A novel memory decoder for video captioning. After obtaining representation of each frame through a pre-trained network, they first fuse the visual and lexical information. Then, at each time step, they construct a multi-layer MemNet-based decoder, i.e., in each layer, we employ a memory set to store previous information and an attention mechanism to select the information related to the current input.


🔗 http://arxiv.org/abs/2002.11886

Via: @cedeeplearning 📌
Other social media: https://linktr.ee/cedeeplearning

🔻Social media can accurately forecast economic impact of natural disaster including COVID-19 pandemic

Credit: by University of Bristol

Social media should be used to chart the economic impact and recovery of businesses in countries affected by the COVID-19 pandemic, according to new research published in Nature Communications. University of Bristol scientists describe a 'real time' method accurately trialed across three global natural disasters which could be used to reliably forecast the financial impact of the current global health crisis.
—————————————
📌Via: @cedeeplearning


https://techxplore.com/news/2020-04-social-media-accurately-economic-impact.html

#machinelearning
#socialmedia
#networkanalysis
#health
#pandemic

🔹Requisites for Operationalizing Your Machine Learning Models

there’s a lot that goes in the backend of creating a machine learning predictive model, but all of these efforts are for naught if you don’t operationalize your model effectively with a proper amount of forethought and rigor. The scoping. The preparation. The building and inferring. Each of these is a crucial initial step of the overall model lifecycle.
———————————
📌Via: @cedeeplearning
📌Other social media: https://linktr.ee/cedeeplearning

link: https://www.rocketsource.co/blog/machine-learning-models/

#machinelearning
#AI
#deeplearning
#datascience
#prediction

🔹Using LIME to Understand a Machine Learning Model’s #Predictions

Using a record explainer mechanism like Local Interpretable #Model_Agnostic Explanations (LIME) is an important technique to filter through the predicted outcomes from any machine learning model. This technique is powerful and fair because it focuses more on the inputs and outputs from the model, rather than on the model itself.
#LIME works by making small tweaks to the input #data and then observing the impact on the output data. By #filtering through the model’s findings and delivering a more digestible explanation, humans can better gauge which predictions to trust and which will be the most valuable for the organization.
———————————
📌Via: @cedeeplearning
📌Other social media: https://linktr.ee/cedeeplearning

link: https://www.rocketsource.co/blog/machine-learning-models/

#machinelearning
#datascience
#deeplearning
#AI

✔️Successfully Deploying Machine Learning Models

There are various opinions and assertions out there regarding the end-to-end process of building and deploying predictive models. We strongly assert that the deployment process is not a process at all — it’s a lifecycle. Why? It’s an infinite process of iterations and improvements. Model deployment is in no way synonymous with model completion. We will go deeper into the reasons for this in the section below as we address the requisite steps for operationalizing a model, but the high-level post-deployment steps are called out in the following diagram. Here’s what that deployment looks like in action
———————————
📌Via: @cedeeplearning
📌Other social media: https://linktr.ee/cedeeplearning

#machinelearning
#lifecycle
#deployment
#datascience
#deeplearning

👆🏻👆🏻Successfully Deploying Machine Learning Models

1. Validate Use Case
2. Data Finalization
3. Explore and Diagnose
4. Cleanse
5. Develop
6. Features
7. Build
8. Infer
9. Publish
10. Deploy
11. Consume
———————————
📌Via: @cedeeplearning

#machinelearning
#datascience
#deployment
#lifecycle
#AI
#data
#deeplearning

🔹Generative vs. Discriminative Algorithms

To understand GANs, you should know how generative #algorithms work, and for that, contrasting them with discriminative algorithms is instructive. Discriminative algorithms try to classify input data; that is, given the features of an instance of data, they predict a label or category to which that data belongs.

Another way to think about it is to distinguish discriminative from generative like this:

1. #Discriminative models learn the boundary between classes
2. #Generative models model the #distribution of individual classes
——————————
📌Via: @cedeeplearnig
📌Other social media: https://linktr.ee/cedeeplearning

link: https://pathmind.com/wiki/generative-adversarial-network-gan

#GAN
#deeplearning
#neuralnetworks
#machinelearning

A self-supervised audio-video synchronization learning method to address the problem of speaker diarization without massive labeling effort.

https://arxiv.org/abs/2002.05314

Via 📌: @CEdeeplearning
Other social media 📌: https://linktr.ee/cedeeplearning

🔻A Beginner's Guide to Convolutional Neural Networks (#CNNs)

Convolutional neural networks are neural networks used primarily to classify images (i.e. name what they see), cluster images by similarity (photo search), and perform object recognition within scenes. For example, convolutional neural networks (ConvNets or CNNs) are used to identify faces, individuals, street signs, tumors, platypuses (platypi?) and many other aspects of visual data.
————————————
📌Via: @cedeeplearning
📌Other social media: https://linktr.ee/cedeeplearning

link: https://pathmind.com/wiki/convolutional-network

#deeplearning
#neuralnetworks
#machinelearning
#math
#datascience

🔻Data for Deep Learning

🔹Types of Data:
1. sound
2. text
3. images
4. time series
5. video

🔹Use Cases:
1. classification
2. clustering
3. predictions

🔹Data Attributes:
1. relevancy
2. proper classification
3. formatting
4. accessibility

🔹Minimum Data Requirement:
The minimums vary with the complexity of the problem, but 100,000 instances in total, across all categories, is a good place to start.
———————————
📌Via: @cedeeplearning
📌Other social media: https://linktr.ee/cedeeplearning

link: https://pathmind.com/wiki/data-for-deep-learning

#deeplearning
#machinelearning
#neuralnetworks
#classification
#clustering
#data

🔹Deep Autoencoders

A deep autoencoder is composed of two, symmetrical deep-belief networks that typically have four or five shallow layers representing the encoding half of the net, and second set of four or five layers that make up the decoding half.
The layers are restricted Boltzmann machines, the #building_blocks of deep-belief networks, with several peculiarities that we’ll discuss below. Here’s a simplified schema of a deep autoencoder’s structure, which we’ll explain below.
————————————
📌Via: @cedeeplearning
📌Other social media: https://linktr.ee/cedeeplearning

link: https://pathmind.com/wiki/deep-autoencoder

#autoencoder
#deepbeliefnetwork
#neuralnetworks
#machinelearning

🔹HOW COMPUTER VISION, AI, AR AND OTHERS ARE ENHANCING IN-VEHICLE EXPERIENCES?

By Smriti Srivastava

Some latest emerging in-vehicle technologies that are changing how people interact with cars:

🔹Authentication Through Biometric

🔹In-vehicle Voice Assistant


🔹Augmented Reality for Heads-up Displays

🔹Reducing Human Error Through Vision-based Monitoring

🔹Retail and Entertainment

Tech-optimized Parking
———————————
📌Via: @cedeeplearning

https://www.analyticsinsight.net/computer-vision-ai-ar-others-enhancing-vehicle-experiences/

#selfdrivingcar
#deeplearning
#AI
#computervision

🔻GOOGLE LEVERAGES MACHINE LEARNING TO IMPROVE DOCUMENT DETECTION CAPABILITIES

Google has been employing a new scanner that uses machine learning to improve detection. Since the scanner launched, Google has boosted the detection of Office documents by 10%. Impressively, Google’s new scanner is getting better at detecting “adversarial, bursty attacks” with the detection rate jumping by 150%.
Interestingly, Google says that 58% of all malware targeting Gmail users comes from malicious documents, the vast majority of that coming from Office documents alone.
—————————————
📌Via: @cedeeplearning

https://www.analyticsinsight.net/google-leverages-machine-learning-to-improve-document-detection-capabilities/

#AI
#cybersecurity
#machinelearning
#google
#datascience

🔻AI app can detect coronavirus from sound of cough

🔹The app has a 70% accuracy rate.

Researchers have developed a new app that uses artificial intelligence technology to determine whether a person has COVID-19 based on the sound of their cough. The app has a 70% accuracy rate.

Source: EPFL

you can record your cough on a smartphone and find out whether you might have COVID-19. So how can a smartphone app detect the new coronavirus? “According to the World Health Organization, 67.7% of people who have the virus present with a dry cough – producing no mucus – as opposed to the wet cough typical of a cold or allergy,” says David Atienza, a professor at EPFL’s School of Engineering who is also the head of ESL and a member of the Coughvid development team.
———————————
📌Via: @cedeeplearning

https://neurosciencenews.com/ai-cough-coronavirus-16145/

#deeplearning
#neuralnetworks
#AI
#machinelearning
#accuracy

🔹Auto-Regressive Generative Models (#PixelRNN, #PixelCNN++)

Authors : Harshit Sharma, Saurabh Mishra

The basic difference between Generative Adversarial Networks (GANs) and Auto-regressive models is that GANs learn implicit data distribution whereas the latter learns an explicit distribution governed by a prior imposed by model structure.

Some of the advantages of Auto-regressive models over GANs are:

1. Provides a way to calculate likelihood

2. The training is more stable than GANs

3. It works for both discrete and continuous data
————————————
📌Via: @cedeeplearning

https://towardsdatascience.com/auto-regressive-generative-models-pixelrnn-pixelcnn-32d192911173

#neuralnetworks
#deeplearning
#cnn
#rnn
#GANs

Glia are turning out to be central to many neurological functions, including pain perception

https://www.quantamagazine.org/glial-brain-cells-long-in-neurons-shadow-reveal-hidden-powers-20200127/


Via: 📌@cedeeplearning
Other social media: https://linktr.ee/cedeeplearning

🔹Conditional Image Generation with PixelCNN Decoders

By Francisco Ingham

This paper explores the potential for conditional image modelling by adapting and improving a convolutional variant of the PixelRNN architecture. Pixel-CNN can be conditioned on a vector to generate similar images. This vector can be either a series of labels representing ImageNet categories or an embedding produced by a convolutional network trained on face images.
————————————
📌Via: @cedeeplearning
📌Other social media: https://linktr.ee/cedeeplearning

link: https://medium.com/a-paper-a-day-will-have-you-screaming-hurray/day-5-conditional-image-generation-with-pixelcnn-decoders-a8fc68b103a2

#deeplearning
#neuralnetworks
#machinelearning
#cnn
#pixelcnn
#pixelrnn

🧠Army develops big data approach to neuroscience - Neuroscience News
https://neurosciencenews-com.cdn.ampproject.org/c/s/neurosciencenews.com/big-data-neuroscience-15626/amp/


Via: 📌@cedeeplearning
Other social media: https://linktr.ee/cedeeplearning

🔻Intro to TensorFlow for Deep Learning (free course from Udacity)

Learn how to build deep learning applications with TensorFlow. This course was developed by the TensorFlow team and Udacity as a practical approach to deep learning for software developers. You'll get hands-on experience building your own state-of-the-art image classifiers and other deep learning models. You'll also use your #TensorFlow w models in the real world on mobile devices, in the cloud, and in browsers. Finally, you'll use advanced techniques and algorithms to work with large datasets. By the end of this course, you'll have all the skills necessary to start creating your own AI applications.
——————————————
📌Via: @cedeeplearning


https://www.udacity.com/course/intro-to-tensorflow-for-deep-learning--ud187

#free #machinelearning
#datascience #math
#deeplearning #udacity