Shrinking-deep-learnings-carbon-footprint-MIT-News.jpg

Shrinking deep learning’s carbon footprint | MIT News

news .mit.edu/sites/default/files/styles/news_article__cover_image__original/public/images/202008/brain-computing.jpg?itok=GZ230mkN” />

In June, OpenAI unveiled the largest language model in the world, a text-generating tool called GPT-3 that can write creative fiction, translate legalese into plain English, and answer obscure trivia questions. It’s the latest feat of intelligence achieved by deep learning, a machine learning method patterned after the way neurons in the brain process and store information.

But it came at a hefty price: at least $4.6 million and 355 years in computing time, assuming the model was trained on a standard neural network chip, or GPU. The model’s colossal size — 1,000 times larger than a typical language model — is the main factor in its high cost.

“You have to throw a lot more computation at something to get a little improvement in performance,” says Neil Thompson, an MIT researcher who has tracked deep learning’s unquenchable thirst for computing. “It’s unsustainable. We have to find more efficient ways to scale deep learning or develop other technologies.”

Some of the excitement over AI’s recent progress has shifted to alarm. In a study last year, researchers at the University of Massachusetts at Amherst estimated that training a large deep-learning model produces 626,000 pounds of planet-warming carbon dioxide, equal to the lifetime emissions of five cars. As models grow bigger, their demand for computing is outpacing improvements in hardware efficiency. Chips specialized for neural-network processing, like GPUs (graphics processing units) and TPUs (tensor processing units), have offset the demand for more computing, but not by enough. 

“We need to rethink the entire stack — from software to hardware,” says Aude Oliva, MIT director of the MIT-IBM Watson AI Lab and co-director of the MIT Quest for Intelligence. “Deep learning has made the recent AI revolution possible, but its growing cost in energy and carbon emissions is untenable.”

Computational limits have dogged neural networks from their earliest incarnation — the perceptron — in the 1950s. As computing power exploded, and the internet unleashed a tsunami of data, they evolved into powerful engines for pattern recognition and prediction. But each new milestone brought an explosion in cost, as data-hungry models demanded increased computation. GPT-3, for example, trained on half a trillion words and ballooned to 175 billion parameters — the mathematical operations, or weights, that tie the model together — making it 100 times bigger than its predecessor, itself just a year old.

In work posted on the pre-print server arXiv, Thompson and his colleagues show that the ability of deep learning models to surpass key benchmarks tracks their nearly exponential rise in computing power use. (Like others seeking to track AI’s carbon footprint, the team had to guess at many models’ energy consumption due to a lack of reporting requirements). At this rate, the researchers argue, deep nets will survive only if they, and the hardware they run on, become radically more efficient.

Toward leaner, greener algorithms

The human perceptual system is extremely efficient at using data. Researchers have borrowed this idea for recognizing actions in video and in real life to make models more compact. In a paper at the European Conference on Computer Vision (ECCV) in August, researchers at the MIT-IBM Watson AI Lab describe a method for unpacking a scene from a few glances, as humans do, by cherry-picking the most relevant data.

Take a video clip of someone making a sandwich. Under the method outlined in the paper, a policy network strategically picks frames of the knife slicing through roast beef, and meat being stacked on a slice of bread, to represent at high resolution. Less-relevant frames are skipped over or represented at lower resolution. A second model then uses the abbreviated CliffsNotes version of the movie to label it “making a sandwich.” The approach leads to faster video classification at half the computational cost as the next-best model, the researchers say.

“Humans don’t pay attention to every last detail — why should our models?” says the study’s senior author, Rogerio Feris, research manager at the MIT-IBM Watson AI Lab. “We can use machine learning to adaptively select the right data, at the right level of detail, to make deep learning models more efficient.”

In a complementary approach, researchers are using deep learning itself to design more economical models through an automated process known as neural architecture search. Song Han, an assistant professor at MIT, has used automated search to design models with fewer weights, for language understanding and scene recognition, where picking out looming obstacles quickly is acutely important in driving applications. 

In a paper at ECCV, Han and his colleagues propose a model architecture for three-dimensional scene recognition that can spot safety-critical details like road signs, pedestrians, and cyclists with relatively less computation. They used an evolutionary-search algorithm to evaluate 1,000 architectures before settling on a model they say is three times faster and uses eight times less computation than the next-best method. 

In another recent paper, they use evolutionary search within an augmented designed space to find the most efficient architectures for machine translation on a specific device, be it a GPU, smartphone, or tiny Raspberry Pi. Separating the search and training process leads to huge reductions in computation, they say.

In a third approach, researchers are probing the essence of deep nets to see if it might be possible to train a small part of even hyper-efficient networks like those above. Under their proposed lottery ticket hypothesis, PhD student Jonathan Frankle and MIT Professor Michael Carbin proposed that within each model lies a tiny subnetwork that could have been trained in isolation with as few as one-tenth as many weights — what they call a “winning ticket.” 

They showed that an algorithm could retroactively find these winning subnetworks in small image-classification models. Now, in a paper at the International Conference on Machine Learning (ICML), they show that the algorithm finds winning tickets in large models, too; the models just need to be rewound to an early, critical point in training when the order of the training data no longer influences the training outcome. 

In less than two years, the lottery ticket idea has been cited more than more than 400 times, including by Facebook researcher Ari Morcos, who has shown that winning tickets can be transferred from one vision task to another, and that winning tickets exist in language and reinforcement learning models, too. 

“The standard explanation for why we need such large networks is that overparameterization aids the learning process,” says Morcos. “The lottery ticket hypothesis disproves that — it’s all about finding an appropriate starting point. The big downside, of course, is that, currently, finding these ‘winning’ starting points requires training the full overparameterized network anyway.”

Frankle says he’s hopeful that an efficient way to find winning tickets will be found. In the meantime, recycling those winning tickets, as Morcos suggests, could lead to big savings.

Hardware designed for efficient deep net algorithms

As deep nets push classical computers to the limit, researchers are pursuing alternatives, from optical computers that transmit and store data with photons instead of electrons, to quantum computers, which have the potential to increase computing power exponentially by representing data in multiple states at once.

Until a new paradigm emerges, researchers have focused on adapting the modern chip to the demands of deep learning. The trend began with the discovery that video -game graphical chips, or GPUs, could turbocharge deep-net training with their ability to perform massively parallelized matrix computations. GPUs are now one of the workhorses of modern AI, and have spawned new ideas for boosting deep net efficiency through specialized hardware. 

Much of this work hinges on finding ways to store and reuse data locally, across the chip’s processing cores, rather than waste time and energy shuttling data to and from a designated memory site. Processing data locally not only speeds up model training but improves inference, allowing deep learning applications to run more smoothly on smartphones and other mobile devices.

Vivienne Sze, a professor at MIT, has literally written the book on efficient deep nets. In collaboration with book co-author Joel Emer, an MIT professor and researcher at NVIDIA, Sze has designed a chip that’s flexible enough to process the widely-varying shapes of both large and small deep learning models. Called Eyeriss 2, the chip uses 10 times less energy than a mobile GPU.

Its versatility lies in its on-chip network, called a hierarchical mesh, that adaptively reuses data and adjusts to the bandwidth requirements of different deep learning models. After reading from memory, it reuses the data across as many processing elements as possible to minimize data transportation costs and maintain high throughput. 

“The goal is to translate small and sparse networks into energy savings and fast inference,” says Sze. “But the hardware should be flexible enough to also efficiently support large and dense deep neural networks.”

Other hardware innovators are focused on reproducing the brain’s energy efficiency. Former Go world champion Lee Sedol may have lost his title to a computer, but his performance was fueled by a mere 20 watts of power. AlphaGo, by contrast, burned an estimated megawatt of energy, or 500,000 times more.

Inspired by the brain’s frugality, researchers are experimenting with replacing the binary, on-off switch of classical transistors with analog devices that mimic the way that synapses in the brain grow stronger and weaker during learning and forgetting.

An electrochemical device, developed at MIT and recently published in Nature Communications, is modeled after the way resistance between two neurons grows or subsides as calcium, magnesium or potassium ions flow across the synaptic membrane dividing them. The device uses the flow of protons — the smallest and fastest ion in solid state — into and out of a crystalline lattice of tungsten trioxide to tune its resistance along a continuum, in an analog fashion.

“Even though the device is not yet optimized, it gets to the order of energy consumption per unit area per unit change in conductance that’s close to that in the brain,” says the study’s senior author, Bilge Yildiz, a professor at MIT.

Energy-efficient algorithms and hardware can shrink AI’s environmental impact. But there are other reasons to innovate, says Sze, listing them off: Efficiency will allow computing to move from data centers to edge devices like smartphones, making AI accessible to more people around the world; shifting computation from the cloud to personal devices reduces the flow, and potential leakage, of sensitive data; and processing data on the edge eliminates transmission costs, leading to faster inference with a shorter reaction time, which is key for interactive driving and augmented/virtual reality applications.

“For all of these reasons, we need to embrace efficient AI,” she says.

Covid Abruzzo Basilicata Calabria Campania Emilia Romagna Friuli Venezia Giulia Lazio Liguria Lombardia Marche Molise Piemonte Puglia Sardegna Sicilia Toscana Trentino Alto Adige Umbria Valle d’Aosta Veneto Italia Agrigento Alessandria Ancona Aosta Arezzo Ascoli Piceno Asti Avellino Bari Barletta-Andria-Trani Belluno Benevento Bergamo Biella Bologna Bolzano Brescia Brindisi Cagliari Caltanissetta Campobasso Carbonia-Iglesias Caserta Catania Catanzaro Chieti Como Cosenza Cremona Crotone Cuneo Enna Fermo Ferrara Firenze Foggia Forlì-Cesena Frosinone Genova Gorizia Grosseto Imperia Isernia La Spezia L’Aquila Latina Lecce Lecco Livorno Lodi Lucca Macerata Mantova Massa-Carrara Matera Messina Milano Modena Monza e della Brianza Napoli Novara Nuoro Olbia-Tempio Oristano Padova Palermo Parma Pavia Perugia Pesaro e Urbino Pescara Piacenza Pisa Pistoia Pordenone Potenza Prato Ragusa Ravenna Reggio Calabria Reggio Emilia Rieti Rimini Roma Rovigo Salerno Medio Campidano Sassari Savona Siena Siracusa Sondrio Taranto Teramo Terni Torino Ogliastra Trapani Trento Treviso Trieste Udine Varese Venezia Verbano-Cusio-Ossola Vercelli Verona Vibo Valentia Vicenza Viterbo

Deep-Learning-Models-for-Multi-Output-Regression.jpg

Deep Learning Models for Multi-Output Regression

Last Updated on August 28, 2020

Multi-output regression involves predicting two or more numerical variables.

Unlike normal regression where a single value is predicted for each sample, multi-output regression requires specialized machine learning algorithms that support outputting multiple variables for each prediction.

Deep learning neural networks are an example of an algorithm that natively supports multi-output regression problems. Neural network models for multi-output regression tasks can be easily defined and evaluated using the Keras deep learning library.

In this tutorial, you will discover how to develop deep learning models for multi-output regression.

After completing this tutorial, you will know:

  • Multi-output regression is a predictive modeling task that involves two or more numerical output variables.
  • Neural network models can be configured for multi-output regression tasks.
  • How to evaluate a neural network for multi-output regression and make a prediction for new data.

Let’s get started.

Deep Learning Models for Multi-Output Regression

Deep Learning Models for Multi-Output Regression
Photo by Christian Collins, some rights reserved.

Tutorial Overview

This tutorial is divided into three parts; they are:

  • Multi-Output Regression
  • Neural Networks for Multi-Outputs
  • Neural Network for Multi-Output Regression
  • Multi-Output Regression

    Regression is a predictive modeling task that involves predicting a numerical output given some input.

    It is different from classification tasks that involve predicting a class label.

    Typically, a regression task involves predicting a single numeric value. Although, some tasks require predicting more than one numeric value. These tasks are referred to as multiple-output regression, or multi-output regression for short.

    In multi-output regression, two or more outputs are required for each input sample, and the outputs are required simultaneously. The assumption is that the outputs are a function of the inputs.

    We can create a synthetic multi-output regression dataset using the make_regression() function in the scikit-learn library.

    Our dataset will have 1,000 samples with 10 input features, five of which will be relevant to the output and five of which will be redundant. The dataset will have three numeric outputs for each sample.

    The complete example of creating and summarizing the synthetic multi-output regression dataset is listed below.


    Running the example creates the dataset and summarizes the shape of the input and output elements.

    We can see that, as expected, there are 1,000 samples, each with 10 input features and three output features.


    Next, let’s look at how we can develop neural network models for multiple-output regression tasks.

    Neural Networks for Multi-Outputs

    Many machine learning algorithms support multi-output regression natively.

    Popular examples are decision trees and ensembles of decision trees. A limitation of decision trees for multi-output regression is that the relationships between inputs and outputs can be blocky or highly structured based on the training data.

    Neural network models also support multi-output regression and have the benefit of learning a continuous function that can model a more graceful relationship between changes in input and output.

    Multi-output regression can be supported directly by neural networks simply by specifying the number of target variables there are in the problem as the number of nodes in the output layer. For example, a task that has three output variables will require a neural network output layer with three nodes in the output layer, each with the linear (default) activation function.

    We can demonstrate this using the Keras deep learning library.

    We will define a multilayer perceptron (MLP) model for the multi-output regression task defined in the previous section.

    Each sample has 10 inputs and three outputs, therefore, the network requires an input layer that expects 10 inputs specified via the “input_dim” argument in the first hidden layer and three nodes in the output layer.

    We will use the popular ReLU activation function in the hidden layer. The hidden layer has 20 nodes, which were chosen after some trial and error. We will fit the model using mean absolute error (MAE) loss and the Adam version of stochastic gradient descent.

    The definition of the network for the multi-output regression task is listed below.


    You may want to adapt this model for your own multi-output regression task, therefore, we can create a function to define and return the model where the number of input and number of output variables are provided as arguments.


    Now that we are familiar with how to define an MLP for multi-output regression, let’s explore how this model can be evaluated.

    Neural Network for Multi-Output Regression

    If the dataset is small, it is good practice to evaluate neural network models repeatedly on the same dataset and report the mean performance across the repeats.

    This is because of the stochastic nature of the learning algorithm.

    Additionally, it is good practice to use k-fold cross-validation instead of train/test splits of a dataset to get an unbiased estimate of model performance when making predictions on new data. Again, only if there is not too much data and the process can be completed in a reasonable time.

    Taking this into account, we will evaluate the MLP model on the multi-output regression task using repeated k-fold cross-validation with 10 folds and three repeats.

    Each fold the model is defined, fit, and evaluated. The scores are collected and can be summarized by reporting the mean and standard deviation.

    The evaluate_model() function below takes the dataset, evaluates the model, and returns a list of evaluation scores, in this case, MAE scores.


    We can then load our dataset and evaluate the model and report the mean performance.

    Tying this together, the complete example is listed below.


    Running the example reports the MAE for each fold and each repeat, to give an idea of the evaluation progress.

    Note: Your results may vary given the stochastic nature of the algorithm or evaluation procedure, or differences in numerical precision. Consider running the example a few times and compare the average outcome.

    At the end, the mean and standard deviation MAE is reported. In this case, the model is shown to achieve a MAE of about 8.184.

    You can use this code as a template for evaluating MLP models on your own multi-output regression tasks. The number of nodes and layers in the model can easily be adapted and tailored to the complexity of your dataset.


    Once a model configuration is chosen, we can use it to fit a final model on all available data and make a prediction for new data.

    The example below demonstrates this by first fitting the MLP model on the entire multi-output regression dataset, then calling the predict() function on the saved model in order to make a prediction for a new row of data.


    Running the example fits the model and makes a prediction for a new row.

    Note: Your results may vary given the stochastic nature of the algorithm or evaluation procedure, or differences in numerical precision. Consider running the example a few times and compare the average outcome.

    As expected, the prediction contains three output variables required for the multi-output regression task.


    Further Reading

    This section provides more resources on the topic if you are looking to go deeper.

    Summary

    In this tutorial, you discovered how to develop deep learning models for multi-output regression.

    Specifically, you learned:

    • Multi-output regression is a predictive modeling task that involves two or more numerical output variables.
    • Neural network models can be configured for multi-output regression tasks.
    • How to evaluate a neural network for multi-output regression and make a prediction for new data.

    Do you have any questions?
    Ask your questions in the comments below and I will do my best to answer.

    Develop Deep Learning Projects with Python!

    Deep Learning with Python
     What If You Could Develop A Network in Minutes

    …with just a few lines of Python

    Discover how in my new Ebook:
    Deep Learning With Python

    It covers end-to-end projects on topics like:
    Multilayer Perceptrons, Convolutional Nets and Recurrent Neural Nets, and more…

    Finally Bring Deep Learning To

    Your Own Projects

    Skip the Academics. Just Results.

    See What’s Inside

    Covid Abruzzo Basilicata Calabria Campania Emilia Romagna Friuli Venezia Giulia Lazio Liguria Lombardia Marche Molise Piemonte Puglia Sardegna Sicilia Toscana Trentino Alto Adige Umbria Valle d’Aosta Veneto Italia Agrigento Alessandria Ancona Aosta Arezzo Ascoli Piceno Asti Avellino Bari Barletta-Andria-Trani Belluno Benevento Bergamo Biella Bologna Bolzano Brescia Brindisi Cagliari Caltanissetta Campobasso Carbonia-Iglesias Caserta Catania Catanzaro Chieti Como Cosenza Cremona Crotone Cuneo Enna Fermo Ferrara Firenze Foggia Forlì-Cesena Frosinone Genova Gorizia Grosseto Imperia Isernia La Spezia L’Aquila Latina Lecce Lecco Livorno Lodi Lucca Macerata Mantova Massa-Carrara Matera Messina Milano Modena Monza e della Brianza Napoli Novara Nuoro Olbia-Tempio Oristano Padova Palermo Parma Pavia Perugia Pesaro e Urbino Pescara Piacenza Pisa Pistoia Pordenone Potenza Prato Ragusa Ravenna Reggio Calabria Reggio Emilia Rieti Rimini Roma Rovigo Salerno Medio Campidano Sassari Savona Siena Siracusa Sondrio Taranto Teramo Terni Torino Ogliastra Trapani Trento Treviso Trieste Udine Varese Venezia Verbano-Cusio-Ossola Vercelli Verona Vibo Valentia Vicenza Viterbo

    Multi-Label-Classification-with-Deep-Learning.jpg

    Multi-Label Classification with Deep Learning

    Last Updated on August 31, 2020

    Multi-label classification involves predicting zero or more class labels.

    Unlike normal classification tasks where class labels are mutually exclusive, multi-label classification requires specialized machine learning algorithms that support predicting multiple mutually non-exclusive classes or “labels.”

    Deep learning neural networks are an example of an algorithm that natively supports multi-label classification problems. Neural network models for multi-label classification tasks can be easily defined and evaluated using the Keras deep learning library.

    In this tutorial, you will discover how to develop deep learning models for multi-label classification.

    After completing this tutorial, you will know:

    • Multi-label classification is a predictive modeling task that involves predicting zero or more mutually non-exclusive class labels.
    • Neural network models can be configured for multi-label classification tasks.
    • How to evaluate a neural network for multi-label classification and make a prediction for new data.

    Let’s get started.

    Multi-Label Classification with Deep Learning

    Multi-Label Classification with Deep Learning
    Photo by Trevor Marron, some rights reserved.

    Tutorial Overview

    This tutorial is divided into three parts; they are:

    • Multi-Label Classification
    • Neural Networks for Multiple Labels
    • Neural Network for Multi-Label Classification

    Multi-Label Classification

    Classification is a predictive modeling problem that involves outputting a class label given some input

    It is different from regression tasks that involve predicting a numeric value.

    Typically, a classification task involves predicting a single label. Alternately, it might involve predicting the likelihood across two or more class labels. In these cases, the classes are mutually exclusive, meaning the classification task assumes that the input belongs to one class only.

    Some classification tasks require predicting more than one class label. This means that class labels or class membership are not mutually exclusive. These tasks are referred to as multiple label classification, or multi-label classification for short.

    In multi-label classification, zero or more labels are required as output for each input sample, and the outputs are required simultaneously. The assumption is that the output labels are a function of the inputs.

    We can create a synthetic multi-label classification dataset using the make_multilabel_classification() function in the scikit-learn library.

    Our dataset will have 1,000 samples with 10 input features. The dataset will have three class label outputs for each sample and each class will have one or two values (0 or 1, e.g. present or not present).

    The complete example of creating and summarizing the synthetic multi-label classification dataset is listed below.


    Running the example creates the dataset and summarizes the shape of the input and output elements.

    We can see that, as expected, there are 1,000 samples, each with 10 input features and three output features.

    The first 10 rows of inputs and outputs are summarized and we can see that all inputs for this dataset are numeric and that output class labels have 0 or 1 values for each of the three class labels.


    Next, let’s look at how we can develop neural network models for multi-label classification tasks.

    Neural Networks for Multiple Labels

    Some machine learning algorithms support multi-label classification natively.

    Neural network models can be configured to support multi-label classification and can perform well, depending on the specifics of the classification task.

    Multi-label classification can be supported directly by neural networks simply by specifying the number of target labels there is in the problem as the number of nodes in the output layer. For example, a task that has three output labels (classes) will require a neural network output layer with three nodes in the output layer.

    Each node in the output layer must use the sigmoid activation. This will predict a probability of class membership for the label, a value between 0 and 1. Finally, the model must be fit with the binary cross-entropy loss function.

    In summary, to configure a neural network model for multi-label classification, the specifics are:

    • Number of nodes in the output layer matches the number of labels.
    • Sigmoid activation for each node in the output layer.
    • Binary cross-entropy loss function.

    We can demonstrate this using the Keras deep learning library.

    We will define a Multilayer Perceptron (MLP) model for the multi-label classification task defined in the previous section.

    Each sample has 10 inputs and three outputs; therefore, the network requires an input layer that expects 10 inputs specified via the “input_dim” argument in the first hidden layer and three nodes in the output layer.

    We will use the popular ReLU activation function in the hidden layer. The hidden layer has 20 nodes that were chosen after some trial and error. We will fit the model using binary cross-entropy loss and the Adam version of stochastic gradient descent.

    The definition of the network for the multi-label classification task is listed below.


    You may want to adapt this model for your own multi-label classification task; therefore, we can create a function to define and return the model where the number of input and output variables is provided as arguments.


    Now that we are familiar with how to define an MLP for multi-label classification, let’s explore how this model can be evaluated.

    Neural Network for Multi-Label Classification

    If the dataset is small, it is good practice to evaluate neural network models repeatedly on the same dataset and report the mean performance across the repeats.

    This is because of the stochastic nature of the learning algorithm.

    Additionally, it is good practice to use k-fold cross-validation instead of train/test splits of a dataset to get an unbiased estimate of model performance when making predictions on new data. Again, only if there is not too much data that the process can be completed in a reasonable time.

    Taking this into account, we will evaluate the MLP model on the multi-output regression task using repeated k-fold cross-validation with 10 folds and three repeats.

    The MLP model will predict the probability for each class label by default. This means it will predict three probabilities for each sample. These can be converted to crisp class labels by rounding the values to either 0 or 1. We can then calculate the classification accuracy for the crisp class labels.


    The scores are collected and can be summarized by reporting the mean and standard deviation across all repeats and cross-validation folds.

    The evaluate_model() function below takes the dataset, evaluates the model, and returns a list of evaluation scores, in this case, accuracy scores.


    We can then load our dataset and evaluate the model and report the mean performance.

    Tying this together, the complete example is listed below.


    Running the example reports the classification accuracy for each fold and each repeat, to give an idea of the evaluation progress.

    Note: Your results may vary given the stochastic nature of the algorithm or evaluation procedure, or differences in numerical precision. Consider running the example a few times and compare the average outcome.

    At the end, the mean and standard deviation accuracy is reported. In this case, the model is shown to achieve an accuracy of about 81.2 percent.

    You can use this code as a template for evaluating MLP models on your own multi-label classification tasks. The number of nodes and layers in the model can easily be adapted and tailored to the complexity of your dataset.


    Once a model configuration is chosen, we can use it to fit a final model on all available data and make a prediction for new data.

    The example below demonstrates this by first fitting the MLP model on the entire multi-label classification dataset, then calling the predict() function on the saved model in order to make a prediction for a new row of data.


    Running the example fits the model and makes a prediction for a new row. As expected, the prediction contains three output variables required for the multi-label classification task: the probabilities of each class label.


    Further Reading

    This section provides more resources on the topic if you are looking to go deeper.

    Summary

    In this tutorial, you discovered how to develop deep learning models for multi-label classification.

    Specifically, you learned:

    • Multi-label classification is a predictive modeling task that involves predicting zero or more mutually non-exclusive class labels.
    • Neural network models can be configured for multi-label classification tasks.
    • How to evaluate a neural network for multi-label classification and make a prediction for new data.

    Do you have any questions?
    Ask your questions in the comments below and I will do my best to answer.

    Develop Deep Learning Projects with Python!

    Deep Learning with Python
     What If You Could Develop A Network in Minutes

    …with just a few lines of Python

    Discover how in my new Ebook:
    Deep Learning With Python

    It covers end-to-end projects on topics like:
    Multilayer Perceptrons, Convolutional Nets and Recurrent Neural Nets, and more…

    Finally Bring Deep Learning To

    Your Own Projects

    Skip the Academics. Just Results.

    See What’s Inside

    Covid Abruzzo Basilicata Calabria Campania Emilia Romagna Friuli Venezia Giulia Lazio Liguria Lombardia Marche Molise Piemonte Puglia Sardegna Sicilia Toscana Trentino Alto Adige Umbria Valle d’Aosta Veneto Italia Agrigento Alessandria Ancona Aosta Arezzo Ascoli Piceno Asti Avellino Bari Barletta-Andria-Trani Belluno Benevento Bergamo Biella Bologna Bolzano Brescia Brindisi Cagliari Caltanissetta Campobasso Carbonia-Iglesias Caserta Catania Catanzaro Chieti Como Cosenza Cremona Crotone Cuneo Enna Fermo Ferrara Firenze Foggia Forlì-Cesena Frosinone Genova Gorizia Grosseto Imperia Isernia La Spezia L’Aquila Latina Lecce Lecco Livorno Lodi Lucca Macerata Mantova Massa-Carrara Matera Messina Milano Modena Monza e della Brianza Napoli Novara Nuoro Olbia-Tempio Oristano Padova Palermo Parma Pavia Perugia Pesaro e Urbino Pescara Piacenza Pisa Pistoia Pordenone Potenza Prato Ragusa Ravenna Reggio Calabria Reggio Emilia Rieti Rimini Roma Rovigo Salerno Medio Campidano Sassari Savona Siena Siracusa Sondrio Taranto Teramo Terni Torino Ogliastra Trapani Trento Treviso Trieste Udine Varese Venezia Verbano-Cusio-Ossola Vercelli Verona Vibo Valentia Vicenza Viterbo

    Recent Posts

    Archives