Experiment tracking (MLflow) in ML Python - Model Metrics & Evaluation

Metrics & Evaluation - Experiment tracking (MLflow)

Which metric matters for Experiment Tracking and WHY

Experiment tracking helps you keep a clear record of your model tests. The key metrics to track depend on your goal, like accuracy, loss, precision, or recall. Tracking these lets you compare models easily and pick the best one. Without tracking, you might forget which model worked best or why.

Confusion Matrix or Equivalent Visualization

While MLflow itself does not create confusion matrices, it stores metrics like True Positives (TP), False Positives (FP), True Negatives (TN), and False Negatives (FN). You can log these values and then visualize the confusion matrix externally to understand model performance.

Confusion Matrix Example:

          Predicted
          P     N
Actual P  TP=50 FN=10
       N  FP=5  TN=35

Total samples = 50 + 10 + 5 + 35 = 100
    

Precision vs Recall Tradeoff with Concrete Examples

When tracking experiments, logging both precision and recall helps you see tradeoffs. For example:

• Spam filter: High precision means fewer good emails marked as spam. You want to avoid false alarms.
• Cancer detection: High recall means catching most cancer cases, even if some false alarms happen.

MLflow lets you track these metrics side by side to choose the best balance for your needs.

What "Good" vs "Bad" Metric Values Look Like for Experiment Tracking

Good experiment tracking means:

• Consistent logging of all important metrics (accuracy, loss, precision, recall, F1).
• Clear naming so you know which model and parameters produced which results.
• Comparing metrics across runs to find improvements.

Bad tracking means missing metrics, unclear records, or no way to compare runs. This leads to confusion and wasted time.

Common Pitfalls in Experiment Tracking Metrics

• Accuracy paradox: High accuracy can be misleading if data is imbalanced. Tracking precision and recall helps avoid this.
• Data leakage: If your experiment accidentally uses future data, metrics look too good but won't work in real life.
• Overfitting indicators: Tracking training vs validation metrics helps spot if your model only memorizes training data.
• Inconsistent metric definitions: Make sure you log metrics the same way every time.

Self Check: Your model has 98% accuracy but 12% recall on fraud. Is it good?

No, this model is not good for fraud detection. Even though accuracy is high, recall is very low. This means the model misses most fraud cases, which is dangerous. For fraud, catching as many frauds as possible (high recall) is critical, even if some false alarms happen.

Key Result

Experiment tracking focuses on logging key metrics like accuracy, precision, and recall to compare models and avoid pitfalls like misleading accuracy or data leakage.

Practice

1. What is the main purpose of using MLflow in machine learning projects?

easy

A. To write machine learning algorithms from scratch

B. To visualize data without any coding

C. To track and manage machine learning experiments

D. To deploy machine learning models to mobile apps

Solution

1. Step 1: Understand MLflow's role

   MLflow is designed to help track experiments by logging parameters, metrics, and models.

2. Step 2: Compare options

   Options A, C, and D describe tasks outside MLflow's main purpose.

3. Final Answer:

   To track and manage machine learning experiments -> Option C

4. Quick Check:

   MLflow = experiment tracking [OK]

Hint: MLflow is for tracking experiments, not coding models [OK]

Common Mistakes:

• Confusing MLflow with model building libraries
• Thinking MLflow is only for visualization
• Assuming MLflow deploys models directly

2. Which MLflow function is used to start a new experiment run in Python?

easy

A. mlflow.init_experiment()

B. mlflow.start_run()

C. mlflow.create_run()

D. mlflow.begin_experiment()

Solution

1. Step 1: Recall MLflow run start syntax

   The correct function to start a run is mlflow.start_run().

2. Step 2: Check other options

   Functions like init_experiment, create_run, and begin_experiment do not exist in MLflow API.

3. Final Answer:

   mlflow.start_run() -> Option B

4. Quick Check:

   Start run = mlflow.start_run() [OK]

Hint: Use start_run() to begin tracking an experiment run [OK]

Common Mistakes:

• Using non-existent MLflow functions
• Confusing experiment and run functions
• Misspelling function names

3. What will be the output of this MLflow code snippet?

import mlflow
with mlflow.start_run():
    mlflow.log_param('alpha', 0.1)
    mlflow.log_metric('accuracy', 0.95)
    run_info = mlflow.active_run().info
print(run_info.run_id is not None)

medium

A. True

B. False

C. None

D. Error

Solution

1. Step 1: Understand mlflow.start_run() context

   The code starts a run, logs parameters and metrics, then gets the active run info.

2. Step 2: Check run_id existence

   Inside the run context, run_id is assigned and not None, so print(run_info.run_id is not None) outputs True.

3. Final Answer:

   True -> Option A

4. Quick Check:

   Active run has run_id = True [OK]

Hint: Active run always has a run_id during context [OK]

Common Mistakes:

• Assuming run_id is None before logging
• Confusing run info with experiment info
• Expecting print to show run_id string

4. Identify the error in this MLflow code snippet:

import mlflow
mlflow.log_param('learning_rate', 0.01)
with mlflow.start_run():
    mlflow.log_metric('loss', 0.2)

medium

A. log_param called outside a run context

B. start_run() missing parentheses

C. log_metric called before log_param

D. No error, code runs fine

Solution

1. Step 1: Check order of MLflow calls

   log_param is called before starting a run, which is not allowed.

2. Step 2: Validate run context requirement

   MLflow requires logging functions inside a run context started by start_run().

3. Final Answer:

   log_param called outside a run context -> Option A

4. Quick Check:

   Logging outside run = error [OK]

Hint: Always log params and metrics inside start_run() [OK]

Common Mistakes:

• Calling log_param before start_run()
• Ignoring run context for logging
• Assuming log_metric order matters

5. You want to compare two ML models' performance using MLflow. Which steps correctly describe this process?

hard

A. Only log parameters for both models, metrics are not needed for comparison

B. Log both models' metrics in the same run, then compare metrics manually

C. Use mlflow.log_artifact() to save models, no need to log metrics or parameters

D. Run each model in separate mlflow.start_run() blocks, log metrics, then use MLflow UI to compare

Solution

1. Step 1: Understand MLflow run separation

   Each model should be run and logged in its own mlflow.start_run() block to keep results separate.

2. Step 2: Use MLflow UI for comparison

   After logging metrics and parameters, MLflow UI allows easy comparison of runs.

3. Final Answer:

   Run each model in separate mlflow.start_run() blocks, log metrics, then use MLflow UI to compare -> Option D

4. Quick Check:

   Separate runs + UI = compare models [OK]

Hint: Use separate runs and MLflow UI to compare models easily [OK]

Common Mistakes:

• Logging multiple models in one run
• Skipping metric logging
• Ignoring MLflow UI for comparison