Stephen Smith's Blog

Musings on Machine Learning…

Archive for August 2019

Raspberry Pi 4 as a Desktop Computer

leave a comment »


The Raspberry Pi Foundation is promoting the Raspberry Pi 4 as a full desktop computer for only $35. I’ve had my Raspberry Pi 4 for about a month now and in this article we’ll discuss if it really is a full desktop computer replacement. This partly depends on what you use your desktop computer for. My answer is that the $35 price is misleading, you need to add quite a few other things to make it work well.

Making the Raspberry Pi 4 into a Decent Desktop

The Raspberry Pi has always been a barebones computer. You’ve always needed to add a case, a keyboard, a mouse, a monitor, a power supply, a video cable and a microSD card. Many people already have these kicking around, so they don’t need to buy them when they get their Pi. For instance, I already had a keyboard and monitor. The Raspberry Pi 4 even supports two monitors.

Beyond the bare bones, you need two more things for a decent desktop, namely:

  1. The 4GB version of the Raspberry Pi 4
  2. A good USB SSD drive

With these, it starts to feel like you are playing with a regular desktop computer. You now have enough RAM to run multiple programs and any good SSD will greatly enhance the performance of thee system, only using the microSD card to boot the Pi.

The Raspberry Pi 3 is a great little computer. Its main limitation is that if you run too many programs or open too many browser tabs, it bogs down and you have a painful process of closing windows (that aren’t responding well), until things pick up again. Now the Raspberry Pi 4 with 4GB of RAM really opens up the number of things you can do at once. Running multiple browser tabs, LibreOffice and a programming IDE are no problem.

The next thing you run into with the Raspberry Pi 4 is the performance of the SD card. Since I needed a video cable and a new case, I ordered a package deal that also included a microSD card containing Raspbian. Sadly, these bundled microSD cards are the cheapest, and hence slowest available. Having Raspbian bundled on a slow card is just a waste. Switching to a Sandisk Extreme 64GB made a huge difference. The speed was much better. When buying a microSD card watch the speed ratings, often the bigger cards (64GB or better) are twice as fast as the smaller cards (32GB or less). With a good microSD card the Raspberry Pi 4 can read and write microSD twice as fast as a Raspberry Pi 3.

I’ve never felt I could truly trust running off a microSD card. I’ve never had one fail, but people report problems all the time. Further, the performance of microSD cards is only a fraction of what you can get from good SSDs. The Raspberry Pi 4 comes with two USB 3 ports which have a theoretical performance ten times that of the microSD port. If you shop around you will find M.2 and SATA SSDs for prices less than those of microSD cards. I purchased a Kingston A1000 M.2 drive which was on sale cheap because the A2000 cards just started shipping. I had to get an M.2 USB caddy to contain it, but combined this was less than $100 and USB caddies are always useful.

Unfortunately, you can’t boot the Raspberry Pi 4 directly off a USB port yet. The Raspberry Pi foundation say this is coming, but not quite here yet. What you can do is have the entire root file system on the USB drive, but the boot partition must be on a microSD card. Setting up the SSD was easier than I thought it would be. I had to partition it, format it, copy everything over to the SSD and then edit /boot/config.txt to say where the root of the main file system is.

With this done, I feel like I’m using a real desktop computer. I’m confident my data is being stored reliably, the performance is great.


The Raspberry Pi 4 uses more power than previous Pis. This means there is more heat to dissipate. The case I received with my Pi 4 didn’t have any ventilation holes and would get quite hot. I solved the problem by removing the top of the case. This let enough heat out that I could run fine for most things. People report that when using a USB SSD that the USB controller chip will overheat and the data throughput will be throttled. I haven’t run into this, but it is something to be aware of.

I installed Tensorflow, Google’s open source AI toolkit. Training a data model with Tensorflow does make my Pi 4 overheat. I suspect Tensorflow is keeping all four CPU cores busy and producing a maximum amount of heat. This might drive me to add a cooling fan. I like the way the Pi runs so quietly, with no fan, it makes no noise. I might try using a small fan blowing down on the Pi to see is that helps.


Is the Raspberry Pi 4 a complete desktop computer for $35? No. But if you get the 4GB model for $55 and then add a USB 3 SSD, then you do have a good workable desktop computer. The CPU power of the Pi has been compared to a typical 2012 desktop computer. But for the cost that is pretty good. I suspect the Wifi/Lan and SSD are quite a bit better than that 2012 computer.

Keep in mind the Raspberry Pi runs Linux, which isn’t for everyone. A typical low cost Windows desktop goes for around $500 these days. You can get a refurbished one for $200-$300. A refurbished desktop can be a good inexpensive option.

I like the Raspberry Pi, partly because you are cleanly out of the WinTel world. No Windows, no Intel. The processor is ARM and the operating system is Raspbian based on Debian Linux. A lot of things you do are DIY, but I enjoy that. With over 25 million Raspberry Pis sold worldwide, there is a lot of community support and you join quite an enthusiastic thriving group.


Written by smist08

August 26, 2019 at 8:17 pm

Playing the Kaggle Two Sigma Challenge 2018/2019

leave a comment »



A couple of years ago, I entered a Kaggle data science competition sponsored by Two Sigma for stock market prediction. I blogged about this in part 1, part 2, part 3, part 4 and part 5. The upshot of this was that although I put in a lot of work, I performed quite poorly in the final stages. I learned a lot about machine learning and data science along the way and was keen to have another go, when another Two Sigma sponsored competition rolled around starting last year.

In this competition we had three months to create our models in the fall of 2018, then they used our models to predict the stock market through the first half of 2019. With my learnings from the last competition, I was able to do much better this time around.

Don’t Overfit

My big lesson from the first competition was to not overfit the model to the training data. This is equivalent to having ten data points and fitting them perfectly with a 9th degree polynomial. There is no error in predicting the ten points, but the model is useless at predicting anything else, and in fact gives about the worst predictions possible.

A more subtle form of overfitting is trying hundreds of models and fiddling with their parameters until they work really well with the training data. This is a lot of work and won’t help you on any data outside of the training data. I did this for the first competition, it was a lot of work, and it performed badly.

Avoiding overfitting means doing less work, which is good. I spent very little time on this competition and got quite good results.

Kaggle Virtual Environment

One of the things I like about this competition is that you play it in Google/Kaggle’s virtual environment. You have a fixed set of computer resources and everyone plays in the same environment. This levels the playing field with people who have access to very high powered equipment at corporations or Universities. This year the environment included a GPU and we could run for six hours on a high powered server.

This does limit the models you can use, I wasn’t successful at using a neural networks, probably because historical stock market data is very flat and it is hard to get these models to converge. I ended up using an Extra Trees model in SciKit Learn.

Make the Program Robust

Usually in this sort of competition, when you run your model, behind the scenes Kaggle runs it on the secret test data as well, so if you run successfully on the provided test data, you know you also run on the secret data you are going to be scored against. In this case the secret test data didn’t exist yet. This led to the worry that sometime in the six months that they would be running our program, something unexpected would appear in the data and cause my program to crash, knocking me out of the competition.

I was careful to put in try/catch statements and added extra checks to try and keep my program running. The other thing with Python programs, is that sometimes they work, but throw a memory exception when they shutdown. I spent some time tracking down a number of these bugs and made sure my program could exit gracefully without any errors.

From the message board for the competition, it appears quite a few competitors were knocked out during the run on the new data.

Don’t Cheat

Some people spend all their time trying to cheat. Trying to hack the system to gain access to the secret test data. For the purposes of the leaderboard, there was secret test data to give people a score during the model building phase, but this data wouldn’t be used in the real competition. There was no real protection on this data, since using it to cheat would be useless. However quite a few people did cheat to move to the top of the leaderboard before the real competition started. These programs crashed when the real competition started.

There are rumours that people have succeeded in other Kaggle competitions by cheating, but in this one, since it was based on stock market data generated after the models were frozen, it wasn’t going to work.

My Model

The intent of the competition was to try to use news data to enhance your stock prediction algorithm. I don’t think this worked well, I used an Extra Trees variation on a Random Forest, and the news data never seemed to contribute much. Many other competitors didn’t even use it. The competition metric was a Sharpe Ratio. A couple of observations about the Sharpe Ratio, one is that it has the standard deviation of the estimates in the denominator. This means volatile stocks will hurt you even if you predict them accurately. Second, you enter a confidence value on whether you think the stock will do well and this confidence can be negative if you think it will go down. If you get the sign wrong it will doubly hurt you.

Given the Sharpe Ratio as the metric, I decided to sort the stocks by standard deviation of their returns and then rate any with a high standard deviation as zero, which would exclude them from the model. This meant I was building a portfolio of a subset of all the stocks present. I wanted stocks I could predict accurately and that had a lower volatility.

In reading the discussion board after the competition, it appears many of the top performers took this approach.


I enjoyed this competition. I liked working with the high powered servers in the competition’s virtual environment. The lesson learned to not overfit, greatly reduced the amount of work. Whenever I was tempted to tune my model, I just said no. I ended up receiving a silver medal, coming in 145 out of 2927 competitors. With each competition I learn a bit more and I look forward to the next one.

Written by smist08

August 16, 2019 at 5:43 pm

Low Cost Linux Notebooks

leave a comment »


Theoretically, a notebook running Linux should be inexpensive, since you don’t need a Windows license and Linux runs well without premium hardware. In reality, buying a Linux notebook tends to be expensive on premium hardware. There are companies like Purism and System76 that produce Linux only laptops but these are high-end expensive. Similarly, companies like Dell seem to charge extra if you want Linux. In this article we’ll look at some options for running Linux inexpensively. We’ll look at the tradeoffs, including privacy and security.

Used, Refurbished or Discounted Windows Notebooks

Windows Notebooks have the advantage of mass-production and competition. There are tons of companies producing Windows notebooks. You can find great deals on sale, plus there is a huge market of refurbished lease returns that offer great deals. Also, companies take returns from retailers like Amazon, make sure they are ok and then sell them at a big discount. You then need to install your favorite Linux distribution and then you are up and running. You can even set it up so you can dual boot either Linux or Windows.

If you are concerned about privacy and security, then the downside of Windows notebooks is that they run the UEFI BIOS. This BIOS has backdoors built in so the NSA, and probably other governments, can remotely take control of your computer.

All that being said, if a notebook runs Windows well, it will run Linux better. A great way to bring an old slow laptop or notebook back to life, is to wipe Windows and replace it with Linux. I’m writing this on an old HP laptop which became slower and slower running Windows 10. Now with Ubuntu Linux, it runs great. No more Windows bitrot and it has a whole new life.


Even cheaper than Windows notebooks, are Chromebooks. These are notebooks designed to run Google’s ChromeOS. These notebooks are cheaper because they don’t require a Windows license and they usually don’t include a harddrive. Instead of a harddrive they have a small memory card usually 16Gig or 32Gig. Chrome OS is based on a Linux kernel, but restricts you in a few ways. You need to sign on using a Google ID, then you install Apps (basically Android apps) via the Google Play store.

Earlier versions couldn’t run regular Linux apps; however, Google has been relaxing this and now allows you to install and run many Linux apps and run a terminal window. Over time Chrome OS has been slowly morphing into full Linux. From being just a portal to Google’s web apps to being a full client operating system. However, I find Chrome OS is still too limiting and there is the issue of having to sign on with Google.

Out of the box, you can’t just install Linux on a Chromebook. The BIOS is locked to only running Chrome OS. The BIOS in Chromebooks is based on Coreboot the open source which is good, however they modified it without providing the source code, so we don’t know if they added hooks for the NSA to spy on you. The Google BIOS does provide a developer mode, this developer mode gives you a root access terminal session and allows you to install and run flavours of Linux from inside Chrome OS using a set of shell scripts called crouton. Many people prefer this method as they get both Linux and Chrome OS at the same time.

Upgrade the BIOS

If you want to boot directly into an alternate OS, you usually need to upgrade the Chromebook’s BIOS to allow this. I bought an inexpensive refurbished Dell Chromebook 11 off Amazon for $100 (CAD). There are two ways to do this, one is reversible, the other isn’t and you run the risk of bricking your device. The Dell’s BIOS is divided into two parts, one is upgradable, and can be reversed using a recovery USB stick. The other requires disassembling the notebook, removing a BIOS write protect tab and then burning the whole BIOS.

I went the reversible route. I made a recovery USB stick and upgraded the BIOS to support booting other operating systems. This isn’t perfect as you are still using Google’s unknown BIOS and you have to hit control-L everytime you boot to run your alternate operating system.

The reason people will risk replacing their whole BIOS is to get a pure version of Coreboot that hasn’t been tampered with by Google. You then have full control of your computer, no developer mode and no control-L to boot. Perhaps one day I’ll give this a try.

Once you have your BIOS updated, you can install Linux from a USB stick. I chose to install GalliumOS, which is tailored for Chromebooks. It installs a minimal Linux, since it knows Chromebooks don’t have much disk space. It also includes all the drivers needed for typical Chromebook trackpads, bluetooth and Wifi. The Gallium OS website has great information, with links to how to upgrade your BIOS and otherwise prepare and complete a successful upgrade.

Another choice is LUbuntu (Light Ubuntu), which is Ubuntu Linux optimized for low memory hardware. I didn’t like this distro as much, probably because it is so optimized for low memory, whereas I have 4GB memory, it is disk space I’m short of (only 16GB). So I didn’t really need the low memory desktop, and would have preferred LibreOffice being left out.

A great source of info on updating Chromebook BIOS’s is MrChromebox. Its interesting because they also have lots of information on how to install UEFI BIOS on a Chromebook, so you can use it as a cheap Windows notebook. You could install UEFI and then run Linux, but why would you want to? Unless you want to be helpful to the NSA and other government spy agencies.


Sadly, running Linux on a converted Windows notebook gives the better experience. At this point, despite the privacy concerns, the UEFI BIOS works better with Linux than Coreboot. On the Chromebook, besides the nuisance of having to hit control-L every time it boots, I found some things just didn’t work well. The main problem I had was closing and opening the lid on the notebook, that Linux’s suspend function didn’t work properly. Often when I opened the lid, Linux didn’t unsuspend and I’d have to do a hard power off- power on which then resulted in a disk corruption scan.  Otherwise bluetooth, wifi and the trackpad work fine.

I also think the small memory cards are a problem. I think you’re better off booting from a regular SSD hard drive. These are inexpensive and give you way more space with better performance. I wish there was a cheap Chromebook with an M.2 interface. Or even one where the memory card isn’t glued to the motherboard and in an accessible location.

I really want an inexpensive notebook with privacy and security. The best option right now is to convert a Chromebook over to full Coreboot and then run a privacy oriented version of Linux like PureOS, but right now this is quite a DIY project.


Written by smist08

August 9, 2019 at 6:46 pm

Posted in Business

Tagged with , , , , , ,

Raspberry Pi 4 First Impressions

leave a comment »


I’ve received my Raspberry Pi 4B with 4GB or RAM a few weeks ago. I’ve been using it to write my forthcoming book on Raspberry Pi Assembly Language Programming, so I thought I’d give a few of my first impressions. The biggest change for the Raspberry Pi 4 is the support for three memory sizes, 1GB, 2GB and 4GB. This overcomes the biggest complaint against the Raspberry Pi 3, that it bogs down too quickly as you run browser tabs and multiple windows.

Some of the other hardware improvements are:

  • Dual 4K monitor support with dual micro-HDMI ports.
  • Two of the four USB ports are USB-3.
  • The ethernet is now gigabit and the WiFi is faster.
  • A 1.5GHz quad-core 64-bit ARM Cortex-A72 CPU.
  • The SDRAM is now LPDDR4.
  • The GPU is upgraded to Broadcom’s VideoCore VI.
  • Hardware HEVC video support for 4Kp60 video.

On paper, this makes the Raspberry Pi 4 appear far superior to its predecessors, In this article, I’ll discuss what is much better and a few of the drawbacks. This release will squash a lot of the compatible Pi competitors, but I’ll compare it to my NVidia Jetson Nano and mention a few places where these products still surpass the Pi.

Raspbian Buster

At the same time the Raspberry Pi Foundation released the Raspberry Pi 4, they also released the new “Buster” version of Raspbian, the Debian Linux derived operating system tailored specifically to the Raspberry Pi. On the day this was announced, I ordered my Raspberry Pi 4, then went and downloaded the new Buster release, then installed it on my Raspberry Pi 3B.

If you have a Raspberry Pi 4, then you must run the Buster release as older versions of Raspbian don’t have support for the newer hardware. If you are running an older Pi then you can keep running the older version or upgrade as you like.

Is it 64-Bits?

The first rumour that was squashed was that Raspbian would move to 64-bits. This didn’t happen. Raspbian is a 32-bit operating system. The Raspberry Pi Foundation says it will stay this way for the foreseeable future. The first reason is that the Raspberry Pi 1 and Raspberry Pi Zero use a much older ARM processor that doesn’t support 64-bits. The Raspberry Pi Foundation still supports and sells these models and they are quite popular due to their low price. They don’t want to support two operating systems, so they stick to one 32-bit version that will run on every Raspberry Pi ever made. Perhaps other hardware vendors should look at this level of support for older models.

Even though 32-bit implies a 32-bit virtual address space for processes, which limits an individual process to 4GB of memory, the ARM SoC used in the Pi has memory access hardware for 48-bit addresses. This allows the operating system to give each process a different 4GB address space, so if Raspberry Pi models with more than 4GB of memory are released, Raspbian can utilize this memory.

Another problem with going to 64-bits is that all the previous Raspberry Pi models, and one version of the Raspberry Pi 4 only have 1GB of RAM. This isn’t sufficient to run a 64-bit operating system. You can do it, but the operating system takes all the RAM, and once you run a program or two, everything bogs down. This is due to all addresses and most integers becoming 64-bits, and hence twice as large. A definite nice feature of Raspbian is that it can run effectively in only 1GB or memory.

Based on Debian Buster

Raspbian is notorious for lagging behind the mainstream releases of Linux. The benefit of this is that Raspbian has always been very stable and reliable. It works well and avoids the problems that happen at the bleeding edge. The downside is that it can contain security vulnerabilities or bugs that are fixed in the newer versions.

With Buster, Raspbian released its version ahead of Debian releasing the main version. Linus Torvalds himself was involved in moving the Pi up to a newer version of the Linux kernel. His concern is that as other hardware platforms adopt proprietary software like UEFI firmware, with government mandated backdoors, that the benefits of open source are being lost. The Raspberry Pi, including its firmware are all open source and there is a feeling in the open source community that this is the future to fight for.

Some Software Not Ported Yet

As a result of the move to Buster, some software that Raspberry users are accustomed to is missing. The most notable case is Mathematica. A port of this is underway and it is promised to be included in a future upgrade.

I had problems with CubicSDR, a Software Defined Radio (SDR) program. It could detect my SDR USB device, but didn’t run properly, just displaying a blank screen when receiving.

Heat Dissipation

The Raspberry Pi 4 uses more power than previous models. It requires a USB-C power adapter which means you can’t use a power adapter from previous models. I bought my Pi 4 from and got the bundle with a case, power adapter, heat sinks and micro-HDMI cable. I needed the cables. The case is their Raspberry Pi 3 case, with the holes for the cables moved for the slightly different Pi 4 configuration. The case lacked any ventilation holes and the Pi would throttle due to overheating fairly easily. My solution was to run it with the top of the case removed. This seems to provide enough air circulation that I haven’t seen any overheating since.

Some people claim the Raspberry Pi 4 requires a fan for cooling, but that hasn’t been my experience. I think the cases need properly thought out ventilation and that is all that is needed. I think a bigger heatsink like the one included with the NVidia Jetson Nano would be warranted as well. I don’t like fans and consider the quietness of the Pi as one of its biggest features.


All this sounds great, but what are the downsides of the Raspberry Pi 4?

All New Cables

I purchased an NVidia Jetson Nano and to run it, I just unplugged the cables from my Raspberry Pi 3 and plugged them into the Jetson and away it went. Not new cables required.

The Raspberry Pi required a new USB-C power supply and a lot has been made of how you can’t use Apple laptop power supplies, but I think the real issue is you can’t use an older Pi power supply, even if it can provide sufficient power.

To support dual monitors, the Pi went to micro-HDMI ports to fit both connectors. This means you need either new cables or at least micro- to regular-HDMI adapters. The NVidia Jetson supports dual monitors but annoyingly with two different cables, HDMI and a DisplayPort cable. At least the cables are the same for the two video ports.

Otherwise all my USB devices that I was using with the Raspberry Pi 3 seem to work with the Pi 4.

SDCard Bottleneck

They have improved the data transfer speed to and from the microSD card with the Pi 4, but this is still a bottleneck. I would have loved it if they had added a M.2 SSD interface to the board. You can improve on the microSD card speed by using a USB 3 external SSD. The problem is that you can’t boot from this USB 3 drive. You can copy the root filesystem over to the drive and run mostly from the USB and although I haven’t tried it yet, people report this is an improvement.

Raspberry Pi promote the 4 as a desktop computer replacement and it definitely has the processing power. However, I don’t think this really holds up without something better than running off a microSD card. The Raspberry Pi Foundation say they will add boot from USB support in a future firmware update, but it isn’t there yet. Although the speed of USB 3 is better than the microSD interface, it still isn’t nearly as good as you can obtain with M.2 and a good new SSD.

No 64-Bits Yet

The Raspberry Pi Foundation, caught everyone by surprise with their release. This included the people that maintain alternate operating systems for the Raspberry Pi. There is a good Ubuntu Mate 64-bit version that runs on the Raspberry Pi 3. It is slow and you can’t run many programs, but it does work and you can experiment with things like ARM 64-bit Assembly programming.

The person that maintains this had to order his Raspberry Pi 4, like everyone else and hasn’t produced a Pi 4 version yet. It would have been nice if the Raspberry Pi Foundation had seeded some early models to the people that develop alternate operating systems for the Pi.

As of this writing, Raspbian is the only operating system that runs on the Raspberry Pi, but hopefully the others won’t take too long to modify what they need to.

The Raspberry Pi 4 with 4GB is the first Raspberry Pi that has the power to run a true 64-bit operating system, so it would be nice to play with.


The Raspberry Pi 4 is still dirt cheap, $35 for the 1GB model and $55 for the 4Gig model. This upgrade is a bit more expensive since you need a new power adapter, new video cables and a new case as well. I think the extra $20 for the extra memory is well worth it.

Compared to the NVidia Jetson Nano

The Raspberry Pi 4 blows most of the current crop of Pi clones out of the water. One exception is the NVidia Jetson Nano. This single board computer has 4GB of memory and runs full 64-bit Ubuntu Linux and as a consequence feels more powerful than the Pi 4.

The Pi 4 has a more powerful ARM CPU, but the Jetson has 4 USB-C ports and a 128 core CUDA GPU. The CUDA GPU is used by software like CubicSDR for DSP like processing, along with most AI toolkits like Tensorflow.

The NVidia Jetson costs $99, so is nearly twice as expensive as a Pi 4. However if you want to experiment with AI, the 128-core CUDA GPU is an excellent entry level system. 


I got used to the Raspberry Pi 4 fairly quickly and after a couple of weeks thought it was pretty similar to the Raspberry Pi 3. I then needed to do something on my Raspberry 3 and booted it up. After using the Pi 4, going back to the Pi 3, felt like I was working in molasses, everything was so slow. This is a real testament to how good the new Pi is, especially with 4GB of memory.

Yes, there are some teething problems with the new model, as there often is at the bleeding edge. But overall the Raspberry Pi 4 is a solid upgrade, and once you adopt it, you really can’t go back. 


Written by smist08

August 2, 2019 at 7:09 pm