How far we can go in bringing innovations in the architectures of Deep Convolutional Neural Networks ?

More Asifullah Khan's questions See All

In case of intrusion detection through Machine learning, how could one tackle the polymorphism aspect of a malware?

Detecting behaviour of a malware with polymorphism attribute.

05 June 2019 7,862 4 View

In case of light weight CNN architectures, Which strategy is best?

For limited hardware resources, which ones are best? Is it the CNN architectures based on depthwise convolution or any other concept?

04 May 2019 9,354 2 View

In Machine learning, is it necessary that the training and test distribution be the same?

What if the test distribution is a little different than the training distribution.

03 April 2019 5,063 12 View

Can we train an AI system to guide us about the importance and worth of recent innovations?

I mean how can we tell that an existing idea/innovation is going to be very fruitful in next ten years or so. Can we train an AI system for this that can predict the dominance/worth of existing...

04 May 2018 1,728 2 View

Is it its possible that in future, each person has an adaptive AI System (like adaptive Deep NN) and wearable gadgets?

I think, its possible to have in future, an Adaptive AI System (like Adaptive Deep NN ) and wearable gadgets for each person. The Adaptive Deep NN will be smart enough to collect and analyse the...

04 May 2018 9,422 1 View

What is the best weight initialization strategy in Deep Convolutional Neural Notworks?

Setting initial parameters is quite important in many ML and specifically, in deep neural networks. Weights and biases are the parameters of a deep neural Network. So what is the best weight...

02 March 2018 7,954 4 View

Is ResNet a type of an ensemble Neural Network?

Resnet and other recent deep NN architectures seem to be Ensemble Neural Networks. Combiner/ MetaClassifiers

01 February 2018 9,192 1 View

Is split,Transform, and Merge concept good for Recent Convolutional Neural Networks?

It is was proposed in Inception Architecture and is being exploited in recent CNN Architectures. Is it good in combination with depth of the architecture.

31 December 2017 7,637 1 View

What could be the strategies to avoid overfitting in deep neural Networks

Deep neural networks have high representational capacity and have gained much success in recent years. However, with the ability of high representational capacity, it can also suffer from...

05 June 2017 3,600 8 View

Is deep learning good for regression problems?

Is deep Learning ( deep neural networks, convolutional deep neural networks, etc.) good for regression problems?

01 February 2015 6,810 14 View

What is the most Operating System (OS) which can be INSTALL on Tiny hard devices like Raspberry Pi (RPi) hardware ?

I'm targeting to deploy a mesh network and manually configure MANET routing protocols. I'm preparing scenarios, architectures, and hard devices needed to do that. Are there some step-by-step...

03 March 2021 1,931 5 View

Why should we choose deep learning to machine learning? Why is CNN better for image classification? Can I get Typical Answer?

What Characteristics makes CNN work better?

03 March 2021 1,458 4 View

What are some of the research gaps in machine learning and artificial intelligence in africa?

i would to know some of the research gaps in the artificial intelligence field in most african countries.

03 March 2021 6,145 3 View

Anyone plz suggest relevant topic for research in image processing+Deep learning.?

I have selected brain tumor images ...but now found that already lots of research done n this topic.

03 March 2021 5,774 3 View

How to increase a model accuracy ?

dear community, my model is based feature extraction from non stationary signals using discrete Wavelet Transform and then using statistical features then machine learning classifiers in order to...

03 March 2021 6,994 5 View

Is there a powerful system for the security of the systems distributed on IoT systems?

02 March 2021 3,858 10 View

Can you share me how teacher education is managed in your respective national educational system?

I feel that the practice in teacher education in my country is below the expected performance level due to very poor management system. Hope I will learn something from your experiences.

02 March 2021 1,516 4 View

How Relativistic effects is one reason for lanthanide contraction?

The Lanthanide contraction which is the decrease in ionic radii of the elements in the lanthanide series from atomic number 57, lanthanum, to 71, lutetium, is due to poor shielding of the 4f...

01 March 2021 2,272 4 View

Can NFL theorem be valid in infinite search space in ML?

NFL theorem is valid for algorithms training in fixed training set. However, the general characteristic of algorithms in expanded or open dataset has not been proved yet. Could you show your...

01 March 2021 1,189 3 View

Is stone a sustainable material? Is it a sustainable approach to use natural stone waste in architecture?

I'm designing a zero-waste restaurant and considering using marble waste to make some tables and a bar countertop. My clients think it's not "sustainable enough" and I would like to have more...

01 March 2021 4,326 7 View

Michael George

There is no fixed number it's all based on the use case,

Basically you need to consider two things before building a large network:

server processing capacity (that will tells you how powerfully processing support you can count on to increase your nodes and layers) for that super processing is needed and maybe quantum in future. and this point we can say it's costing problem which might challenge the features increment.

how many hidden layers you need to add before you lose the control on output quality, in other word, does the hidden layers perform good output with the minimum possible number? coz in some cases the massive network can make random results which is hard to understand how to fix it even. it will be matter of probabilities especially if some neurons fetch their values from huge and sensitive analysis.

and out of this two points, there is no limit to go deep with the network.

Muhammad Aamir

its depends on your computational power rest there is no limit. Deep , deeper or deepen all you need is machine to handle it.

Asifullah Khan

Yes but my point is not depth related only, rather generic. The different architectural innovations that are coming in; Concept of High-Way Networks (multiple controlled paths), ResNet (Skip connections to address the problem of vanishing gradients), Split-Transform-Merge concept (Inception), Squeeze and Excitation concept (exploiting multi-channel based information for for achieving discerning capabilities and adjusting accordingly).