Technology
Artificial Intelligence
Computer Science
Image Processing
Speech Recognition
Algebra
Weather Prediction
One of the biggest open problems in computer science is the complexity of matrix multiplications
Alpha Tensor is the first AI system for discovering the fastest way to multiply two matrices
Here's all you need to know about Deepminds recent Alpha Tensor:
Alpha Tensor is the first AI system for discovering the fastest way to multiply two matrices
Here's all you need to know about Deepminds recent Alpha Tensor:
We all know matrix multiplication is one of the simplest operations in algebra.
But this is fundamental for image processing, speech recognization, weather prediction, and many other applications
But this is fundamental for image processing, speech recognization, weather prediction, and many other applications
So, even significantly minor improvements to the efficiency of matrix multiplication can have a huge impact and saves a lot of time and money
AlphaTensor is an extension of AlphaZero:
We all know AlphaZero which is a reinforcement learning agent that outperformed humans in board games like chess, Go, and shogi
Checkout here: deepmind.com
We all know AlphaZero which is a reinforcement learning agent that outperformed humans in board games like chess, Go, and shogi
Checkout here: deepmind.com
The researchers trained a new version of AlphaZero.
Alpha Tensor started without any knowledge of existing matrix multiplication algorithms
Alpha Tensor started without any knowledge of existing matrix multiplication algorithms
Through learning, AlphaTensor gradually improved over time, re-discovering historical fast matrix multiplication
Finally, it came out with the best by surpassing the previously efficient Algorithms
checkout this blog by deepmind:
deepmind.com
Finally, it came out with the best by surpassing the previously efficient Algorithms
checkout this blog by deepmind:
deepmind.com
Finally for Comparision:
Number of steps needed to multiply two 9x9 matrices is 511 by the current existing algorithms.
The Alpha Tensor reduced that to 498.
Number of steps needed to multiply two 9x9 matrices is 511 by the current existing algorithms.
The Alpha Tensor reduced that to 498.
That's all for this thread. If you find this useful:
1. Leave your thoughts below
2. Follow me @Sumanth_077 if you don't want to miss content on Python, Data Science and Machine Learning
1. Leave your thoughts below
2. Follow me @Sumanth_077 if you don't want to miss content on Python, Data Science and Machine Learning
Loading suggestions...