Virtual Reality (VR): In AR and VR environments, NST can enhance the visual experience by applying artistic styles in real-time, creating immersive and engaging worlds for users. Product Design: NST can be used in product design to create visually appealing prototypes and presentations, allowing designers to experiment with different artistic styles quickly. Therapeutic Settings for Mental Health: There's growing interest in using NST in therapeutic settings. By creating soothing and pleasant images, it can be used as a tool for relaxation and stress relief, contributing positively to mental health and well-being. Educational Tools: NST can also be used as an educational tool in art and design schools, helping students understand the nuances of different artistic styles and techniques. These diverse applications showcase the versatility of NST, demonstrating its potential beyond the realm of digital art creation.

Limitations and Challenges of Neural Style Transfer

Computational Intensity: - NST, especially when using deep learning models like VGG19, is computationally demanding. It requires significant processing power, often necessitating the use of GPUs to achieve reasonable processing times. Balancing Content and Style: - Achieving the right balance between content and style in the output image can be challenging. It often requires careful tuning of the algorithm's parameters and may involve a lot of trial and error. Unpredictability of Results: - The outcome of NST can be unpredictable. The results may vary widely based on the chosen content and style images and the specific configurations of the neural network. Quality of Output: - The quality of the generated image can sometimes be lower than expected, with issues like distortions in the content or the style not being accurately captured. Training Data Limitations: - The effectiveness of NST is also influenced by the variety and quality of images used to train the underlying model. Limited or biased training data can affect the versatility and effectiveness of the style transfer. Overfitting: - There's a risk of overfitting, especially when the style transfer model is trained on a narrow set of images. This can limit the model's ability to generalize across different styles and contents. These challenges highlight the need for ongoing research and development in the field of NST to enhance its efficiency, versatility, and accessibility.

Necessary Hardware Resources for AI and Machine Learning in Art Generation

To effectively work with AI and machine learning algorithms for art generation, which can be computationally intensive, certain hardware resources are essential: High-Performance GPUs: - Graphics Processing Units (GPUs) are crucial for their ability to handle parallel tasks, making them ideal for the intensive computations required in training and running neural networks. - GPUs significantly reduce the time required for training models and generating art, compared to traditional CPUs. Sufficient RAM: - Adequate Random Access Memory (RAM) is important for handling large datasets and the high memory requirements of deep learning models. - A minimum of 16GB RAM is recommended, but 32GB or higher is preferable for more complex tasks. Fast Storage Solutions: - Solid State Drives (SSDs) are preferred over Hard Disk Drives (HDDs) for their faster data access speeds, which is beneficial when working with large datasets and models. High-Performance CPUs: - While GPUs handle most of the heavy lifting, a good CPU can improve overall system performance and efficiency. - Multi-core processors with high clock speeds are recommended. Cloud Computing Platforms: - Cloud computing resources like AWS, Google Cloud Platform, or Microsoft Azure offer powerful hardware for AI and machine learning tasks without the need for local installation. - These platforms provide scalability, allowing you to choose resources as per the project's requirements. Adequate Cooling Solutions: - High computational tasks generate significant heat. Therefore, a robust cooling solution is necessary to maintain optimal hardware performance and longevity. Reliable Power Supply: - A stable and reliable power supply is crucial, especially for desktop setups, to ensure uninterrupted processing and to protect the hardware from power surges. Investing in these hardware resources can greatly enhance the efficiency and capabilities of AI and machine learning algorithms in art generation and other computationally demanding tasks.

Limitations and Challenges of Neural Style Transfer

Neural Style Transfer (NST), despite its innovative applications in art and technology, faces several limitations and challenges: Computational Resource Intensity: - NST is computationally demanding, often requiring powerful GPUs and significant processing power. This can be a barrier for individuals or organizations without access to high-end computing resources. Quality and Resolution of Output: - The quality and resolution of the output images can sometimes be less than satisfactory. High-resolution images may lose detail or suffer from distortions after the style transfer. Balancing Act Between Content and Style: - Achieving a harmonious balance between the content and style in the output image can be challenging. It often requires fine-tuning of parameters and multiple iterations. Generalization and Diversity: - NST models might struggle with generalizing across vastly different styles or content types. This can limit the diversity of styles that can be effectively transferred. Training Data Biases: - The effectiveness of NST can be limited by the biases present in the training data. A model trained on a narrow range of styles may not perform well with radically different artistic styles. Overfitting Risks: - There's a risk of overfitting when the style transfer model is exposed to a limited set of images, leading to reduced effectiveness on a broader range of styles. Real-Time Processing Challenges: - Implementing NST in real-time applications, such as video styling, can be particularly challenging due to the intensive computational requirements. Understanding and addressing these limitations and challenges is crucial for the advancement and wider application of NST technologies.

Trends and Innovations in Neural Style Transfer (NST)

Neural Style Transfer (NST) is an evolving field with continuous advancements and innovations. These developments are broadening its applications and enhancing its efficiency: Improving Efficiency: - Research is focused on making NST algorithms faster and more resource-efficient. This includes optimizing existing neural network architectures and developing new methods to reduce computational requirements. Adapting to Various Artistic Styles: - Innovations in NST are enabling the adaptation to a wider range of artistic styles. This includes the ability to mimic more complex and abstract art forms, providing artists and designers with more diverse creative tools. Extending Applications Beyond Visual Art: - NST is finding applications in areas beyond traditional visual art. This includes video game design, film production, interior design, and even fashion, where NST can be used to create unique patterns and designs. Real-Time Style Transfer: - Advances in real-time processing capabilities are enabling NST to be applied in dynamic environments, such as live video feeds, augmented reality (AR), and virtual reality (VR). Integration with Other AI Technologies: - NST is being combined with other AI technologies like Generative Adversarial Networks (GANs) and reinforcement learning to create more sophisticated and versatile style transfer tools. User-Friendly Tools and Platforms: - The development of more user-friendly NST tools and platforms is democratizing access, allowing artists and non-technical users to experiment with style transfer without deep technical knowledge. These trends and innovations are propelling NST into new realms of creativity and practical application, making it a rapidly growing area in the field of AI and machine learning.

🌐 Sources

- Neural Style Transfer: Trends, Innovations, and Benefits - Challenges and Limitations of Deep Learning for Style Transfer - Neural Style Transfer: A Critical Review - Neural Style Transfer for So Long, and Thanks for all the Fish - Advantages and disadvantages of two methods of Neural Style Transfer - Evaluate and improve the quality of neural style transfer - Neural Style Transfer: Creating Artistic Images with Deep Learning - Classic algorithms, neural style transfer, GAN - Ricardo Corin - Mastering Neural Style Transfer - Neural Style Transfer Papers on GitHub - How to Make Artistic Images with Neural Style Transfer - Artificial Intelligence and Applications: Neural Style Transfer - Neural Style Transfer with Deep VGG model - Style Transfer using Deep Neural Network and PyTorch - Neural Style Transfer on Real Time Video (With Full Implementable Code) - How to Code Neural Style Transfer in Python? - Implementing Neural Style Transfer Using TensorFlow 2.0 - Neural Style Transfer (NST). Using Deep Learning Algorithms - Neural Read the full article

[globaltechnologysolution]

Facial Segmentation Service:-

Facial segmentation is the process of separating different parts of a face image into distinct regions or segments. This can be accomplished using machine learning techniques.

Visit: https://gts.ai/

[govindhtech]

Convolutional Neural Network & AAAI 2024 vision transformer

How Does AMD Improve AI Algorithm Hardware Efficiency?

Convolutional Neural Network(CNN) Unified Progressive Depth Pruner and AAAI 2024 Vision Transformer. Users worldwide have acknowledged AMD, one of the biggest semiconductor suppliers in the world, for its innovative chip architectural design and AI development tools. As AI advances so quickly, one of Their goals is to create high-performance algorithms that work better with AMD hardware.

Inspiration

Deep neural networks (DNNs) have achieved notable breakthroughs in a wide range of tasks, leading to impressive achievements in industrial applications. Model optimization is one of these applications that is in high demand since it can increase model inference speed while reducing accuracy trade-offs. This effort involves several methods, including effective model design, quantization, and model pruning. A common method for optimizing models in industrial applications is model trimming.

Model pruning is a major acceleration technique that aims to remove unnecessary weights intentionally while preserving accuracy. Because of sparse computation and fewer parameters, depth-wise convolutional layers provide difficulties for the traditional channel-wise pruning approach. Furthermore, channel-wise pruning techniques would make efficient models thinner and sparser, which would result in low hardware utilization and lower possible hardware efficiency.

Moreover, current model platforms favor a larger degree of parallel computation, such as GPUs. Depth Shrinker and Layer-Folding are suggested as ways to optimize MobileNetV2 in order to solve these problems by using reparameterization approaches to reduce model depth.

These techniques do have some drawbacks, though, such as the following:

The process of fine-tuning a subnet by eliminating activation layers directly may jeopardies the integrity of baseline model weights, making it more difficult to achieve high performance.

These techniques have usage restrictions.

They cannot be used to prune models that have certain normalization layers, such as Layer Norm.

Because Layer Norm is present in vision transformer models, these techniques cannot be applied to them for optimization.

Convolutional Neural Network

In order to address these issues, they suggest a depth pruning methodology that can prune Convolutional Neural Network(CNN) and vision transformer models, together with a novel block pruning method and progressive training strategy. Higher accuracy can be achieved by using the progressive training technique to transfer the baseline model structure to the subnet structure with high utilization of baseline model weights.

The current normalization layer problem can be resolved by their suggested block pruning technique, which in theory can handle all activation and normalization layers. As a result, vision transformer models can be pruned using the AMD method, which is incompatible with current depth pruning techniques.

Important Technologies

Rather than just removing the block, the AMD depth pruning approach proposes a novel block pruning strategy with reparameterization technique in an effort to reduce model depth. In block merging, the AMD block trimming technique transforms a complicated and sluggish block into a simple and fast block, as seen in Figure.

Figure : The suggested depth pruner framework by AMD. To speed up and conserve memory, each baseline block that has been pruned will progressively grow into a smaller merged block. Four baselines are tested: one vision transformer network (DeiT-Tiny) and three CNN-based networks (ResNet34, MobileNetV2, and ConvNeXtV1).

Supernet training, Subnet finding, Subnet training, and Subnet merging are the four primary phases that make up the technique. As seen in Figure , users first build a Supernet based on the basic architecture and modify blocks inside it. An ideal subnet is found via a search algorithm following Supernet training. It then use a progressive training approach that has been suggested to optimize the best Subnet with the least amount of accuracy loss. In the end, the reparameterization process would combine the Subnet into a shallower model.

Advantages

Key contributions are summarized below:

A novel block pruning strategy using reparameterization technique.

A progressive training strategy for subnet optimization.

Conducting extensive experiments on both Convolutional Neural Network(CNN) and vision transformer models to showcase the superb pruning performance provided by depth pruning method.

A unified and efficient depth pruning method for both Convolutional Neural Network(CNN) and vision transformer models.

With the AMD approach applied on ConvNeXtV1, they got three pruned ConvNeXtV1 models, which outperform popular models with similar inference performance, as illustrates, where P6 represents pruning 6 blocks of the model. Furthermore, this approach beats existing state-of-the-art methods in terms of accuracy and speedup ratio, as demonstrated. With only 1.9% top-1 accuracy reductions, the suggested depth pruner on AMD Instinct MI100 GPU accelerator achieves up to 1.26X speedup.

ConvNeXtV1 depth pruning findings on ImageNet performance. A batch size of 128 AMD Instinct MI100 GPUs is used to test speedups. Use the slowest network (EfficientFormerV2) in the table as the benchmark (1.0 speedup) for comparison.

The findings of WD-Pruning (Yu et al. 2022) and S2ViTE (Tang et al. 2022) are cited in their publication. The results of XPruner (Yu and Xiang 2023) and HVT (Pan et al. 2021), as well as SCOP (Tang et al. 2020), are not publicly available.

In summary

They have implemented this method on several Convolutional Neural Network(CNN) models and transformer models, to provide a unified depth pruner for both effective Convolutional Neural Network(CNN) and visual transformer models to prune models in the depth dimension. The benefits of this approach are demonstrated by the SOTA pruning performance. They plan to investigate the methodology on additional transformer models and workloads in the future.

Read more on Govindhtech.com

[nucotdata]

Nucot training in Data science with python and IT Service management.

[feitgemel]

Discover how to build a CNN model for skin melanoma classification using over 20,000 images of skin lesions

We'll begin by diving into data preparation, where we will organize, clean, and prepare the data form the classification model.

Next, we will walk you through the process of build and train convolutional neural network (CNN) model. We'll explain how to build the layers, and optimize the model.

Finally, we will test the model on a new fresh image and challenge our model.

Check out our tutorial here : https://youtu.be/RDgDVdLrmcs

Enjoy

Eran

#Python #Cnn #TensorFlow #deeplearning #neuralnetworks #imageclassification #convolutionalneuralnetworks #SkinMelanoma #melonomaclassification

[span-arch]

Today I'm going to talk about Urban Planning and Artifcial Intelligence. Especially about Data Bias and how sloppy labeling can affect design goals massively. This is for a get together called "Cidade Habitação e Participação - politicas publicas para construir uma cidade justa para o seculo XXI" organized by the "Laboratório de Habitación Basica" Please join us on zoom starting 4pm EST: https://us02web.zoom.us/j/84134631331 pw: city #artificialintelligence #architecture #design #agency #authorship #MachineLearning #neuralnetworks #posthuman #postdigital #school #neuralarchitecture #archdaily #spanarch #archinect #archilovers #ConvolutionalNeuralNetworks #3DgraphCNN #settlement #architecture_theory #ML #AI (at Ann Arbor, Michigan) https://www.instagram.com/p/CJJ4FyssSQj/?igshid=nv85cv807sg7

[clunite]

Using Deep Learning to detect a face. It gives much better accuracy as compare to OpenCV's Haar Cascades. ⠀ ⠀ Also, at the same time (in the background) building my custom dataset of images to further perform facial recognition and collect training data.⠀ ⠀ Thanks for Adrian @PyImageSearch for his amazing tutorials. ⠀ ⠀ ⠀ ⠀ #WeekEndGrind #WeekEnd #WeekEndProject #DeepLearning #AI #ArtificalIntelligence #MachineLearning #NerualNetwork #NN #CaffeNet #OpenCV #FaceDetection #FaceRecognition #DeepNeuralNetwork #DNN #ConvolutionalNeuralNetwork #CNN #CaffeNet #OpenCV2 #ComputerVision (at Brampton, Ontario) https://www.instagram.com/p/B5SaITAHJT5/?igshid=7s79ggft6z0u

[amitesh-ranjan]

This painting is my personal favorite. When I was conceiving the idea, I felt a strong drive to achieve beauty in intensity and I kept working on this painting for months till the point when I felt totally satisfied with the result. Here I present before you an output from running style transfer on the same artwork. A lot of appreciation and thanks goes to Ankita @bipolarraf for being so creative in her photographs. May her beauty shine forever. #science #art #neuralstyletransfer #painting #photograph #nst #technology #ai #computer #artificialintelligence #beauty #mumbai #india #styletransfer #computerscience #tensorflow #neuralnetwork #cnn #convolutionalneuralnetwork (at India Mumbai) https://www.instagram.com/p/B3OuYKXlkZ1/?igshid=juuj33mtfbaq

[aihindishow-blog]

Hello India, Welcome to India’s first podacst specialized in artificial intellegnce, machine learning and python programming. You will get one episode everyday on different topic but related to artificial intelligence. . . Follow this page for updates and to get some of our best podcast. . . . Just go to spotify and search for AI Hindi show you will get our podcast. Thank you. . . . . #artificialintelligence #machinelearning #pythonprogramming #india #hindi #pythoncode #pythonindia #indianpython #aihindishow #python3 #artificialintelligenceai #artificialinteligence #matplotlib #seaborn #numpy #linux #convolutionalneuralnetwork #neuralnetworks #neuramisdeep #machinelearningalgorithms #data #datascience #datascienceindia #bigdata #bigdataanalytics (at India) https://www.instagram.com/p/B5VCp0blbVI/?igshid=1qa5ebsmnkffs

[rcgopitechie]

Day 96 - Facing Image Region Missing On Your Photo? Let's Check The Image On Missing Pixel Filler https://www.gopichandrakesan.com/day-96-facing-image-region-missing-on-your-photo-lets-check-the-image-on-missing-pixel-filler/?feed_id=567&_unique_id=60fb17f920d6b #100dayschallenge #ArtificialIntelligence #convolutionalneuralnetwork #DeepLearning #NeuralNetwork

[lovemarketingie]

Reposting @xpstrategist: 5 Key Artificial Intelligence Predictions For 2018: How Machine Learning Will Change Everything 2018 will be the year of Artificial Intelligence (AI) and Machine Learning (ML). It will be less about hype and more about real world implementations. Here I look at some of the top tends and how they will change our lives in 2018. http://crwd.fr/2BV88Ok #ai #artificialintelligence #machinelearning #deeplearning #convolutionalneuralnetwork #technology #exponential #lovemarketing #smallbiz #smallbusiness #digitalmarketing #❤️marketing #welovemarketing #onlinemarketing #marketing #internetmarketing #marketingonline #boba #seo #socialmedia #marketingtips #google #marketingagency #hustle #mmga #growthhacking #digitaldesign #digitallife #digitalmedia #entrepreneur (at Canary Islands)

[trituenhantaoio]

#cnn #convolutionalneuralnetwork #invented #neuroscience #visualcortex https://www.instagram.com/p/B-s9e_HFwuk/?igshid=1scehu2wbu9le

[incegna]

Advanced AI: Deep Reinforcement Learning in Python.The Complete Guide to Mastering Artificial Intelligence using Deep Learning and Neural Networks.Use Convolutional Neural Networks with Deep Q-Learning. Interested people can share me your details. Grab a chance and get an additional discount. Check our Info : www.incegna.com Reg Link for Programs : http://www.incegna.com/contact-us Follow us on Facebook : www.facebook.com/INCEGNA/? Follow us on Instagram : https://www.instagram.com/_incegna/ For Queries : enquiry@incegna.com #deeplearning,#advancedartificialintelligence,#datascience,#neuralnetwork,#convolutionalneuralnetwork,#qlearning,#deepqlearning,#theano,#tensorflow,#reinforcementlearning,#machinelearningprogrammers,#machinelearningalgorithms,#pythondevelopers,#aidevelopers,#datascientist https://www.instagram.com/p/B8Gj7oVgrxG/?igshid=rgwwerukc306

[kebourne]

This Google service describe in this link (Google Nowcast) is definitely going to be useful, and that's great, but I’d like to point out the implications of how they are doing it.  Over the past year, I’ve been following closely the use of CNNs in building predictive models.  The concept is relatively new.  Traditionally, LSTM-RNNs were the defacto model for predictive models, but there is now a growing movement to use CNNs instead, which are traditionally used for image analysis / computer vision.  The idea is that you give the CNN a visual representation of the data you want to build a predictive model around and the model can pick up on more patterns than what it could typically see in just a text-intensive model line RNNs.  I haven’t seen it put to actual use yet, until now, where it looks like Google is applying this to near-term weather predictions (think Dark Skies “its going to rain in 5 minutes” kind of forecasts).  And it looks like Google is finding great success with it.  I’ve also talked with several other data scientists that are using predictive models in their research, and everyone that has tried CNNs and compared them to LSTMs has said that the CNNs performed better.  This definitely looks like the future for predictive modeling. Of course, it is much more computationally expensive, but those costs are coming down and within reach of most companies.  Being someone who does a lot of AI in constrained environments, like mobile, this also poses a challenge there.  But it's a great start and something to be very excited about in the predictive model world!