Notes

In the last few months, I’ve intermittently thought about getting an Online Masters in Computer Science. And then wasted a few days mulling over it. Here are a few reasons why I think about it. Go deeper into CS in a more structured manner. The OMSCS program from Georgia Tech seems best for this: it’s both cheap and reputed. However, people also say online that it’s rigorous and it’ll require dedicated time every week. So, instead of signing up for something like that, a better approach would be to rigorously study something on my own. For e.g., I could take a goal to read up two books on Artificial Intelligence this quarter. They take so many students that I won’t be guaranteed to get into the classes I want in every semester. I am part of an industry that hardly cares about credentials. (In fact, many self-taught programmers are able to make it big.) So, such a degree won’t help me in any professional way. Get a degree for the sake of it. If so, there are two ways of going about it: Find a shitty college that doesn’t have a high bar in handing out degrees. (The college doesn’t necessarily need to be in the US.) Pursue a degree from some other field, such as quantum physics or arts. That would be more fun and challenging. I am going to drop this idea for now. And focus the little time I get away from work and parenting to studying Artificial Intelligence. I should probably take a goal of finishing the D2L book and Hands on Machine Learning this quarter.

Setup, for Backblaze Create a configuration file: cat > ~/.config/restic-backup.conf << 'EOF' BACKUP_PATHS=/media/unencrypted_drive/SharedFolder BACKUP_EXCLUDES='--exclude-caches --exclude /media/unencrypted_drive/SharedFolder/... --exclude /media/unencrypted_drive/SharedFolder/... --exclude "._*"' RETENTION_DAYS=14 RETENTION_WEEKS=8 RETENTION_MONTHS=6 B2_ACCOUNT_ID=... B2_ACCOUNT_KEY=... RESTIC_REPOSITORY=b2:restic-personal-backup-1:home-hard-drive RESTIC_PASSWORD=... EOF To source these as environment variables, simply do: set -a source ~/.config/restic-backup.conf set +a # To unset. Setup: restic init mkdir -p ~/.config/systemd/user/ cat > ~/.config/systemd/user/restic-backup.service << 'EOF' [Unit] Description=Restic backup service [Service] Type=oneshot ExecStart=restic backup --verbose $BACKUP_EXCLUDES $BACKUP_PATHS ExecStartPost=restic forget --verbose --keep-daily $RETENTION_DAYS --keep-weekly $RETENTION_WEEKS --keep-monthly $RETENTION_MONTHS --prune EnvironmentFile=%h/.config/restic-backup.conf EOF cat > ~/.config/systemd/user/restic-backup.timer << 'EOF' [Unit] Description=Backup with restic daily [Timer] OnCalendar=daily Persistent=true [Install] WantedBy=timers.target EOF systemctl --user daemon-reload systemctl --user enable --now restic-backup.timer Then: ...

import os import sys module_path = os.path.abspath(os.path.join('whatever_path')) if module_path not in sys.path: sys.path.append(module_path) Copied from here.

Cuda versions: pip install torch torchvision \ mxnet-cu102 \ pandas \ scikit-learn \ jupyterlab \ matplotlib \ ipykernel \ boto3

Install docker run \ -d \ -v /opt/wallabag/data:/var/www/wallabag/data \ -v /opt/wallabag/images:/var/www/wallabag/web/assets/images \ -e SYMFONY__ENV__DOMAIN_NAME="https://wallabag.flister.cf" \ -e SYMFONY__ENV__FOSUSER_REGISTRATION=false \ --restart unless-stopped \ -p 8091:80 wallabag/wallabag # Setup TLS sudo certbot --nginx -d wallabag.flister.cf Add the following to the relevant section of the Nginx config file: location / { proxy_pass http://localhost:8091; proxy_set_header X-Forwarded-Host $server_name; proxy_set_header X-Forwarded-Proto https; proxy_set_header X-Forwarded-For $remote_addr; } Note: Detailed instructions here. The domain in SYMFONY__ENV__DOMAIN_NAME is the one you use to access the website. So, if you access via some other IP address or domain name, CSS styles won’t load. Upgrade Upgrade Docker image by following this. If needed, run database migration by following this.

Setup Installation guide: https://docs.docker.com/engine/install/ubuntu/ No sudo: Check this to allow Docker containers to run as non-root. See what’s running docker ps Container-specific commands # SSH into a running container. docker exec -it 0061ecc7ef37 /bin/bash # Restart docker restart 52835f0907e8 # Stop docker container stop 23642fbe8515 # Stop all docker stop $(docker ps -a -q) # Prune all stopped containers docker container prune Parameters for docker run -d: Run as daemon. -p 8439:8080: Map port A externally to B internally. -v ~/Documents/TiddlyWikiDockerVolume:/var/lib/tiddlywiki: Map volume A externally to B internally. -e USERNAME="abc": Environment variable. --restart unless-stopped: Automatically bring up the container on Docker daemon restart. Compose # Assuming current directory contains a file named *docker-compose.yml* docker-compose up # Or docker-compose down Uprade an existing service The following assumes that the containers are stateless. ...

Deep Learning: https://www.fast.ai http://deeplearning.net/ http://neuralnetworksanddeeplearning.com/index.html https://d2l.ai Also, see this for resources on practicing these skills.

Project ideas: https://github.com/NirantK/awesome-project-ideas Datasets: http://academictorrents.com/ https://www.dataquest.io/blog/free-datasets-for-projects/

Bag of Tricks for Image Classification with Convolutional Neural Networks: This paper reviews a few tricks that have been proposed in the past for image classification using CNNs and proposes a few of its own. It then documents how those tricks improve training time or model accuracy. I found the following interesting: Increasing the batch size during training will decrease the variance (and, therefore, noise) of stochastic gradient, so we can increase learning rate and, therefore, reduce training time. Learning rates should be kept low initially, otherwise they can lead to numerical instability. Apply weight decay only to weights, not to biases and $\gamma$ and $\beta$ in Batch Norm layers. You can use 16-bit floating point precision to decrease training time. Cosine decay for learning rate. I didn’t know about the teacher student model, where we make the student imitate the teacher. This is different from transfer learning where we use one model as the starting point to train another. Mixup training: An augmentation technique that randomly chooses two samples and blends them together with a weighted linear interpolation. I couldn’t imagine something like that would work. I didn’t fully understand: Label smoothing.

DeepAnT: A Deep Learning Approach forUnsupervised Anomaly Detectionin Time Series: It’s a long paper but, at it’s core, it presents a straightforward idea. Use CNN as a forecasting model to predict data for a given timestamp. Calculate an anomaly score by comparing the prediction against the observed value and call the time-series anomalous (for that timestamp) if the score is higher than some threshold. I found two problems with their proposal of calculating anomaly scores: It doesn’t take variance into account and only uses absolute difference. If the latter is fixed, shouldn’t anomaly score be inversely proportional to variance? How do I interpret that score in business terminology? They used Mean Absolute Error as the loss function for the model. Why not Mean Squared Error? The idea of choosing time-series specific thresholds would be tricky but the paper doesn’t clarify how to automate that at scale.