farhanhubble

## Computes cosine similarity between two matrices.
## similarity[i][j] = A[i,:] * B[:,j] / |A[i,:]| * |B[:,j]|
def cosine_similarity(A,B):
    B_T = B.T
    
    product = np.matmul(A,B_T)

    normA = np.linalg.norm(A,axis=1)
    normA = normA.reshape(normA.size,1)

farhanhubble

void filter(VectorXd &x, MatrixXd &P) {

	for (unsigned int n = 0; n < measurements.size(); ++n) {

		VectorXd z = measurements[n];
		//YOUR CODE HERE
		
		// KF Prediction
		VectorXd x_ = F*x + u;
		MatrixXd P_ = F*P*F.transpose() + Q;

farhanhubble

import json
import asyncio
import websockets

def _check_mode(msg):
    return  "Auto" if msg and msg[:2] == '42' else "Manual"

def _parse_telemetry(msg):
    msg_json = msg[2:]
    parsed = json.loads(msg_json)

farhanhubble

TypeError                                 Traceback (most recent call last)
<ipython-input-44-a553646a889c> in <module>
    22     for i, samples in enumerate([samples_1, samples_10, samples_100]):
    23         results[clf_name][i] = \
---> 24         train_predict(clf, samples, X_train, y_train, X_test, y_test)
    25
    26 # Run metrics visualization for the three supervised learning models chosen

<ipython-input-36-2fd67492e721> in train_predict(learner, sample_size, X_train, y_train, X_test, y_test)
    19     # TODO: Fit the learner to the training data using slicing with ‘sample_size’ using .fit(training_features[:], training_labels[:])

farhanhubble

world =  [['▢', '▣', '▢'],
          ['▢', '▣', '▢'],
          ['▢', '▣', '▢']]


## Code to update the world one step.

nb_rows = len(world)
nb_cols = len(world[0])

farhanhubble

1. List the space used by the root parition and its directories excluding other paritions 1

du -h -x -d 1 /

2. Get rid of trash in common cleanable folders as described here
3. If the /var directory has a huge size try cleaning up Packagekit cache

sudo pkcon refresh force -c -1

farhanhubble

1. Install the correct drivers from within the system (Software): 495.46 as of Jan 1,2022. Compatible Devices
2. Reboot
3. Check the latest CUDA version available here
4. Install an appropriate version of CUDA using the instructions here.
   • The CUDA installer has bundled Nvidia drivers. Deselect these drivers if the driver version installed earlier is higher.
5. The installer will recommend adding certain directories to the PATH and LD_LIBRARY_PATH. Edit the ~/.bashrc to add the paths.
6. Run source ~/.bashrc to pull the changes in the currect terminal.
7. Reboot [Optional].
8. Go to the samples directory (Will be inside a folder named something like NVIDIA_CUDA-11.5_Samples that is inside the home folder,

farhanhubble

Use the entr command to watch the Notebook for changes.

ls myfile.ipynb| entr -r jupyter nbconvert myfile.ipynb --to slides --post serve --ServePostProcessor.open_in_browser=False

The URL of the slides gets printed in the terminal:

Serving your slides at http://127.0.0.1:8000/myfile.slides.html

Sources:

farhanhubble

Neural networks that perform classification(predict the class of an input) produce a vector of C raw numbers for every input, where C is the total number of classes, for example 10 for MNIST. This vector is called a logit.

If we feed an MNIST image for digit 3 to a classification model, it may produce the a logit vector like this (top row added to show the class ID's:

0	1	2	3	4	5	6	7	8	9
-121.2	-212.3	81.1	171.1	-55.0	132.5	-13.2	63.5	99.2	-10.9