An educational crypto, finance, and AI vocabulary puzzle about wallet signatures, business accounting, neural network training settings, and evaluation data.
Wallet Key Basics
These concepts distinguish secret signing material from public verification information.
- Private Key: A private key is secret cryptographic material used to create digital signatures.
- Public Key: A public key is cryptographic information that others can use to verify signatures made with the corresponding private key.
- Seed Phrase: A seed phrase is a sequence of words used by compatible wallets to derive and recover keys.
- Digital Signature: A digital signature allows verification that a message was signed with a particular private key and has not been altered.
Balance Sheet Basics
The accounting equation relates assets to liabilities and equity at a particular date.
AssetLiabilityEquityRetained Earnings
- Asset: An asset is a resource controlled by a business from which future economic benefits are expected.
- Liability: A liability is a present obligation of a business to transfer economic resources.
- Equity: Equity is the residual interest in assets after deducting liabilities.
- Retained Earnings: Retained earnings are accumulated profits less accumulated losses and distributions to owners, reported within equity.
Training Hyperparameters
These settings govern update size, examples per update, data passes, and a common regularization method.
Learning RateBatch SizeEpoch CountDropout Rate
- Learning Rate: The learning rate scales the parameter updates made by an optimization algorithm during training.
- Batch Size: Batch size specifies how many training examples are processed together for a training step.
- Epoch Count: Epoch count specifies how many complete passes a training process makes through the training dataset.
- Dropout Rate: The dropout rate specifies the probability of disabling an activation during training when dropout regularization is used.
Dataset Partitions
Separating data by purpose, time, or entity helps prevent evaluation results from benefiting from leaked information.
Training SetValidation SetTest SetTemporal Split
- Training Set: A training set contains the examples used to fit a model during learning.
- Validation Set: A validation set is held apart from parameter fitting and used to compare models or tune hyperparameters.
- Test Set: A test set is reserved for evaluating a chosen model and should not guide training or hyperparameter selection.
- Temporal Split: A temporal split separates data by time, commonly using earlier observations for training and later observations for evaluation.