Deep learning

---

Assalamualaikum steemians

---

How are you?Hope so everyone would be safe and sound just like me as I am also safe Alhamdulillah...

---

Deep learning is my today's topic to share with all of you guys so I would really like to delve into its depth....

Source

If I talk about deep learning then this is basically subset of machine learning which utilise artificial neural networks for the purpose of analysing and interpreting information. Deep learning is basically something which is made by taking inspiration from brain structure and function of a human and by automatic adjustment of their internal parameters it can be enhance regarding their working. This is really very useful for enabling them in recognition of some of the patterns, in classification of data and in making predictions with unprecedented preciseness.

---

Deep learning is known from 1940 if we go back into its history but in 1990 researchers started to explore multi layer neural network in 2012. There was a deep neural network which won ImageNet large scale we will recognition challenge and it was all about demonstration of potential present in deep learning.

---

If I talk about types of deep learning models then there are radius particular models which include convolutional neural network that are used for making analysis of images and videos. There are recurrent neural networks that are useful for sequential information and generative adversarial networks for the purpose of generation of any new kind of information. There are other some of the most well known models which include autoencoders, long short term memory network etc.

---

If I talk about convolutional neural networks then these are basically designed for processing data with grid like topology. This is basically using convolutional and pulling layers for the purpose of extracting the features which is followed by connected layers for classifying.In image recognizing, detection of object and segmentation convolutional neural networks have achieved state of art performance.

---

If I talk about recurrent neural networks then these are basically suitable for sequential information such as in speech, in text or in time series information. These are used for making recurrent connections for capturing temporal dependencies. There are some of the gated recurrent unit which are most famous recurrent neural network variants which are used for mitigating vanishing gradient.

Source

If I talk about generative adversarial networks then these consist of two neural networks in which first one is a generator and second one is a discriminator. Generator is used for creation of new data samples but discriminatory is used for evaluation authenticity and originally of these samples. These type of neural networks are used for generation of images,videos and music based upon reality.

---

Deep learning have played its major role in different industries like healthcare in which for making diagnosis and for medical imaging it is in use as well as in the field of finance for analysis of risk ,for optimisation of portfolios and in the field of transport for autonomous vehicle it is used.

---

If I talk about some of the limitations that deep learning is facing then these are of vanishing gradient, over fitting and computational needs as well as deep learning models can also be very difficult in terms of interpreting as well as they may also perpetuate more biasness present in training information.

If I conclude my words then deep learning is revolutionising artificial intelligence Research and application and meanwhile it is also offering and unparalleled performance in some particular but diverse domains. Last but not please I have talked about some of challenges that deep learning is facing. I invite everyone to read my post and share your feedback!

---

Thanks

---