How bad is your memory?
We all know that one person, who religiously forgets, or even better, we might be that person. This is my journey from 'I don't remember, I have a bad memory' to 'I remember that, we talked over it for a minute in a tech huddle about resource points, 3 months ago.' [Just a little bit of exaggeration, but still close]
Initially I started with a structured storage. Since the data I was consuming, was unstructured, the first step was to structure the data. This worked for simple things like Grocery lists, appointments etc. I started taking help of external tools like planners, calendars, and note books to store my indexes. These indexes also started to work as triggers. This approach of remembering things took a lot of energy in the form of building indexes, and I had hit the same problem a database would face. Also I soon realised that the input data was never structured and it took a lot of energy to re-structure the data. Even more energy was used to build indexes. So this was also a failed attempt at boosting my memory, but there were some take aways. I had disciplined my self, and now I knew how to use indexes and triggers to remember things.
I was in dire need of a mechanism to index my memory, and neural network's capabilities to label information by forming networks, which were activated or deactivated based on the occurrence of an event was my next experiment. I started linking events with other events and memories, but it worked at certain times, and not always. I started keeping tab of events which worked in my favour. We human beings have a really good memory when it comes to remembering a negative event, this in some extreme cases, turns into a painful mental condition, like PTSD ( Post traumatic Stress Disorder).
1. I had realized that I could store infinite amount of information, but needed a way to index it
2. I could index information, but indexing took more time, as the data was unstructured.
3. I could generate indexes and labels for information by linking it to other events, but only negative events were able to generate useful indexes.
From the third point I realised, I needed a better way to label the information, as I was eventually getting pulled into a depressive way of life. I started looking for take-aways rather than events, which I could use as a label to index my memory. Every conversation or action I did, had a take-away, a learning - negative or positive, which I used as a reward point to label my memory, and it helped me both in acquiring new knowledge as well as helped me look at the world in a different light altogether.
Bonus: A Machine Learning Approach to Databases Indexes
1. Memory Castle
My first approach started with the 'Memory Castle'. My workplace was converted into my memory storage. The security desk started storing my passwords - each drawer storing a different one. The printer started storing facts and figures. Cafeteria was turned into a storage for conversations. It felt good, finally being able to remember things, but soon I started over storing things. Storing memory took time, retrieving was a nightmare. I was overwhelmed with information, and I was running short of place in my memory castle to categorically store information. I had to expand my Memory Castle, and It started taking a lot of time and energy to maintain it and eventually everything was garbled up.2. Database: Indexing Memory
With the Memory Castle experiment I had established the fact that I could store Information, I could retrieve information, so storage was not an issue, neither was retrieval. The one thing Memory Castle lacked was a proper way to index information.Initially I started with a structured storage. Since the data I was consuming, was unstructured, the first step was to structure the data. This worked for simple things like Grocery lists, appointments etc. I started taking help of external tools like planners, calendars, and note books to store my indexes. These indexes also started to work as triggers. This approach of remembering things took a lot of energy in the form of building indexes, and I had hit the same problem a database would face. Also I soon realised that the input data was never structured and it took a lot of energy to re-structure the data. Even more energy was used to build indexes. So this was also a failed attempt at boosting my memory, but there were some take aways. I had disciplined my self, and now I knew how to use indexes and triggers to remember things.
3 Neural Networks: Labelling and Classifying Information
Artificial Intelligence as a concept fascinated me, specifically Neural Networks. Neural Networks are not used to store things, neither are they used to retrieve stored information, but they are good at classifying information.I was in dire need of a mechanism to index my memory, and neural network's capabilities to label information by forming networks, which were activated or deactivated based on the occurrence of an event was my next experiment. I started linking events with other events and memories, but it worked at certain times, and not always. I started keeping tab of events which worked in my favour. We human beings have a really good memory when it comes to remembering a negative event, this in some extreme cases, turns into a painful mental condition, like PTSD ( Post traumatic Stress Disorder).
4. Reinforcement learning: Extracting rewarding labels
At this juncture, I had enough data points from my experiments,1. I had realized that I could store infinite amount of information, but needed a way to index it
2. I could index information, but indexing took more time, as the data was unstructured.
3. I could generate indexes and labels for information by linking it to other events, but only negative events were able to generate useful indexes.
From the third point I realised, I needed a better way to label the information, as I was eventually getting pulled into a depressive way of life. I started looking for take-aways rather than events, which I could use as a label to index my memory. Every conversation or action I did, had a take-away, a learning - negative or positive, which I used as a reward point to label my memory, and it helped me both in acquiring new knowledge as well as helped me look at the world in a different light altogether.
5. Conclusion
I had arrived at a strategy which would work for me always without driving me nuts, or making me depressed. I had arrived at the solution by marrying concepts from Human behavioural psychology with that of computer science.
Human memory is like Santa's bag. There is no limit to what we can store, if we know how to retrieve it efficiently.
Bonus: A Machine Learning Approach to Databases Indexes
Comments
Post a Comment