Glue

F* the algo.

[Glue @isGLUE]

This used to be the thing. Today all LLMs are terribly overfitted.

[Glue @isGLUE]

When a machine learning model is trained on a lot of features, it becomes overfitted and more and more reliant on the training data, defeating the purpose by performing poorly on real data.

Avoiding overfitting is a major motivation for performing dimensionality reduction.

[Glue @isGLUE]

But all outdated now. No vibes.

[Glue @isGLUE]

Wonderfully explained!

[The Innovation Materials @Innov_Materials]

We keep hearing about Artificial Intelligence, Machine Learning, Neural Network, and Deep Learning. Do you know their relationship?
https://t.co/qA8rufKH8M

[Glue @isGLUE]

@debayan Interns are usually happier

[Naval @naval]

The Internet only tolerates complete centralization and complete decentralization.

Everything in between gets destroyed.

[Glue @isGLUE]

@debayan Happy birthday!

[Glue @isGLUE]

1. Boosts your confidence

[Glue @isGLUE]

Know + Do, that’s how you can get the ball rolling!📈

[Glue @isGLUE]

The fewer features our training data has, the fewer assumptions our model makes and the simpler it will be.

Dimensionality reduction has a lot more advantages to offer, like

1. Fewer dimensions mean less computing. Less data means that algorithms train faster.

[Glue @isGLUE]

1. AdaDelta Optimizer: It can be understood as a more robust version of the AdaGrad Optimizer. It was created to address the significant shortcomings of AdaGrad and the RMS prop optimizer and is based on adaptive learning.

[Glue @isGLUE]

RMSprop is a gradient-based optimization technique used in training neural networks. It was proposed by the father of back-propagation, Geoffrey Hinton.

[Glue @isGLUE]

1. Mini Batch Gradient Descent Optimizer: In this variation of gradient descent, only a portion of the dataset is used to calculate the loss function rather than the entire training set.

[Glue @isGLUE]

Stochastic Gradient Descent with Momentum Optimizer: We know that stochastic gradient descent takes a much more noisy path than the gradient descent algorithm.

[Glue @isGLUE]

1. Gradient Descent Optimizer: It is the most basic but most commonly used optimization algorithm.

It is widely used in linear regression and classification algorithms. The gradient descent algorithm is also used in neural network backpropagation.

[Glue @isGLUE]

But what are multi-class classification problems?

Modern machine learning problems involve complex tasks where one has to make the optimal decision given hundreds or thousands of possible actions.

[Glue @isGLUE]

It involves dividing the dataset into different and unique classes based on different parameters so that a new and unseen record can be put into one of the classes.

[Glue @isGLUE]

Mean Absolute Error Loss (MAE): It is defined as the #average of the absolute difference between the actual and predicted values.

It is also known as the L1 loss function.