Artificial Neuron In Practice

Part 1: The Underlying Mechanics

Problem 1: Linear Score Computation ( $z = w \cdot x + b$ ) Given a 3-dimensional input feature vector $x = \begin{bmatrix} 2.0 \\ -1.0 \\ 3.0 \end{bmatrix}$ , a weight vector $w = \begin{bmatrix} 1.5 \\ 4.0 \\ -0.5 \end{bmatrix}$ , and a baseline bias $b = -2.5$ , calculate the linear score $z = w \cdot x + b$ .

Reveal Solution

\begin{aligned} &\mathbf{w \cdot x} = (1.5)(2.0) + (4.0)(-1.0) + (-0.5)(3.0) = 3.0 - 4.0 - 1.5 = \mathbf{-2.5} \\ &\mathbf{z = (w \cdot x) + b} = -2.5 + (-2.5) = \mathbf{-5.0} \\ &\text{Final linear score: } \mathbf{z = -5.0} \end{aligned}

Problem 2: Linear Score to Sigmoid Activation

A neuron has weight vector $w = \begin{bmatrix} 2.0 \\ -1.0 \end{bmatrix}$ and baseline bias $b = -3.0$ .

Part A: Calculate the linear score $z = w \cdot x + b$ for the input feature vector $x = \begin{bmatrix} 3.0 \\ 1.0 \end{bmatrix}$ .
Part B: Using the Sigmoid formula $\sigma(z) = \frac{1}{1 + e^{-z}}$ and the reference table below, calculate the neuron's output probability $\sigma(z)$ : $\text{Reference: } e^{-1.0} \approx 0.3679, \quad e^{-2.0} \approx 0.1353, \quad e^{-3.0} \approx 0.0498$

Reveal Solution

Part A:

\begin{aligned} &\mathbf{w \cdot x} = (2.0)(3.0) + (-1.0)(1.0) = 6.0 - 1.0 = 5.0 \\ &\mathbf{z = (w \cdot x) + b} = 5.0 + (-3.0) = \mathbf{2.0} \end{aligned}

Part B:

\mathbf{\sigma(2.0)} = \frac{1}{1 + e^{-2.0}} = \frac{1}{1 + 0.1353} = \frac{1}{1.1353} \approx \mathbf{0.8808} \quad (\mathbf{88.1\%})

Final linear score: $z = \mathbf{2.0}$ , Sigmoid output: $\sigma(z) \approx \mathbf{88.1\%}$

Problem 3: Baseline Bias and the Decision Threshold

A neuron receives input feature vector $x = \begin{bmatrix} 3.0 \\ 4.0 \end{bmatrix}$ with weight vector $w = \begin{bmatrix} 2.0 \\ 1.5 \end{bmatrix}$ .

Part A: Calculate the raw dot product $w \cdot x$ .
Part B: What exact baseline bias $b$ places the neuron on the decision threshold ( $z = 0$ , yielding $\sigma(z) = 0.50$ )?
Part C: To produce an output above the decision threshold ( $\sigma(z) > 0.50$ ), must the chosen bias $b$ be greater than or less than the value found in Part B?

Reveal Solution

Part A:

\mathbf{w \cdot x} = (2.0)(3.0) + (1.5)(4.0) = 6.0 + 6.0 = \mathbf{12.0}

Part B: To place the neuron on the decision threshold ( $z = 0$ ):

\mathbf{z = w \cdot x + b = 0} \implies 12.0 + b = 0 \implies \mathbf{b = -12.0}

Part C: Because Sigmoid is strictly increasing, an output above the decision threshold ( $\sigma(z) > 0.50$ ) requires $z > 0$ :

\mathbf{w \cdot x + b > 0} \implies 12.0 + b > 0 \implies \mathbf{b > -12.0}

The bias must be greater than $-12.0$ (such as $-11.0$ or $0.0$ ). Any bias greater than $-12.0$ results in a positive linear score ( $z > 0$ ), yielding $\sigma(z) > 0.50$ .

Problem 4: Complete Artificial Neuron Pipeline

A binary classification neuron evaluates input feature vector $x = \begin{bmatrix} 3.0 \\ 2.0 \end{bmatrix}$ with weight vector $w = \begin{bmatrix} 1.0 \\ -1.0 \end{bmatrix}$ and baseline bias $b = -2.0$ .

Part A: Calculate the raw dot product $w \cdot x$ .
Part B: Calculate the linear score $z = w \cdot x + b$ .
Part C: Using the Sigmoid formula $\sigma(z) = \frac{1}{1 + e^{-z}}$ and the reference table below, calculate the neuron's output probability $\sigma(z)$ : $\text{Reference: } e^{0.0} = 1.0000, \quad e^{1.0} \approx 2.7183, \quad e^{2.0} \approx 7.3891$
Part D: State the binary classification output ( $1$ or $0$ ) using the standard decision threshold of $0.50$ .

Reveal Solution

Part A:

\mathbf{w \cdot x} = (1.0)(3.0) + (-1.0)(2.0) = 3.0 - 2.0 = \mathbf{1.0}

Part B:

\mathbf{z = (w \cdot x) + b} = 1.0 + (-2.0) = \mathbf{-1.0}

Part C:

\mathbf{\sigma(-1.0)} = \frac{1}{1 + e^{-(-1.0)}} = \frac{1}{1 + e^{1.0}} = \frac{1}{1 + 2.7183} = \frac{1}{3.7183} \approx \mathbf{0.2689} \quad (\mathbf{26.9\%})

Part D:

Because $\sigma(z) \approx 0.2689 < 0.50$ (corresponding to $z < 0$ ), the neuron outputs $0$ (Inactive / Negative Class).

This traces the complete 4-stage artificial neuron pipeline:

Stage 1 (Feature weighting & dot product): $w \cdot x = 1.0$
Stage 2 (Baseline bias offset): $z = 1.0 + (-2.0) = -1.0$
Stage 3 (Sigmoid compression): $\sigma(-1.0) \approx 0.2689$ ( $26.9\%$ )
Stage 4 (Decision threshold): $26.9\% < 50\% \implies \mathbf{0}$

Part 2: Applied Scenario: The VC Investment Predictor

In Topic 1, two startups submitted pitch decks to a Venture Capital firm evaluating Unicorn Potential across four features [Team Experience, Market Size, Competition, Risk]:

x_{\text{OmniFlow}} = \begin{bmatrix} 1.0 \\ 0.3 \\ 1.0 \\ 0.2 \end{bmatrix} \qquad\qquad x_{\text{Solaris}} = \begin{bmatrix} 0.3 \\ 1.0 \\ 0.0 \\ 0.8 \end{bmatrix}

The Venture Capital firm evaluates Unicorn Potential using the weight vector $w$ :

w = \begin{bmatrix} 4.0 \\ 2.0 \\ -3.0 \\ 1.0 \end{bmatrix} \begin{matrix} \text{(Team Experience Weight)} \\ \text{(Market Size Weight)} \\ \text{(Competition Weight)} \\ \text{(Risk Weight)} \end{matrix}

In Topic 1, their raw dot products were:

$w \cdot x_{\text{OmniFlow}} = (4.0)(1.0) + (2.0)(0.3) + (-3.0)(1.0) + (1.0)(0.2) = \mathbf{+1.8}$
$w \cdot x_{\text{Solaris}} = (4.0)(0.3) + (2.0)(1.0) + (-3.0)(0.0) + (1.0)(0.8) = \mathbf{+4.0}$

Now, macroeconomic conditions change. The broader venture capital market enters a downturn marked by rising interest rates and conservative capital deployment. To reflect higher investment standards, the investment committee introduces a baseline bias:

b = -2.5

To pass initial screening, the firm requires a Sigmoid probability of at least $50\%$ ( $\sigma(z) \ge 0.50$ , corresponding to $z \ge 0$ ) to classify a startup as High Potential.

Problem 5: Startup Screening Under Macro Downturn

Part A: Compute the linear score $z = w \cdot x + b$ for both OmniFlow and Solaris AI.
Part B: Using the Sigmoid formula $\sigma(z) = \frac{1}{1 + e^{-z}}$ and the reference table below, calculate the screening probability $\sigma(z)$ for both startups: $\text{Reference: } e^{-1.5} \approx 0.2231, \quad e^{-0.7} \approx 0.4966, \quad e^{0.7} \approx 2.0138, \quad e^{1.5} \approx 4.4817$
Part C: State which startup qualifies as High Potential ( $\sigma(z) \ge 0.50$ ), and explain how the negative bias ( $b = -2.5$ ) altered the outcome for OmniFlow compared to its raw dot product in Topic 1.

Reveal Solution

Part A: Linear Score Computation

For OmniFlow:

\mathbf{z_{\text{OmniFlow}}} = (w \cdot x_{\text{OmniFlow}}) + b = 1.8 + (-2.5) = \mathbf{-0.7}

For Solaris AI:

\mathbf{z_{\text{Solaris}}} = (w \cdot x_{\text{Solaris}}) + b = 4.0 + (-2.5) = \mathbf{+1.5}

Part B: Sigmoid Probability

For OmniFlow ( $z = -0.7$ ):

\mathbf{\sigma(-0.7)} = \frac{1}{1 + e^{-(-0.7)}} = \frac{1}{1 + e^{0.7}} = \frac{1}{1 + 2.0138} = \frac{1}{3.0138} \approx \mathbf{0.3318} \quad (\mathbf{33.2\%})

For Solaris AI ( $z = +1.5$ ):

\mathbf{\sigma(+1.5)} = \frac{1}{1 + e^{-1.5}} = \frac{1}{1 + 0.2231} = \frac{1}{1.2231} \approx \mathbf{0.8176} \quad (\mathbf{81.8\%})

Part C: Decision & Analysis

Solaris AI qualifies as a High Potential candidate with a probability of $81.8\%$ ( $\sigma(z) \ge 0.50$ ). Its strong dot product ( $+4.0$ ) absorbs the negative bias ( $b = -2.5$ ) and keeps the linear score positive ( $z = +1.5$ ).
OmniFlow falls into Low Potential with a probability of $33.2\%$ ( $\sigma(z) < 0.50$ ). Although its dot product was positive in Topic 1 ( $+1.8$ ), adding the negative bias pulls its linear score into negative territory ( $z = -0.7$ ), dropping its probability below the $50\%$ threshold.

Key Takeaway: The baseline bias $b$ acts as an environmental threshold on Unicorn Potential. When market conditions tighten ( $b = -2.5$ ), borderline candidates with small positive dot products get rejected ( $z < 0$ ), while high-conviction opportunities remain positive ( $z > 0$ ).