ºÝºÝߣ

ºÝºÝߣShare a Scribd company logo

Sequence Hacking Pitch at Data Science Hackathon

•Download as PPTX, PDF•
•
A
ams345

This document discusses sequences and their importance in learning. It provides examples of sequences without memory to predict the next element. The document also includes musical sequences, such as a Bach invention and lyrics from Eminem's "Lose Yourself". Finally, it lists flexibility, autonomy and efficiency as important concepts.

1 of 16
Download to read offline
Sequence Hackin’ Rian Shams 02/14/2016
What is a sequence? A collection of ordered elements Why are sequences important? Learning occurs sequentially
A B C A B C A ? Goal: Predict next element in this sequence Assume: No prior memory
A B C A B C A B Goal: Predict next element in this sequence Assume: No prior memory
AA B C A B C B Past Present Future
AA B C A B C B Past Present Future Sequence Memory
A B C D X B C Y A B C D X B C Y A B C D X B C ? Goal: Predict next element in this sequence Assume: No prior memory
A B C D X B C Y A B C D X B C Y A B C D X B C Y Goal: Predict next element in this sequence Assume: No prior memory
E J E G E G J I F J B J E D A D D B I I C B F D C B J J F F D G I F A J J C F H G I F C F G C G G F E J E G E G J I F J B J E D A D D B I I C B F D C B J J F F D G I F A J J C F H G I F C F G C G G F E J E G E G J I F J B J E D A D D B I I C B F D C B J J F F D G I F A J J C F H G I F C F G C G G F E J E G ?
MUSIC
Johann Sebastian Bach 1685 – 1750
Invention 1
Marshall Bruce Mathers III (aka Eminem)
His palms are sweaty, knees weak, arms are heavy There's vomit on his sweater already, mom's spaghetti He's nervous, but on the surface he looks calm and ready to drop bombs, But he keeps on forgetting what he wrote down, The whole crowd goes so loud He opens his mouth, but the words won't come out He's choking how, everybody's joking now The clock's run out, time's up, over, bloah! Snap back to reality, Oh there goes gravity Oh, there goes Rabbit, he choked He's so mad, but he won't give up that Easy, no He won't have it, he knows his whole back's to these ropes It don't matter, he's dope He knows that but he's broke He's so stagnant, he knows When he goes back to his mobile home, that's when it's Back to the lab again, yo This whole rhapsody He better go capture this moment and hope it don't pass him You better lose yourself in the music, the moment You own it, you better never let it go (go) You only get one shot, do not miss your chance to blow This opportunity comes once in a lifetime yo You better lose yourself in the music, the moment You own it, you better never let it go (go) You only get one shot, do not miss your chance to blow This opportunity comes once in a lifetime yo Lose Yourself
FLEXIBILITY AUTONOMY EFFICENCY
Rian Shams rian.shams@gmail.com Contact Info

More Related Content

Sequence Hacking Pitch at Data Science Hackathon

  • 2. What is a sequence? A collection of ordered elements Why are sequences important? Learning occurs sequentially
  • 3. A B C A B C A ? Goal: Predict next element in this sequence Assume: No prior memory
  • 4. A B C A B C A B Goal: Predict next element in this sequence Assume: No prior memory
  • 5. AA B C A B C B Past Present Future
  • 6. AA B C A B C B Past Present Future Sequence Memory
  • 7. A B C D X B C Y A B C D X B C Y A B C D X B C ? Goal: Predict next element in this sequence Assume: No prior memory
  • 8. A B C D X B C Y A B C D X B C Y A B C D X B C Y Goal: Predict next element in this sequence Assume: No prior memory
  • 9. E J E G E G J I F J B J E D A D D B I I C B F D C B J J F F D G I F A J J C F H G I F C F G C G G F E J E G E G J I F J B J E D A D D B I I C B F D C B J J F F D G I F A J J C F H G I F C F G C G G F E J E G E G J I F J B J E D A D D B I I C B F D C B J J F F D G I F A J J C F H G I F C F G C G G F E J E G ?
  • 10. MUSIC
  • 11. Johann Sebastian Bach 1685 – 1750
  • 13. Marshall Bruce Mathers III (aka Eminem)
  • 14. His palms are sweaty, knees weak, arms are heavy There's vomit on his sweater already, mom's spaghetti He's nervous, but on the surface he looks calm and ready to drop bombs, But he keeps on forgetting what he wrote down, The whole crowd goes so loud He opens his mouth, but the words won't come out He's choking how, everybody's joking now The clock's run out, time's up, over, bloah! Snap back to reality, Oh there goes gravity Oh, there goes Rabbit, he choked He's so mad, but he won't give up that Easy, no He won't have it, he knows his whole back's to these ropes It don't matter, he's dope He knows that but he's broke He's so stagnant, he knows When he goes back to his mobile home, that's when it's Back to the lab again, yo This whole rhapsody He better go capture this moment and hope it don't pass him You better lose yourself in the music, the moment You own it, you better never let it go (go) You only get one shot, do not miss your chance to blow This opportunity comes once in a lifetime yo You better lose yourself in the music, the moment You own it, you better never let it go (go) You only get one shot, do not miss your chance to blow This opportunity comes once in a lifetime yo Lose Yourself

Editor's Notes

  • #2: My name is Rian And for my demo we are going to be doing some sequence hacking.
  • #3: Sequence A sequence is A collection of Ordered elements. Numbers are sequences. Words are sequences. My speech is a sequence. Sequences are important Because learning occurs sequentially. Think about anything you have ever learned; Counting, writing, programming, walking, singing, dancing, brushing your teeth, getting dressed These are all sequences That involve a start, an end, and a sequence of steps in between.
  • #4: Take a look at this sequence. A B C A B C A ? If I were to ask you to predict the next element in this sequence, assuming no prior memory, what would you predict? No prior memory -> no knowledge of what the alphabet and what these symbols represent.
  • #5: No prior memory -> no knowledge of what the alphabet and what these symbols represent. Cut to the demo here
  • #6: Why do we predict B? We use our knowledge of the present (in this case, the blue A) and our knowledge of the past (the repetitive ABC structure) To generate a prediction about what the future might look like (B). If we were to expose Sequence Memory to this same sequence and also ask it to predict the next element in the sequence we would also expect it to predict B because Sequence memory learns to predict the future based on the past and present. Lets test this out. (demo) During Demo: -Pre-run The programming language I am using for this demo is called Racket. Racket is a dialect of Lisp, which was traditionally used for AI applications. Lisp is known for its use of parentheses. The input is ABCABCA. The name of the function that will operate on this sequence to make predictions is sequence memory. The display function displays the output. -Post-run The first column is the input data (raw values) The second column is the predicted output (expected values) Each row is input and output at a timestep. First when this sequence memory algorithm is exposed to A it doesn’t make a prediction Because A is a novel input it has never been exposed to before. The same is true for B and C, But when Sequence Memory is exposed to A for the second time it predicts B -Conclusion What we have learned from this example is that Sequence Memory can learn to make predictions from a short, simple sequence.
  • #7: Why do we predict B? We use our knowledge of the present (in this case, the blue A) and our knowledge of the past (the repetitive ABC structure) To generate a prediction about what the future might look like (B). If we were to expose Sequence Memory to this same sequence and also ask it to predict the next element in the sequence we would also expect it to predict B because Sequence memory learns to predict the future based on the past and present. Lets test this out. (demo) During Demo: -Pre-run The programming language I am using for this demo is called Racket. Racket is a dialect of Lisp, which was traditionally used for AI applications. Lisp is known for its use of parentheses. The input is ABCABCA. The name of the function that will operate on this sequence to make predictions is sequence memory. The display function displays the output. -Post-run The first column is the input data (raw values) The second column is the predicted output (expected values) Each row is input and output at a timestep. First when this sequence memory algorithm is exposed to A it doesn’t make a prediction Because A is a novel input it has never been exposed to before. The same is true for B and C, But when Sequence Memory is exposed to A for the second time it predicts B -Conclusion What we have learned from this example is that Sequence Memory can learn to make predictions from a short, simple sequence.
  • #8: No prior memory -> no knowledge of what the alphabet and what these symbols represent.
  • #9: No prior memory -> no knowledge of what the alphabet and what these symbols represent.
  • #10: Lets move on to the next example Which is even longer & even more complicated. If I were to show you this sequence: (point to sequence) and ask you To predict the next element in this sequence based on the present element which in this case is G (in blue), and the entire past which in this case is everything that precedes the blue G What would you predict? If it is at all possible to make a prediction From this input sequence This example is more challenging that the last couple of examples. But if I were to show you this differently
  • #12: Johann Sebastian Bach is one of the greatest composers and pianists ever.
  • #13: Before Demo This melody called Invention 1 is one of his shorter pieces. What I am going to show now is the SAME Sequence Memory algorithm Learning to play this song. All I am doing is feeding in data Corresponding to What the note is (C, D, E, B-flat) The duration of the note (how long it’s pressed) And where in the sequence the note occurs. After Demo This is remarkable because the same algorithm that just a few slides ago was learning 123’s and ABC’s, that started with no prior knowledge, Or in computer science lingo as an empty list, Has learned the sequence of these notes Learned the overall structure of the song (where it starts and ends) After only being exposed to the song, or hearing the song twice, In like 20 seconds. I used to play the piano, In fact I was classically trained for about 10 years And in my first 20 seconds of learning to play the piano I could not play this. No one can, not even the great Bach To play a song like this well takes at least 3-4 years of study if you are good and practice a lot. Re-Emphasize So I want to re-emphasize that this is the same model, This same exact algorithm used in past examples, So now lets move on from one legendary musician, to another…
  • #14: Marshall Bruce Mathers III Aka Eminem Eminem is considered to be one of the greatest lyricists and rappers Of all time. For those of you familiar with Eminem and his music Know that one of his most famous songs is called ….
  • #15: Before Demo Lose Yourself. These are the lyrics to Lose Yourself. I wanted to find out if I could feed these lyrics into Sequence Memory Algorithm and have it learn to rap this song Lose Yourself. After Demo This is also very impressive because After being exposed to this sequence of lyrics Just twice, it learned it almost perfectly. Another nice feature about Sequence Memory Is that unlike a biological cortex It won’t ever forget a sequence unless you tell it to.
  • #16: Flexible Sequence Learning Sequence Memory should be used because it is flexible sequence learning meaning it can learn an arbitrary amount of sequences of arbitrary type, arbitrary length, And arbitrary context (assuming your computer meets the required memory capacity). Autonomous Sequence Learning Sequence Memory is autonomous sequence learning Meaning it can learn sequences with little to no human supervision. This is unsupervised machine learning. Robust Sequence Learning Sequence Memory is robust sequence learning, meaning it is fault tolerant and can handle noisy data well.