Guess the Deep Learning Technique

Riddle

(1)

Played 34 %Accuracy 95 Average time 11:28

Create copy

About this activity

Riddles about DL techniques

Created by

Thanuja Viswanadham

India

Download the paper version to play

Top 10 results

#	Player	Accuracy	time
1	Sahasra August 4, 2026	% 100	04:20
2	Anonymous August 4, 2026	% 100	07:53
3	Anonymous August 4, 2026	% 100	08:56
4	Anonymous August 4, 2026	% 100	11:18
5	Anonymous August 4, 2026	% 100	11:33
6	Anonymous August 4, 2026	% 100	11:35
7	Golla Gnaneshwari August 4, 2026	% 100	11:48
8	Anonymous August 4, 2026	% 100	12:14
9	Anonymous August 4, 2026	% 100	12:27
10	Anonymous August 4, 2026	% 100	12:45

Do you want to stay in the Top 10 of this game? Log in to identify yourself.

Make your own free game from our game creator

Compete against your friends to see who gets the best score in this game

Make challenge

Fantastic creatures

Educaplay Educational Resources

Spain

(166)

Guess them using as few clues as possible
The gold riddle

Cody Nakajima

United States

(4)

On your quest you find 3 treasure chests. 1 chest is the gold chest. You need to find the gold chest. But the wrong chests has a deadly gas. And to make matters worse all of the chests look identical....
GUESS, GUESSER...

Esteban PedrazaEnciso

Colombia

(10)

¡ I guess the tracks who it is and have fun...!
Brain Game 2

Kecia Hopper

United States

(8)

Brain teasers
Words About Words Vocabulary 6c

Ginger Jantz

United States

(13)

Words used to discuss words and writing

		%
	Anonymous	%
		%
		%

Print your game

Guess the Deep Learning Technique


Guess the Deep Learning TechniqueOnline version Riddles about DL techniques by Thanuja Viswanadham 1 Riddle 1: I like to clean up unnecessary things. I make many weights become exactly zero. I help the model focus only on the most important features. Who am I? Hints 1 It promotes sparsity 2 A type of penalty that can wipe out weights 3 Think of pruning weights to zero 2 Riddle 2: I don't remove weights completely. Instead, I gently shrink them so they do not become too large. I help the model generalize better without making the weights disappear. Who am I? Hints 1 Improves generalization 2 Shrinks weights gradually 3 A milder penalty than L1 3 Riddle 3: Whenever the model creates very large weights, I add an extra cost. The larger the weights become, the greater the penalty. I encourage simpler models that generalize well. Who am I? Hints 1 Related to weight magnitudes 2 Encourages simplicity 3 Penalizes large parameter norms 4 Riddle 4: I allow the model to learn, but only within certain limits. Instead of giving complete freedom, I make learning follow specific constraints. Think of me as learning with rules. Who am I? Hints 1 Controls learning space 2 Regularization framed as constraints 3 Use of constraints in optimization 5 Riddle 5: I know exactly when enough learning is enough. Before the model begins memorizing the training data, I stop the learning process. This helps the model perform better on unseen data. Who am I? Hints 1 Monitors progress 2 Prevents overfitting 3 Stop training based on validation performance 6 Riddle 6: Some of my examples come with answers, while many do not. Even with only a few labeled examples, I can still learn effectively. I make use of both labeled and unlabeled data. Who am I? Hints 1 Combines labeled and unlabeled data 2 Works with limited labels 3 Uses unlabeled data 7 Riddle 7: Why create new weights everywhere when the same ones can do the job? I reuse parameters across different parts of the model, making it smaller and more efficient. Who am I? Hints 1 Promotes efficiency 2 Reduces model size 3 Share weights across layers or parts 8 Riddle 8: During training, I randomly ask some neurons to sit out for a while. Every neuron learns to work independently, reducing overfitting. Who am I? Hints 1 Improves generalization 2 Prevents co-adaptation 3 Randomly disable units during training 9 Riddle 9: I prepare the model for tricky situations by training it with carefully crafted deceptive examples. Because of me, attackers find it much harder to fool the model. Who am I? Hints 1 Tests model against attacks 2 Improves robustness 3 Train with adversarial examples 10 Riddle 10: Two images may look slightly different because one is shifted, rotated, or scaled. Instead of treating them as different, I measure their similarity by considering these small transformations. Who am I? Hints 1 Geometric distance concept 2 Metrics that handle small changes 3 Transformation-aware similarity

You can integrate the game into an LMS compatible with LTI 1.1 or LTI 1.3 such as Canvas, Moodle, or Blackboard. This way, the scores will be automatically saved into the platform’s gradebook.

Guess the Deep Learning Technique

Riddle

About this activity

Created by

Download the paper version to play

Top Games

Riddle

Fantastic creatures

Riddle

The gold riddle

Riddle

GUESS, GUESSER...

Riddle

Brain Game 2

Riddle

Words About Words Vocabulary 6c

Guess the Deep Learning TechniqueOnline version

by Thanuja Viswanadham

Riddle 1: I like to clean up unnecessary things. I make many weights become exactly zero. I help the model focus only on the most important features. Who am I?

Hints

Riddle 2: I don't remove weights completely. Instead, I gently shrink them so they do not become too large. I help the model generalize better without making the weights disappear. Who am I?

Hints

Riddle 3: Whenever the model creates very large weights, I add an extra cost. The larger the weights become, the greater the penalty. I encourage simpler models that generalize well. Who am I?

Hints

Riddle 4: I allow the model to learn, but only within certain limits. Instead of giving complete freedom, I make learning follow specific constraints. Think of me as learning with rules. Who am I?

Hints

Riddle 5: I know exactly when enough learning is enough. Before the model begins memorizing the training data, I stop the learning process. This helps the model perform better on unseen data. Who am I?

Hints

Riddle 6: Some of my examples come with answers, while many do not. Even with only a few labeled examples, I can still learn effectively. I make use of both labeled and unlabeled data. Who am I?

Hints

Riddle 7: Why create new weights everywhere when the same ones can do the job? I reuse parameters across different parts of the model, making it smaller and more efficient. Who am I?

Hints

Riddle 8: During training, I randomly ask some neurons to sit out for a while. Every neuron learns to work independently, reducing overfitting. Who am I?

Hints

Riddle 9: I prepare the model for tricky situations by training it with carefully crafted deceptive examples. Because of me, attackers find it much harder to fool the model. Who am I?

Hints

Riddle 10: Two images may look slightly different because one is shifted, rotated, or scaled. Instead of treating them as different, I measure their similarity by considering these small transformations. Who am I?

Hints