Week 2 Progress - General Research - Neural Networks

Summary:

I researched Neural networks using three main resources listed at the end of this page, I learned a great deal of overarching concepts and tried to understand them. I have summarized the main points of my learning on general neural networks below. Later I will be delving into neural networks in Python. There are very complex mathematical algorithms involved in the actual neurons and they went a little bit over my head but I will be making an effort to understand them as I move forward and I think once they are translated into python code I will have an easier time with them although no doubt they will still be extremely complex. Below I have also listed the next steps for what I will be trying to learn next!


Overview of Concepts:

Neural Networks are specific tools used within the field of machine learning for a variety of very useful tasks, they are not meant to simulate a human brain but to use the way in which we understand the brain in order to create a program that can learn things in some cases better than a human can. For instance neural networks have been used to learn how to identify unhealthy patients by looking at pictures of thousands of healthy and unhealthy patients. And in some cases they can accomplish these tasks better than humans have been able to alone. Neural networks simulate the overall concept of neurons firing in the brain and use that process in a similar way to how human brains learn to learn a very specific set task. 

 

Supervised Vs. Unsupervised

-Supervised algorithms are algorithms where experts specify which data will lead to the best results, a more traditional method.

-Unsupervised algorithms are ones that have the algorithm specify it's own value on data and can often lead to better results than supervised, a newer method.

 

Neural Network Structure:

a simple neural network 

 Intel Corporation. (2018). “A Simple Neural Network”. [Diagram]. Retrieved from https://newsroom.intel.com/news/many-ways-define-artificial-intelligence/#gs.b18dny

-A Neural network is made of nodes (neurons), the circles in the above image, as well as connections between them.

-Neural networks run cyclically, meaning they run through the network over and over refining which responses are better each time and in this way being trained, learning. 

-The more iterations the better refined the responses are generally.

-Deep learning algorithms refer to the number of layers of neurons in a Neural network, the deeper the number of layers the deeper the learning.

-Computation occurs within nodes and that computation determines what output from that node should be chosen.


Algorithm writing and training:

-Selecting proper training datasets is very important to gaining the desired output, training involves running the algorithm over and over and error correcting each time until it appears to be learning correctly.

-Different weights are assigned to different variables that are identified as important to the specific task, the higher the weight the more that feature will affect the neural network's conclusion.

-Multi-layer perceptions with many different neurons per layer are the most commonly used neural networks today although they were more simple in the past.


Next Steps / Future steps:

-I will try to dive deeper into neural networks by looking more into the inner workings of the networks and try to further understand internal neuron algorithms.

-I will start to learn about neural networks from a Python language perspective and attempt to apply my knowledge of basic Python programming to make sense of simple networks.

-After that I will attempt to look at the actual python code of simple neural networks and try to dissect and understand it as much as possible.


Resources Used:

 I found and used the following resources to various extents, some will be used more thoroughly in future weeks as I refine my learning.

 

-Intel Corporation. (2018). The Many Ways to Define Artificial Intelligence. Retrieved from https://newsroom.intel.com/news/many-ways-define-artificial-intelligence/#gs.b18dny

-Nicholson, Chris. (2020). A Beginner's Guide to Neural Networks and Deep Learning Retrieved from https://wiki.pathmind.com/neural-network

-IBM. (2020). What are neural networks?. Retrieved from https://www.ibm.com/cloud/learn/neural-networks

Comments

Post a Comment

Popular posts from this blog

Week 4 Progress -Applications - Neural Networks in Python

Image
Summary/Overview: During this week I continued to learn more about neural network algorithms and coding using previously found resources and attempted to understand them better. During this time I also read some resources on several uses of neural networks in various fields including in Geographic Information Science (GIS) applications, healthcare and image recognition. Additionally I created a template for my final product slideshow and began planning out the sections that I would go over in it. Healthcare Applications: There are many different applications of neural networks in healthcare fields I explored several from a general perspective. Some of these applications include the creation of diagnostic systems to identify issues, recreating neural systems as shown in the image below, analyzing bichemical samples, analyzing drug trial data to better develop them and the analysis of visual data captured from a wide variety of medical devices. RevolverDeBlake. (2019). "Study shows ...
Read more

Week 6 - Final Project Wrap Up

    Priya Dialani. (2019). “Developing Innovation: Neural Network and Deep Learning”. [Image]. Retrieved from https://www.analyticsinsight.net/developing-innovation-neural-network-and-deep-learning/   Well here we are in the final week of the course and at the end of the Genius Hour Project. During the last six weeks I explored Neural Networks from a Python programming perspective and theiir various applications across multiple industries. I will discuss the the various aspects of the project in the final week in the sections below.   What I Learned: During this Genius Hour Project I learned a great deal more than I knew about neural networks, specifically neural networks in Python as well as how these neural networks can be applied to various industries from a real world perspective. I explored the structure of Neural networks the process of building them and how they work, I learned what supervised vs. unsupervised training was and about Neural network algorithms....
Read more

Week 3 Progress - Specific Research - Neural Networks in Python

Image
Summary/Overview: During this weeks research I dove into several resources listed below that discussed neural networks from a Python programming perspective as well as several that discussed programming simple neural networks from a practical perspective in python. Some of these included psuedo code as well as actual coding examples. Although the actual syntax of the python coding was useful in grounding the concepts of neural networks in my previous Python experience I still struggle somewhat with some of the more advanced mathematical concepts involved in the coding process, It was useful to have a more grounded perspective however and was interesting to see just how simple the code snipets involved in creating a basic low number of layers neural network would be. I also learned more about the role kof modules in Python Neural Network coding as well as specific training datasets, where to find them and how to use them. This weeks research also helped to cement my understanding of wei...
Read more