Its crucial that the time and memory cost of running inference in a ML model is lower than the time and memory cost of training. Model compression is a strategy for reducing the cost of inference. Its main idea is to replace the original, expensive model with a smaller model that requires less memory and runtime to store and evaluate. This strategy can be used when size of the original model needs to prevent overfitting. 

Model Compression

 -> Building systems that have dynamic structure in the graph is a strategy for accelerating data-processing systems. These systems dynamically determine neural network should be run on an input. Conditional Computation refers to the form of dynamic structure inside neural networks. 

-> A strategy for speeding up an inference in a classifier is to employ a cascade of classifiers. This strategy may be used to detect the presence of a rare object.

-> Since decision tree determines which of its subtrees should be evaluated for each input, its an example of dynamic structure. Switch, where a hidden unit can receive input from diﬀerent units depending on the contex, is also a dynamic structure. 

-> One disadvantage of using dynamically structured systems is the reduced degree of parallelism that results from the system following diﬀerent code branches for diﬀerent inputs. 

Dynamic Structure

On the frontier of deep learning applications, research is being done on the representation of knowledge in relational manners to learn reasoning.

Knowledge Representation, Reasoning and Question Answering

In this node, we describe how to use deep learning to solve applications in computer vision, speech recognition, natural language processing, and other areas of commercial interest. 

University of California, Davis

Goodfellow, I., Bengio, Y., & Courville, A. (2016). $\mathit{Deep \ Learning.}$ MIT Press. Retrieved from [www.deeplearningbook.org](https://www.deeplearningbook.org) 

Deep Learning

The universal approximation theorem (Horniket al., 1989; Cybenko, 1989) states that a feedforward network with a linear output layer and at least one hidden layer with any “squashing” activation function (such as the logistic sigmoid activation function) can approximate any Borel measurable function from one ﬁnite-dimensional space to another with any desired nonzero amount of error, provided that the network is given enough hidden units. The derivatives of the feedforward network can also approximate the derivatives of the function arbitrarily well.

It means that regardless of what function we are trying to learn, we know that a large feedforward network will be able to represent this function. 

Universal Approximation Theorem

The illustration below shows the relationship between the fields of Artificial Intelligence. Deep learning is a subset of machine learning. 

Refernce : https://lupinepublishers.com/material-science-journal/pdf/MAMS.MS.ID.000138.pdf

Set relationship of Artificial Intelligence

The possibility of Deep Learning is enormous, but it also has a limitation which we need to understand for better utilization. 

a. The systems need 10,000+ examples to learn a concept
like cows. Humans only need a handful of examples.

b. Deep Learning cannot explain how the systems got an
answer.

c. Deep Learning lacks common sense. This makes the
systems fragile and when errors are made, the errors can be
very large.


Limitation of Deep Learning

- It requires very large amount of data in order to perform better than other techniques.

- It is extremely expensive to train due to complex data models. Moreover deep learning requires expensive GPUs and hundreds of machines. This increases cost to the users.

- There is no standard theory to guide you in selecting right deep learning tools as it requires knowledge of topology, training method and other parameters. As a result it is difficult to be adopted by less skilled people.

- It is not easy to comprehend output based on mere learning and requires classifiers to do so. Convolutional neural network based algorithms perform such tasks.

Drawbacks or disadvantages of Deep Learning

Applications of Deep Learning

One of the most significant advantages of deep learning is its ability to replace the labor-intensive process of manual feature engineering. In traditional machine learning, domain experts crafted specific algorithms to transform raw perceptual data into feature vectors. Deep learning replaces these domain-specific preprocessing steps with automatically tuned filters that are learned jointly from the data, yielding superior accuracy and providing a unified approach across diverse fields.

Learn Before

Related

Learn After