Assignment 5

In this article, the author talked about how memory is an increasingly important topic in the field of cognitive psychology. Recently, substantial progress has been made to attempt to create human like memory in deep learning systems. This process introduces key questions that are important to keep in mind. They are: What makes memory such a complex subject in deep learning systems? What are the main techniques used to represent memories in deep learning models? Where can we draw inspiration about memory architectures? 

Neuroscience and cognitive psychology are among the fields of psychology that have had the most influence on our knowledge of memory. Neuroscience mainly focuses on how memories are created, destroyed, and stored as either short term or long term. One case study that provided a huge breakthrough in this field was Henry Gustav Molaison, or patient HM. He suffered an accident when he was young that ultimately caused him to have seizures on a regular basis. When he became an adult, he went through surgery that aimed to cure his symptoms. Unfortunately, the surgeons accidentally cut a part of his hippocampus, which caused HM to be unable to create new memories. Post-surgery, he was only able to retain information for a few minutes and became unable to form new long-term memories. This helped lead to a better understanding of how memories are stored, created, and recalled. 

A substantial challenge related to the aspect of the neuroscience theory of memory is the binding problem. The binding problem is difficulty scientists have understanding how small fragments of memories can be put together by the mind to form a complete memory. All memories of different events or things are stored in small parts in different areas of the brain, and only one que can cause you to remember the entirety of the event. Although researchers are still unsure of the exact reason for this, a possible explanation is that parts of a full memory are linked by electromagnetic vibrations. These would then cause a link between memory fragments; which would activate them together when recalling a memory.  

To better understand this binding problem, we need to consider other psychological factors that impact memory recollection. A leading theory that explains associations in memory is called the priming effect. The priming effect is one of the most significant explanations for associations in memory. It states that being primed, even with one word, can trigger a flood of memories associated with the word. For example, if you hear the word “college” you may immediately think about classes, exams, or partying. The memories or events that one remembers through priming differs based on personal experience. 

This article was very intriguing to me.  Ever since I was young, I have always wondered how we formed memories and what science was behind it. I am not shocked that researchers still struggle to explain the binding problem. The concept of understanding it completely seems incredibly challenging. The information provided in this article ties in very well to what we have been learning in class and aided a lot in helping to connect aspects of the last couple chapters to real-life research.  

Citation: 

Rodriguez, J. (2018, July 25). Understanding Memory in Deep Learning Systems: The Neuroscience, Psychology and Technology? Retrieved from https://towardsdatascience.com/understanding-memory-in-deep-learning-systems-the-neuroscience-psychology-and-technology-78922ae6a1dd