Neural networks

---

Assalamualaikum steemians

---

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

---

Neural network is my today's topic to share with all of you guys so I would like to delve into its various aspects.

Source

Neural networks are basically computational models which are made by getting inspiration from structure of brain and function of brain of human as well as these networks are made up of interconnected nodes which are connected with the each other and hence they are used for processing and transmitting of information.

---

If I move back into the history of neural networks then its concept arrives in 1940 but in world of computational power and algorithms advancements made later that are in growth recently. Some of the significant milestones achieved during advancement and its growth was the development of perceptron that occurred in 1957 and back propagation that occurred in 1986.

---

Neural network is not generalized. If I talk about its types then there are different types and some of them are here in my words. There are feed forward networks in which information always flow in a forward direction like from input to output. There are recurrent neural networks in which information flows in form of loops which in turn used to enable temporal dependency. Last but not least convolutional neural networks comes that are specifically made for imaging and signaling process.

---

If I talk about architecture of neural networks then first of all input layer comes which is used for receiving data and then hidden layers is used for processing of sensitive information and last but not least output layer is used for generation of predictions. If I talk about learning system of neural networks then they learn by optimisation of algorithm such as stochastic gradient descent which is used for adjustment of biasness for minimization of errors.

---

More than this there are some forms of learning in which first of all there is supervised learning in which there are noodle networks that learnt from data which is labelled and optimisation of performance at a specific task like in classification of image. Secondly there is on supervised learning in which neural networks discover different patterns and relationships in that data that are unlabeled like in the form of clustering.

---

If I talk about applications of neural networks than hair comes the information in which neural networks may be applied so in computer vision like for recognition of type of image and which object is present can be detected by neural network. If I talk about natural language processing then here neural networks also applied for analysis of text and for translation of different type of languages. Last but not least in controlling and navigation and in field of robotics it has wide role.

Source

If I talk about some of the benefits that we can obtain from neural networks then first of all we can recognise patterns like neural networks are helpful in identification of some of the complex patterns as well as these can also provide flexibility in terms of learning from diverse data types and these networks are able to handle large sets of data so this is another plus point.

---

If there are plus points then definitely there are some of the limitations that exist side by side so these networks make it difficult to understand decisions of neural networks so in this way interpretability issue as well as training time is another issue in which these networks need significant computational resources as well as these networks also memorize training data so over fitting is another challenge.

At last I want to explain in a brief way that there are some of the neural network tools and frameworks like if I talk about tensor flow then it is an open source framework as well as similarly there is another neural network tool comes which is pyTorch and this is most famous deep learning library and last but not least keras come with high level neural network API.

That was all about neural networks. I am grateful for own who give their time to read my post and I will be grateful if you all would give feedback about my content!

---

Thanks

---